diff --git "a/checkpoint-15957/trainer_state.json" "b/checkpoint-15957/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-15957/trainer_state.json"
@@ -0,0 +1,95770 @@
+{
+  "best_metric": 0.349341481924057,
+  "best_model_checkpoint": "toxic-comments-classifier-distilbert/checkpoint-15957",
+  "epoch": 1.0,
+  "eval_steps": 1,
+  "global_step": 15957,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-08,
+      "loss": 0.6793,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-08,
+      "loss": 0.6688,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-07,
+      "loss": 0.7643,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6e-07,
+      "loss": 0.6471,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-07,
+      "loss": 0.7596,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4e-07,
+      "loss": 0.5819,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8e-07,
+      "loss": 0.6388,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2e-07,
+      "loss": 0.6856,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6e-07,
+      "loss": 0.6278,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-07,
+      "loss": 0.6342,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.3999999999999997e-07,
+      "loss": 0.6404,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.8e-07,
+      "loss": 0.667,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.2e-07,
+      "loss": 0.7599,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.6e-07,
+      "loss": 0.636,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-07,
+      "loss": 0.8157,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.4e-07,
+      "loss": 0.6085,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.800000000000001e-07,
+      "loss": 0.6541,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.2e-07,
+      "loss": 0.6366,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.599999999999999e-07,
+      "loss": 0.7558,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-07,
+      "loss": 0.6663,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.399999999999999e-07,
+      "loss": 0.7089,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.799999999999999e-07,
+      "loss": 0.6312,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.2e-07,
+      "loss": 0.8188,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.6e-07,
+      "loss": 0.6478,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-06,
+      "loss": 0.6939,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.04e-06,
+      "loss": 0.5541,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.08e-06,
+      "loss": 0.6293,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.12e-06,
+      "loss": 0.8176,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.16e-06,
+      "loss": 0.6934,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-06,
+      "loss": 0.7833,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.24e-06,
+      "loss": 0.5687,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.28e-06,
+      "loss": 0.65,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.32e-06,
+      "loss": 0.7248,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3600000000000001e-06,
+      "loss": 0.5952,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4e-06,
+      "loss": 0.7146,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.44e-06,
+      "loss": 0.7228,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.48e-06,
+      "loss": 0.5672,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5199999999999998e-06,
+      "loss": 0.6201,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5599999999999999e-06,
+      "loss": 0.6189,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6e-06,
+      "loss": 0.7471,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6399999999999998e-06,
+      "loss": 0.7528,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6799999999999998e-06,
+      "loss": 0.6409,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7199999999999998e-06,
+      "loss": 0.4936,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7599999999999999e-06,
+      "loss": 0.7155,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-06,
+      "loss": 0.7212,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.84e-06,
+      "loss": 0.6781,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8799999999999998e-06,
+      "loss": 0.735,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.92e-06,
+      "loss": 0.7923,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.96e-06,
+      "loss": 0.5498,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-06,
+      "loss": 0.7077,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999581773697745e-06,
+      "loss": 0.7924,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999163547395495e-06,
+      "loss": 0.8043,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999874532109324e-06,
+      "loss": 0.5839,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999832709479099e-06,
+      "loss": 0.5817,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9997908868488736e-06,
+      "loss": 0.5462,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9997490642186486e-06,
+      "loss": 0.5507,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999707241588423e-06,
+      "loss": 0.6739,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999665418958198e-06,
+      "loss": 0.6209,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999623596327973e-06,
+      "loss": 0.5248,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9995817736977478e-06,
+      "loss": 0.8509,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9995399510675228e-06,
+      "loss": 0.7532,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9994981284372973e-06,
+      "loss": 0.5628,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9994563058070723e-06,
+      "loss": 0.8068,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999414483176847e-06,
+      "loss": 0.4972,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9993726605466215e-06,
+      "loss": 0.7725,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9993308379163965e-06,
+      "loss": 0.5733,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999289015286171e-06,
+      "loss": 0.8389,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999247192655946e-06,
+      "loss": 0.6216,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9992053700257206e-06,
+      "loss": 0.5923,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9991635473954956e-06,
+      "loss": 0.689,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99912172476527e-06,
+      "loss": 0.4916,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999079902135045e-06,
+      "loss": 0.6997,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99903807950482e-06,
+      "loss": 0.7725,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9989962568745948e-06,
+      "loss": 0.4701,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9989544342443698e-06,
+      "loss": 0.5952,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9989126116141444e-06,
+      "loss": 0.6515,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9988707889839194e-06,
+      "loss": 0.6652,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.998828966353694e-06,
+      "loss": 0.4934,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9987871437234685e-06,
+      "loss": 0.5427,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9987453210932435e-06,
+      "loss": 0.541,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998703498463018e-06,
+      "loss": 0.6484,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998661675832793e-06,
+      "loss": 0.8256,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9986198532025676e-06,
+      "loss": 0.5326,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9985780305723426e-06,
+      "loss": 0.5462,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9985362079421172e-06,
+      "loss": 0.5973,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9984943853118922e-06,
+      "loss": 0.7662,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9984525626816672e-06,
+      "loss": 0.5384,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998410740051442e-06,
+      "loss": 0.6779,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9983689174212168e-06,
+      "loss": 0.8079,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9983270947909914e-06,
+      "loss": 0.5269,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998285272160766e-06,
+      "loss": 0.4909,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998243449530541e-06,
+      "loss": 0.6948,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9982016269003155e-06,
+      "loss": 0.7065,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9981598042700905e-06,
+      "loss": 0.5245,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.998117981639865e-06,
+      "loss": 0.4955,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99807615900964e-06,
+      "loss": 0.6721,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9980343363794147e-06,
+      "loss": 0.53,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9979925137491897e-06,
+      "loss": 0.4924,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9979506911189642e-06,
+      "loss": 0.6462,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9979088684887392e-06,
+      "loss": 0.5345,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997867045858514e-06,
+      "loss": 0.7836,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997825223228289e-06,
+      "loss": 0.6462,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997783400598064e-06,
+      "loss": 0.7954,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9977415779678384e-06,
+      "loss": 0.5075,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997699755337613e-06,
+      "loss": 0.5832,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997657932707388e-06,
+      "loss": 0.7302,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9976161100771625e-06,
+      "loss": 0.4646,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9975742874469375e-06,
+      "loss": 0.4258,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997532464816712e-06,
+      "loss": 0.5045,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997490642186487e-06,
+      "loss": 0.6629,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9974488195562617e-06,
+      "loss": 0.7698,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9974069969260367e-06,
+      "loss": 0.6642,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9973651742958112e-06,
+      "loss": 0.7067,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9973233516655862e-06,
+      "loss": 0.427,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997281529035361e-06,
+      "loss": 0.674,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997239706405136e-06,
+      "loss": 0.6592,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997197883774911e-06,
+      "loss": 0.4563,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9971560611446854e-06,
+      "loss": 0.5904,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99711423851446e-06,
+      "loss": 0.6402,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.997072415884235e-06,
+      "loss": 0.5897,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9970305932540095e-06,
+      "loss": 0.7099,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9969887706237845e-06,
+      "loss": 0.6952,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996946947993559e-06,
+      "loss": 0.6342,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996905125363334e-06,
+      "loss": 0.4268,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9968633027331087e-06,
+      "loss": 0.6161,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9968214801028837e-06,
+      "loss": 0.5237,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9967796574726582e-06,
+      "loss": 0.7058,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9967378348424332e-06,
+      "loss": 0.4822,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996696012212208e-06,
+      "loss": 0.5344,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996654189581983e-06,
+      "loss": 0.4367,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996612366951758e-06,
+      "loss": 0.7173,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9965705443215324e-06,
+      "loss": 0.4674,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996528721691307e-06,
+      "loss": 0.5897,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996486899061082e-06,
+      "loss": 0.7567,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9964450764308565e-06,
+      "loss": 0.814,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9964032538006315e-06,
+      "loss": 0.652,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996361431170406e-06,
+      "loss": 0.7104,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996319608540181e-06,
+      "loss": 0.4999,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9962777859099557e-06,
+      "loss": 0.4892,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9962359632797307e-06,
+      "loss": 0.4667,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9961941406495053e-06,
+      "loss": 0.468,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9961523180192803e-06,
+      "loss": 0.5118,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996110495389055e-06,
+      "loss": 0.4548,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99606867275883e-06,
+      "loss": 0.7331,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9960268501286044e-06,
+      "loss": 0.5749,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9959850274983794e-06,
+      "loss": 0.4405,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995943204868154e-06,
+      "loss": 0.6228,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995901382237929e-06,
+      "loss": 0.4475,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9958595596077035e-06,
+      "loss": 0.6611,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9958177369774785e-06,
+      "loss": 0.6619,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995775914347253e-06,
+      "loss": 0.6722,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995734091717028e-06,
+      "loss": 0.5435,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9956922690868027e-06,
+      "loss": 0.6018,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9956504464565777e-06,
+      "loss": 0.5654,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9956086238263523e-06,
+      "loss": 0.4206,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9955668011961273e-06,
+      "loss": 0.4861,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995524978565902e-06,
+      "loss": 0.7115,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9954831559356764e-06,
+      "loss": 0.4452,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9954413333054514e-06,
+      "loss": 0.425,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9953995106752264e-06,
+      "loss": 0.7467,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995357688045001e-06,
+      "loss": 0.4172,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995315865414776e-06,
+      "loss": 0.5168,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9952740427845506e-06,
+      "loss": 0.7538,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9952322201543256e-06,
+      "loss": 1.0364,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9951903975241e-06,
+      "loss": 0.3586,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995148574893875e-06,
+      "loss": 0.4306,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9951067522636497e-06,
+      "loss": 0.6341,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9950649296334247e-06,
+      "loss": 0.415,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9950231070031993e-06,
+      "loss": 0.7821,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9949812843729743e-06,
+      "loss": 0.3926,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994939461742749e-06,
+      "loss": 0.4738,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9948976391125234e-06,
+      "loss": 0.4014,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9948558164822984e-06,
+      "loss": 0.8513,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994813993852073e-06,
+      "loss": 0.6297,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994772171221848e-06,
+      "loss": 0.7603,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994730348591623e-06,
+      "loss": 0.597,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9946885259613976e-06,
+      "loss": 0.3594,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9946467033311726e-06,
+      "loss": 0.8393,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994604880700947e-06,
+      "loss": 0.7756,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994563058070722e-06,
+      "loss": 0.8128,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9945212354404967e-06,
+      "loss": 0.6317,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9944794128102717e-06,
+      "loss": 1.0091,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9944375901800463e-06,
+      "loss": 0.5845,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9943957675498213e-06,
+      "loss": 0.8334,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994353944919596e-06,
+      "loss": 0.7911,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9943121222893704e-06,
+      "loss": 0.4609,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9942702996591454e-06,
+      "loss": 0.7137,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99422847702892e-06,
+      "loss": 0.6176,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994186654398695e-06,
+      "loss": 0.3867,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99414483176847e-06,
+      "loss": 0.4571,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9941030091382446e-06,
+      "loss": 0.3631,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9940611865080196e-06,
+      "loss": 0.4359,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994019363877794e-06,
+      "loss": 0.7032,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993977541247569e-06,
+      "loss": 0.709,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9939357186173437e-06,
+      "loss": 0.6797,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9938938959871187e-06,
+      "loss": 0.4552,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9938520733568933e-06,
+      "loss": 0.722,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9938102507266683e-06,
+      "loss": 0.7539,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993768428096443e-06,
+      "loss": 0.5624,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9937266054662174e-06,
+      "loss": 0.6515,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9936847828359924e-06,
+      "loss": 0.3294,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993642960205767e-06,
+      "loss": 0.4264,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993601137575542e-06,
+      "loss": 0.6066,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993559314945317e-06,
+      "loss": 0.4513,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9935174923150916e-06,
+      "loss": 0.6352,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9934756696848666e-06,
+      "loss": 0.4183,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993433847054641e-06,
+      "loss": 0.4106,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993392024424416e-06,
+      "loss": 0.6595,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9933502017941907e-06,
+      "loss": 0.6596,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9933083791639657e-06,
+      "loss": 0.9047,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9932665565337403e-06,
+      "loss": 0.3732,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9932247339035153e-06,
+      "loss": 0.4632,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99318291127329e-06,
+      "loss": 0.577,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9931410886430644e-06,
+      "loss": 0.6872,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9930992660128394e-06,
+      "loss": 0.4642,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993057443382614e-06,
+      "loss": 0.4271,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993015620752389e-06,
+      "loss": 0.3634,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9929737981221636e-06,
+      "loss": 0.3998,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9929319754919386e-06,
+      "loss": 0.4666,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9928901528617136e-06,
+      "loss": 0.8933,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992848330231488e-06,
+      "loss": 0.6456,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992806507601263e-06,
+      "loss": 0.4017,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9927646849710377e-06,
+      "loss": 0.3651,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9927228623408127e-06,
+      "loss": 0.4249,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9926810397105873e-06,
+      "loss": 0.3031,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992639217080362e-06,
+      "loss": 0.5684,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992597394450137e-06,
+      "loss": 0.4761,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9925555718199115e-06,
+      "loss": 0.5051,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9925137491896865e-06,
+      "loss": 0.4947,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992471926559461e-06,
+      "loss": 0.7377,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992430103929236e-06,
+      "loss": 0.9503,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9923882812990106e-06,
+      "loss": 0.3373,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9923464586687856e-06,
+      "loss": 0.7766,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9923046360385606e-06,
+      "loss": 0.3814,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992262813408335e-06,
+      "loss": 0.2854,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99222099077811e-06,
+      "loss": 0.3337,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9921791681478847e-06,
+      "loss": 0.3996,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9921373455176597e-06,
+      "loss": 0.7006,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9920955228874343e-06,
+      "loss": 0.3185,
+      "step": 239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.992053700257209e-06,
+      "loss": 0.3566,
+      "step": 240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.992011877626984e-06,
+      "loss": 0.5419,
+      "step": 241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9919700549967585e-06,
+      "loss": 0.3844,
+      "step": 242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9919282323665335e-06,
+      "loss": 0.3206,
+      "step": 243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991886409736308e-06,
+      "loss": 0.6189,
+      "step": 244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991844587106083e-06,
+      "loss": 0.28,
+      "step": 245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9918027644758576e-06,
+      "loss": 0.4334,
+      "step": 246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9917609418456326e-06,
+      "loss": 1.0138,
+      "step": 247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9917191192154076e-06,
+      "loss": 0.7887,
+      "step": 248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991677296585182e-06,
+      "loss": 0.2823,
+      "step": 249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991635473954957e-06,
+      "loss": 1.3859,
+      "step": 250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9915936513247317e-06,
+      "loss": 0.7676,
+      "step": 251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9915518286945067e-06,
+      "loss": 0.2817,
+      "step": 252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9915100060642813e-06,
+      "loss": 0.7203,
+      "step": 253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991468183434056e-06,
+      "loss": 0.2482,
+      "step": 254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991426360803831e-06,
+      "loss": 0.4828,
+      "step": 255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9913845381736055e-06,
+      "loss": 0.6771,
+      "step": 256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9913427155433805e-06,
+      "loss": 0.2292,
+      "step": 257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991300892913155e-06,
+      "loss": 0.6996,
+      "step": 258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99125907028293e-06,
+      "loss": 0.4853,
+      "step": 259
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9912172476527046e-06,
+      "loss": 0.8957,
+      "step": 260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9911754250224796e-06,
+      "loss": 0.9889,
+      "step": 261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991133602392254e-06,
+      "loss": 0.7252,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991091779762029e-06,
+      "loss": 0.396,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.991049957131804e-06,
+      "loss": 0.4212,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9910081345015788e-06,
+      "loss": 0.4256,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9909663118713538e-06,
+      "loss": 0.3365,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9909244892411283e-06,
+      "loss": 0.4939,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990882666610903e-06,
+      "loss": 0.3509,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990840843980678e-06,
+      "loss": 0.5079,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9907990213504525e-06,
+      "loss": 0.5613,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9907571987202275e-06,
+      "loss": 0.3336,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990715376090002e-06,
+      "loss": 0.3664,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990673553459777e-06,
+      "loss": 1.0342,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9906317308295516e-06,
+      "loss": 0.4031,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9905899081993266e-06,
+      "loss": 0.7698,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990548085569101e-06,
+      "loss": 0.3431,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990506262938876e-06,
+      "loss": 0.2658,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990464440308651e-06,
+      "loss": 0.7866,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9904226176784258e-06,
+      "loss": 0.2829,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9903807950482008e-06,
+      "loss": 0.2252,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9903389724179753e-06,
+      "loss": 0.2323,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99029714978775e-06,
+      "loss": 0.8496,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990255327157525e-06,
+      "loss": 0.2513,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9902135045272995e-06,
+      "loss": 0.4465,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9901716818970745e-06,
+      "loss": 0.2183,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990129859266849e-06,
+      "loss": 0.29,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.990088036636624e-06,
+      "loss": 0.573,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9900462140063986e-06,
+      "loss": 0.8715,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9900043913761736e-06,
+      "loss": 0.4173,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989962568745948e-06,
+      "loss": 0.2656,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989920746115723e-06,
+      "loss": 0.37,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989878923485498e-06,
+      "loss": 0.4106,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9898371008552728e-06,
+      "loss": 0.3579,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9897952782250474e-06,
+      "loss": 0.3427,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9897534555948223e-06,
+      "loss": 1.2467,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989711632964597e-06,
+      "loss": 1.1,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989669810334372e-06,
+      "loss": 0.5458,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9896279877041465e-06,
+      "loss": 0.9444,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9895861650739215e-06,
+      "loss": 0.7004,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989544342443696e-06,
+      "loss": 0.2795,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989502519813471e-06,
+      "loss": 0.3925,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9894606971832456e-06,
+      "loss": 0.2717,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9894188745530206e-06,
+      "loss": 0.8097,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9893770519227952e-06,
+      "loss": 0.2901,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98933522929257e-06,
+      "loss": 0.4185,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9892934066623448e-06,
+      "loss": 0.4283,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9892515840321198e-06,
+      "loss": 0.2549,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9892097614018944e-06,
+      "loss": 0.5634,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9891679387716694e-06,
+      "loss": 0.3704,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989126116141444e-06,
+      "loss": 0.2535,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989084293511219e-06,
+      "loss": 0.3641,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9890424708809935e-06,
+      "loss": 0.399,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9890006482507685e-06,
+      "loss": 0.856,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988958825620543e-06,
+      "loss": 0.4576,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988917002990318e-06,
+      "loss": 0.2908,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9888751803600926e-06,
+      "loss": 0.2786,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9888333577298676e-06,
+      "loss": 0.3335,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9887915350996422e-06,
+      "loss": 0.398,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9887497124694172e-06,
+      "loss": 0.4107,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988707889839192e-06,
+      "loss": 0.2484,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988666067208967e-06,
+      "loss": 0.3071,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9886242445787414e-06,
+      "loss": 0.7051,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9885824219485164e-06,
+      "loss": 0.2652,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988540599318291e-06,
+      "loss": 0.181,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988498776688066e-06,
+      "loss": 0.3225,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9884569540578405e-06,
+      "loss": 0.2539,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9884151314276155e-06,
+      "loss": 0.5228,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98837330879739e-06,
+      "loss": 0.3728,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988331486167165e-06,
+      "loss": 0.2261,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9882896635369397e-06,
+      "loss": 0.2662,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9882478409067147e-06,
+      "loss": 0.7225,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9882060182764892e-06,
+      "loss": 1.0346,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9881641956462642e-06,
+      "loss": 0.4102,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988122373016039e-06,
+      "loss": 0.5276,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9880805503858134e-06,
+      "loss": 0.9152,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9880387277555884e-06,
+      "loss": 0.2963,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9879969051253634e-06,
+      "loss": 0.2106,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987955082495138e-06,
+      "loss": 0.3281,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987913259864913e-06,
+      "loss": 0.2723,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9878714372346875e-06,
+      "loss": 0.2082,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9878296146044625e-06,
+      "loss": 0.5394,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987787791974237e-06,
+      "loss": 0.3522,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987745969344012e-06,
+      "loss": 0.3353,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9877041467137867e-06,
+      "loss": 0.1936,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9876623240835617e-06,
+      "loss": 0.2385,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9876205014533362e-06,
+      "loss": 0.8949,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9875786788231112e-06,
+      "loss": 0.1562,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987536856192886e-06,
+      "loss": 0.7092,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9874950335626604e-06,
+      "loss": 0.485,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9874532109324354e-06,
+      "loss": 0.3055,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9874113883022104e-06,
+      "loss": 0.2095,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987369565671985e-06,
+      "loss": 0.2876,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98732774304176e-06,
+      "loss": 0.2598,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9872859204115345e-06,
+      "loss": 0.2561,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9872440977813095e-06,
+      "loss": 0.7804,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987202275151084e-06,
+      "loss": 0.6532,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987160452520859e-06,
+      "loss": 0.9417,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9871186298906337e-06,
+      "loss": 0.5501,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9870768072604087e-06,
+      "loss": 0.2146,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9870349846301832e-06,
+      "loss": 0.4059,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986993161999958e-06,
+      "loss": 0.5365,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986951339369733e-06,
+      "loss": 0.1539,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9869095167395074e-06,
+      "loss": 0.2063,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9868676941092824e-06,
+      "loss": 0.4688,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9868258714790574e-06,
+      "loss": 0.2142,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986784048848832e-06,
+      "loss": 0.1757,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986742226218607e-06,
+      "loss": 0.233,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9867004035883815e-06,
+      "loss": 1.0913,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9866585809581565e-06,
+      "loss": 0.2512,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986616758327931e-06,
+      "loss": 1.0412,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986574935697706e-06,
+      "loss": 0.8241,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9865331130674807e-06,
+      "loss": 0.5769,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9864912904372557e-06,
+      "loss": 0.1711,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9864494678070303e-06,
+      "loss": 0.2392,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986407645176805e-06,
+      "loss": 0.1882,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98636582254658e-06,
+      "loss": 0.8748,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9863239999163544e-06,
+      "loss": 0.1626,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9862821772861294e-06,
+      "loss": 0.6156,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986240354655904e-06,
+      "loss": 1.0769,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986198532025679e-06,
+      "loss": 0.1315,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986156709395454e-06,
+      "loss": 0.4181,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9861148867652285e-06,
+      "loss": 0.1571,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9860730641350035e-06,
+      "loss": 0.2482,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986031241504778e-06,
+      "loss": 0.1582,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985989418874553e-06,
+      "loss": 0.319,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9859475962443277e-06,
+      "loss": 0.4388,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9859057736141027e-06,
+      "loss": 0.44,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9858639509838773e-06,
+      "loss": 1.0315,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985822128353652e-06,
+      "loss": 0.1421,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985780305723427e-06,
+      "loss": 0.5199,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9857384830932014e-06,
+      "loss": 0.1625,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9856966604629764e-06,
+      "loss": 0.6012,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985654837832751e-06,
+      "loss": 1.0827,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985613015202526e-06,
+      "loss": 0.2738,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985571192572301e-06,
+      "loss": 0.2358,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9855293699420756e-06,
+      "loss": 0.2531,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9854875473118506e-06,
+      "loss": 0.2839,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985445724681625e-06,
+      "loss": 0.329,
+      "step": 398
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9854039020514e-06,
+      "loss": 0.9575,
+      "step": 399
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9853620794211747e-06,
+      "loss": 0.3571,
+      "step": 400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9853202567909497e-06,
+      "loss": 0.4814,
+      "step": 401
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9852784341607243e-06,
+      "loss": 0.2758,
+      "step": 402
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.985236611530499e-06,
+      "loss": 0.6626,
+      "step": 403
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.985194788900274e-06,
+      "loss": 0.7826,
+      "step": 404
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9851529662700484e-06,
+      "loss": 0.2322,
+      "step": 405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9851111436398234e-06,
+      "loss": 0.2501,
+      "step": 406
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.985069321009598e-06,
+      "loss": 0.3984,
+      "step": 407
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.985027498379373e-06,
+      "loss": 0.1431,
+      "step": 408
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984985675749148e-06,
+      "loss": 0.1763,
+      "step": 409
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9849438531189226e-06,
+      "loss": 1.0178,
+      "step": 410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9849020304886976e-06,
+      "loss": 0.2146,
+      "step": 411
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984860207858472e-06,
+      "loss": 0.3585,
+      "step": 412
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984818385228247e-06,
+      "loss": 0.1676,
+      "step": 413
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9847765625980217e-06,
+      "loss": 0.1953,
+      "step": 414
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9847347399677967e-06,
+      "loss": 0.3383,
+      "step": 415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9846929173375713e-06,
+      "loss": 0.2022,
+      "step": 416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984651094707346e-06,
+      "loss": 0.254,
+      "step": 417
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984609272077121e-06,
+      "loss": 0.3267,
+      "step": 418
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9845674494468954e-06,
+      "loss": 0.266,
+      "step": 419
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9845256268166704e-06,
+      "loss": 0.3025,
+      "step": 420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984483804186445e-06,
+      "loss": 0.2167,
+      "step": 421
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98444198155622e-06,
+      "loss": 0.2219,
+      "step": 422
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9844001589259946e-06,
+      "loss": 0.1868,
+      "step": 423
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9843583362957696e-06,
+      "loss": 0.3659,
+      "step": 424
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9843165136655446e-06,
+      "loss": 0.8052,
+      "step": 425
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984274691035319e-06,
+      "loss": 0.1313,
+      "step": 426
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984232868405094e-06,
+      "loss": 0.1618,
+      "step": 427
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9841910457748687e-06,
+      "loss": 0.189,
+      "step": 428
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9841492231446433e-06,
+      "loss": 0.8475,
+      "step": 429
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9841074005144183e-06,
+      "loss": 0.298,
+      "step": 430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984065577884193e-06,
+      "loss": 0.142,
+      "step": 431
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.984023755253968e-06,
+      "loss": 0.1166,
+      "step": 432
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9839819326237424e-06,
+      "loss": 0.1909,
+      "step": 433
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9839401099935174e-06,
+      "loss": 0.2148,
+      "step": 434
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983898287363292e-06,
+      "loss": 0.128,
+      "step": 435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983856464733067e-06,
+      "loss": 0.3363,
+      "step": 436
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9838146421028416e-06,
+      "loss": 0.7066,
+      "step": 437
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9837728194726166e-06,
+      "loss": 0.1025,
+      "step": 438
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9837309968423916e-06,
+      "loss": 0.2074,
+      "step": 439
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983689174212166e-06,
+      "loss": 0.1266,
+      "step": 440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983647351581941e-06,
+      "loss": 0.187,
+      "step": 441
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9836055289517157e-06,
+      "loss": 0.6712,
+      "step": 442
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9835637063214903e-06,
+      "loss": 0.097,
+      "step": 443
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9835218836912653e-06,
+      "loss": 0.1412,
+      "step": 444
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98348006106104e-06,
+      "loss": 0.5795,
+      "step": 445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983438238430815e-06,
+      "loss": 0.3231,
+      "step": 446
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9833964158005894e-06,
+      "loss": 0.1637,
+      "step": 447
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9833545931703644e-06,
+      "loss": 0.4288,
+      "step": 448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983312770540139e-06,
+      "loss": 0.1158,
+      "step": 449
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983270947909914e-06,
+      "loss": 0.3036,
+      "step": 450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9832291252796886e-06,
+      "loss": 1.5677,
+      "step": 451
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9831873026494636e-06,
+      "loss": 0.1528,
+      "step": 452
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9831454800192386e-06,
+      "loss": 0.2978,
+      "step": 453
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983103657389013e-06,
+      "loss": 0.163,
+      "step": 454
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983061834758788e-06,
+      "loss": 0.491,
+      "step": 455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9830200121285627e-06,
+      "loss": 1.6287,
+      "step": 456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9829781894983373e-06,
+      "loss": 0.1202,
+      "step": 457
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9829363668681123e-06,
+      "loss": 0.2562,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982894544237887e-06,
+      "loss": 0.1995,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982852721607662e-06,
+      "loss": 1.172,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9828108989774365e-06,
+      "loss": 0.2525,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9827690763472115e-06,
+      "loss": 1.6812,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982727253716986e-06,
+      "loss": 0.5528,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982685431086761e-06,
+      "loss": 0.1161,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9826436084565356e-06,
+      "loss": 0.2499,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9826017858263106e-06,
+      "loss": 1.4384,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982559963196085e-06,
+      "loss": 0.334,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98251814056586e-06,
+      "loss": 1.2484,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982476317935635e-06,
+      "loss": 0.1281,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9824344953054097e-06,
+      "loss": 0.0955,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9823926726751843e-06,
+      "loss": 0.4238,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9823508500449593e-06,
+      "loss": 0.0987,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982309027414734e-06,
+      "loss": 0.1678,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982267204784509e-06,
+      "loss": 0.1035,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9822253821542835e-06,
+      "loss": 0.103,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9821835595240585e-06,
+      "loss": 1.0647,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982141736893833e-06,
+      "loss": 0.1141,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982099914263608e-06,
+      "loss": 0.5272,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9820580916333826e-06,
+      "loss": 0.3931,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9820162690031576e-06,
+      "loss": 0.4057,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981974446372932e-06,
+      "loss": 0.2202,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981932623742707e-06,
+      "loss": 0.1274,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981890801112482e-06,
+      "loss": 0.0981,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9818489784822568e-06,
+      "loss": 0.2609,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9818071558520313e-06,
+      "loss": 0.2577,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9817653332218063e-06,
+      "loss": 0.2853,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981723510591581e-06,
+      "loss": 0.377,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981681687961356e-06,
+      "loss": 0.1503,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9816398653311305e-06,
+      "loss": 0.1227,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9815980427009055e-06,
+      "loss": 1.5409,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98155622007068e-06,
+      "loss": 1.127,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981514397440455e-06,
+      "loss": 0.1337,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9814725748102296e-06,
+      "loss": 0.1764,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9814307521800046e-06,
+      "loss": 0.1601,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981388929549779e-06,
+      "loss": 0.7668,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981347106919554e-06,
+      "loss": 0.1187,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9813052842893288e-06,
+      "loss": 0.9933,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9812634616591038e-06,
+      "loss": 0.9421,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9812216390288783e-06,
+      "loss": 0.48,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9811798163986533e-06,
+      "loss": 0.2834,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981137993768428e-06,
+      "loss": 0.416,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981096171138203e-06,
+      "loss": 0.2488,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9810543485079775e-06,
+      "loss": 0.2904,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9810125258777525e-06,
+      "loss": 1.0721,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980970703247527e-06,
+      "loss": 0.8854,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980928880617302e-06,
+      "loss": 0.2056,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9808870579870766e-06,
+      "loss": 1.0958,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9808452353568516e-06,
+      "loss": 0.3947,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980803412726626e-06,
+      "loss": 0.2669,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9807615900964008e-06,
+      "loss": 0.3458,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9807197674661758e-06,
+      "loss": 0.202,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9806779448359508e-06,
+      "loss": 0.1445,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9806361222057253e-06,
+      "loss": 0.1597,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9805942995755003e-06,
+      "loss": 0.9957,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980552476945275e-06,
+      "loss": 0.861,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98051065431505e-06,
+      "loss": 0.2763,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9804688316848245e-06,
+      "loss": 0.2586,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9804270090545995e-06,
+      "loss": 0.264,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980385186424374e-06,
+      "loss": 0.3476,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980343363794149e-06,
+      "loss": 0.389,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9803015411639236e-06,
+      "loss": 0.1523,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9802597185336986e-06,
+      "loss": 0.1184,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980217895903473e-06,
+      "loss": 0.3251,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9801760732732478e-06,
+      "loss": 0.1002,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9801342506430228e-06,
+      "loss": 0.2048,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9800924280127978e-06,
+      "loss": 0.116,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9800506053825724e-06,
+      "loss": 0.081,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9800087827523474e-06,
+      "loss": 0.1933,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979966960122122e-06,
+      "loss": 1.255,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979925137491897e-06,
+      "loss": 0.3442,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9798833148616715e-06,
+      "loss": 0.1372,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9798414922314465e-06,
+      "loss": 0.1598,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979799669601221e-06,
+      "loss": 0.3943,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979757846970996e-06,
+      "loss": 0.0957,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9797160243407706e-06,
+      "loss": 0.1194,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9796742017105456e-06,
+      "loss": 0.3344,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9796323790803202e-06,
+      "loss": 0.1282,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979590556450095e-06,
+      "loss": 0.146,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.97954873381987e-06,
+      "loss": 0.4522,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9795069111896444e-06,
+      "loss": 0.2985,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9794650885594194e-06,
+      "loss": 1.2649,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9794232659291944e-06,
+      "loss": 0.1542,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979381443298969e-06,
+      "loss": 0.0774,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979339620668744e-06,
+      "loss": 0.2688,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9792977980385185e-06,
+      "loss": 0.1145,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9792559754082935e-06,
+      "loss": 0.1254,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979214152778068e-06,
+      "loss": 0.2425,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979172330147843e-06,
+      "loss": 0.1693,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9791305075176177e-06,
+      "loss": 0.3178,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9790886848873927e-06,
+      "loss": 0.1255,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9790468622571672e-06,
+      "loss": 0.0843,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979005039626942e-06,
+      "loss": 0.4584,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978963216996717e-06,
+      "loss": 0.0991,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9789213943664914e-06,
+      "loss": 0.1371,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9788795717362664e-06,
+      "loss": 0.0948,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9788377491060414e-06,
+      "loss": 0.12,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978795926475816e-06,
+      "loss": 1.8704,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978754103845591e-06,
+      "loss": 0.253,
+      "step": 558
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9787122812153655e-06,
+      "loss": 0.6468,
+      "step": 559
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9786704585851405e-06,
+      "loss": 0.1343,
+      "step": 560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.978628635954915e-06,
+      "loss": 0.2778,
+      "step": 561
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97858681332469e-06,
+      "loss": 0.1482,
+      "step": 562
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9785449906944647e-06,
+      "loss": 0.8678,
+      "step": 563
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9785031680642392e-06,
+      "loss": 0.1648,
+      "step": 564
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9784613454340142e-06,
+      "loss": 0.8925,
+      "step": 565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.978419522803789e-06,
+      "loss": 0.3449,
+      "step": 566
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.978377700173564e-06,
+      "loss": 0.0659,
+      "step": 567
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9783358775433384e-06,
+      "loss": 0.1116,
+      "step": 568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9782940549131134e-06,
+      "loss": 0.1516,
+      "step": 569
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9782522322828884e-06,
+      "loss": 0.0889,
+      "step": 570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.978210409652663e-06,
+      "loss": 0.0737,
+      "step": 571
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.978168587022438e-06,
+      "loss": 0.208,
+      "step": 572
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9781267643922125e-06,
+      "loss": 0.8363,
+      "step": 573
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9780849417619875e-06,
+      "loss": 0.0732,
+      "step": 574
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.978043119131762e-06,
+      "loss": 0.6823,
+      "step": 575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.978001296501537e-06,
+      "loss": 0.2276,
+      "step": 576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9779594738713117e-06,
+      "loss": 0.0848,
+      "step": 577
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9779176512410862e-06,
+      "loss": 0.3855,
+      "step": 578
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9778758286108612e-06,
+      "loss": 0.8563,
+      "step": 579
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977834005980636e-06,
+      "loss": 0.0877,
+      "step": 580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977792183350411e-06,
+      "loss": 0.5052,
+      "step": 581
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9777503607201854e-06,
+      "loss": 0.0602,
+      "step": 582
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9777085380899604e-06,
+      "loss": 1.1192,
+      "step": 583
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977666715459735e-06,
+      "loss": 0.1068,
+      "step": 584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97762489282951e-06,
+      "loss": 0.069,
+      "step": 585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977583070199285e-06,
+      "loss": 0.1584,
+      "step": 586
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9775412475690595e-06,
+      "loss": 0.1198,
+      "step": 587
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9774994249388345e-06,
+      "loss": 0.1611,
+      "step": 588
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977457602308609e-06,
+      "loss": 0.1208,
+      "step": 589
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977415779678384e-06,
+      "loss": 0.2472,
+      "step": 590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9773739570481587e-06,
+      "loss": 0.0639,
+      "step": 591
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9773321344179333e-06,
+      "loss": 0.118,
+      "step": 592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9772903117877083e-06,
+      "loss": 0.1153,
+      "step": 593
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977248489157483e-06,
+      "loss": 1.3868,
+      "step": 594
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977206666527258e-06,
+      "loss": 0.4098,
+      "step": 595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9771648438970324e-06,
+      "loss": 1.0964,
+      "step": 596
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9771230212668074e-06,
+      "loss": 0.1532,
+      "step": 597
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977081198636582e-06,
+      "loss": 0.0741,
+      "step": 598
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.977039376006357e-06,
+      "loss": 0.1068,
+      "step": 599
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976997553376132e-06,
+      "loss": 0.4453,
+      "step": 600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9769557307459065e-06,
+      "loss": 1.3869,
+      "step": 601
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9769139081156815e-06,
+      "loss": 0.2109,
+      "step": 602
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976872085485456e-06,
+      "loss": 0.3085,
+      "step": 603
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976830262855231e-06,
+      "loss": 0.117,
+      "step": 604
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9767884402250057e-06,
+      "loss": 0.0597,
+      "step": 605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9767466175947803e-06,
+      "loss": 0.0972,
+      "step": 606
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9767047949645553e-06,
+      "loss": 0.2438,
+      "step": 607
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97666297233433e-06,
+      "loss": 0.3658,
+      "step": 608
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976621149704105e-06,
+      "loss": 0.1312,
+      "step": 609
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9765793270738794e-06,
+      "loss": 0.1876,
+      "step": 610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9765375044436544e-06,
+      "loss": 0.1214,
+      "step": 611
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976495681813429e-06,
+      "loss": 0.1715,
+      "step": 612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976453859183204e-06,
+      "loss": 0.4359,
+      "step": 613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976412036552979e-06,
+      "loss": 0.1938,
+      "step": 614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9763702139227536e-06,
+      "loss": 0.0884,
+      "step": 615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9763283912925285e-06,
+      "loss": 0.1568,
+      "step": 616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976286568662303e-06,
+      "loss": 0.7749,
+      "step": 617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976244746032078e-06,
+      "loss": 0.1489,
+      "step": 618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9762029234018527e-06,
+      "loss": 0.9698,
+      "step": 619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9761611007716273e-06,
+      "loss": 0.308,
+      "step": 620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9761192781414023e-06,
+      "loss": 0.384,
+      "step": 621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976077455511177e-06,
+      "loss": 0.9369,
+      "step": 622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976035632880952e-06,
+      "loss": 0.078,
+      "step": 623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9759938102507264e-06,
+      "loss": 0.1272,
+      "step": 624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9759519876205014e-06,
+      "loss": 0.1713,
+      "step": 625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975910164990276e-06,
+      "loss": 0.1307,
+      "step": 626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975868342360051e-06,
+      "loss": 0.9392,
+      "step": 627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9758265197298256e-06,
+      "loss": 0.1956,
+      "step": 628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9757846970996006e-06,
+      "loss": 0.0539,
+      "step": 629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9757428744693756e-06,
+      "loss": 0.0546,
+      "step": 630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97570105183915e-06,
+      "loss": 0.0543,
+      "step": 631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9756592292089247e-06,
+      "loss": 0.0954,
+      "step": 632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9756174065786997e-06,
+      "loss": 0.0852,
+      "step": 633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9755755839484743e-06,
+      "loss": 0.1359,
+      "step": 634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9755337613182493e-06,
+      "loss": 0.1038,
+      "step": 635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975491938688024e-06,
+      "loss": 0.1586,
+      "step": 636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975450116057799e-06,
+      "loss": 0.0816,
+      "step": 637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9754082934275734e-06,
+      "loss": 0.0565,
+      "step": 638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9753664707973484e-06,
+      "loss": 0.0669,
+      "step": 639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975324648167123e-06,
+      "loss": 0.0743,
+      "step": 640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975282825536898e-06,
+      "loss": 0.3219,
+      "step": 641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9752410029066726e-06,
+      "loss": 0.0999,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9751991802764476e-06,
+      "loss": 0.1443,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9751573576462226e-06,
+      "loss": 0.0518,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975115535015997e-06,
+      "loss": 0.1828,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9750737123857717e-06,
+      "loss": 0.0833,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9750318897555467e-06,
+      "loss": 1.6288,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9749900671253213e-06,
+      "loss": 0.1275,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9749482444950963e-06,
+      "loss": 0.1859,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974906421864871e-06,
+      "loss": 0.1075,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974864599234646e-06,
+      "loss": 0.0834,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9748227766044204e-06,
+      "loss": 0.1828,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9747809539741954e-06,
+      "loss": 0.1906,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97473913134397e-06,
+      "loss": 0.0649,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974697308713745e-06,
+      "loss": 0.0608,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9746554860835196e-06,
+      "loss": 0.0732,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9746136634532946e-06,
+      "loss": 0.2704,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9745718408230696e-06,
+      "loss": 0.2828,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974530018192844e-06,
+      "loss": 0.4599,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9744881955626187e-06,
+      "loss": 0.1429,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9744463729323937e-06,
+      "loss": 0.1684,
+      "step": 661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9744045503021683e-06,
+      "loss": 1.5451,
+      "step": 662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9743627276719433e-06,
+      "loss": 1.3234,
+      "step": 663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974320905041718e-06,
+      "loss": 0.0993,
+      "step": 664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974279082411493e-06,
+      "loss": 1.5585,
+      "step": 665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9742372597812674e-06,
+      "loss": 0.4722,
+      "step": 666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9741954371510424e-06,
+      "loss": 0.0565,
+      "step": 667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974153614520817e-06,
+      "loss": 0.2291,
+      "step": 668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974111791890592e-06,
+      "loss": 0.4554,
+      "step": 669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9740699692603666e-06,
+      "loss": 0.2268,
+      "step": 670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9740281466301416e-06,
+      "loss": 0.565,
+      "step": 671
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973986323999916e-06,
+      "loss": 0.0741,
+      "step": 672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973944501369691e-06,
+      "loss": 1.6981,
+      "step": 673
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9739026787394657e-06,
+      "loss": 0.0516,
+      "step": 674
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9738608561092407e-06,
+      "loss": 0.0539,
+      "step": 675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9738190334790153e-06,
+      "loss": 0.1514,
+      "step": 676
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9737772108487903e-06,
+      "loss": 1.2066,
+      "step": 677
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973735388218565e-06,
+      "loss": 0.5977,
+      "step": 678
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97369356558834e-06,
+      "loss": 0.1831,
+      "step": 679
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9736517429581144e-06,
+      "loss": 0.0526,
+      "step": 680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9736099203278894e-06,
+      "loss": 0.1052,
+      "step": 681
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973568097697664e-06,
+      "loss": 0.0565,
+      "step": 682
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973526275067439e-06,
+      "loss": 0.0614,
+      "step": 683
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9734844524372136e-06,
+      "loss": 0.4527,
+      "step": 684
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9734426298069886e-06,
+      "loss": 1.7131,
+      "step": 685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973400807176763e-06,
+      "loss": 0.1478,
+      "step": 686
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973358984546538e-06,
+      "loss": 0.0964,
+      "step": 687
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9733171619163127e-06,
+      "loss": 0.1814,
+      "step": 688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9732753392860877e-06,
+      "loss": 1.6218,
+      "step": 689
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9732335166558623e-06,
+      "loss": 0.0531,
+      "step": 690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9731916940256373e-06,
+      "loss": 1.6438,
+      "step": 691
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973149871395412e-06,
+      "loss": 0.1548,
+      "step": 692
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973108048765187e-06,
+      "loss": 0.1078,
+      "step": 693
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9730662261349615e-06,
+      "loss": 0.0981,
+      "step": 694
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9730244035047365e-06,
+      "loss": 0.095,
+      "step": 695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972982580874511e-06,
+      "loss": 0.1011,
+      "step": 696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972940758244286e-06,
+      "loss": 0.0617,
+      "step": 697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9728989356140606e-06,
+      "loss": 0.3238,
+      "step": 698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9728571129838356e-06,
+      "loss": 0.0865,
+      "step": 699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97281529035361e-06,
+      "loss": 0.0428,
+      "step": 700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9727734677233848e-06,
+      "loss": 0.0796,
+      "step": 701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9727316450931597e-06,
+      "loss": 0.1692,
+      "step": 702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9726898224629347e-06,
+      "loss": 0.0482,
+      "step": 703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9726479998327093e-06,
+      "loss": 0.0542,
+      "step": 704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9726061772024843e-06,
+      "loss": 0.62,
+      "step": 705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972564354572259e-06,
+      "loss": 0.0448,
+      "step": 706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972522531942034e-06,
+      "loss": 0.0645,
+      "step": 707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9724807093118085e-06,
+      "loss": 0.0633,
+      "step": 708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9724388866815835e-06,
+      "loss": 0.0979,
+      "step": 709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972397064051358e-06,
+      "loss": 0.3244,
+      "step": 710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972355241421133e-06,
+      "loss": 0.1027,
+      "step": 711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9723134187909076e-06,
+      "loss": 0.0696,
+      "step": 712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972271596160682e-06,
+      "loss": 0.0697,
+      "step": 713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972229773530457e-06,
+      "loss": 0.0768,
+      "step": 714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9721879509002318e-06,
+      "loss": 0.0527,
+      "step": 715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9721461282700068e-06,
+      "loss": 0.1737,
+      "step": 716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9721043056397818e-06,
+      "loss": 1.1312,
+      "step": 717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9720624830095563e-06,
+      "loss": 0.0343,
+      "step": 718
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9720206603793313e-06,
+      "loss": 0.0473,
+      "step": 719
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971978837749106e-06,
+      "loss": 0.0753,
+      "step": 720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971937015118881e-06,
+      "loss": 0.0411,
+      "step": 721
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9718951924886555e-06,
+      "loss": 0.052,
+      "step": 722
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9718533698584305e-06,
+      "loss": 0.3372,
+      "step": 723
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971811547228205e-06,
+      "loss": 0.2828,
+      "step": 724
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.97176972459798e-06,
+      "loss": 0.2157,
+      "step": 725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9717279019677546e-06,
+      "loss": 0.0549,
+      "step": 726
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971686079337529e-06,
+      "loss": 0.0463,
+      "step": 727
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971644256707304e-06,
+      "loss": 0.0813,
+      "step": 728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9716024340770788e-06,
+      "loss": 0.216,
+      "step": 729
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9715606114468538e-06,
+      "loss": 0.0748,
+      "step": 730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9715187888166288e-06,
+      "loss": 0.1937,
+      "step": 731
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9714769661864033e-06,
+      "loss": 0.0322,
+      "step": 732
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9714351435561783e-06,
+      "loss": 0.3634,
+      "step": 733
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971393320925953e-06,
+      "loss": 0.0541,
+      "step": 734
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971351498295728e-06,
+      "loss": 0.1757,
+      "step": 735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9713096756655025e-06,
+      "loss": 0.0362,
+      "step": 736
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9712678530352775e-06,
+      "loss": 0.072,
+      "step": 737
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971226030405052e-06,
+      "loss": 0.0427,
+      "step": 738
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971184207774827e-06,
+      "loss": 0.0481,
+      "step": 739
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9711423851446016e-06,
+      "loss": 0.0717,
+      "step": 740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971100562514376e-06,
+      "loss": 1.464,
+      "step": 741
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.971058739884151e-06,
+      "loss": 0.0472,
+      "step": 742
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9710169172539258e-06,
+      "loss": 0.043,
+      "step": 743
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9709750946237008e-06,
+      "loss": 0.0812,
+      "step": 744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9709332719934753e-06,
+      "loss": 0.0312,
+      "step": 745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9708914493632503e-06,
+      "loss": 1.0341,
+      "step": 746
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9708496267330253e-06,
+      "loss": 1.7694,
+      "step": 747
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9708078041028e-06,
+      "loss": 0.0433,
+      "step": 748
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.970765981472575e-06,
+      "loss": 0.3292,
+      "step": 749
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9707241588423495e-06,
+      "loss": 0.0357,
+      "step": 750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9706823362121245e-06,
+      "loss": 0.0315,
+      "step": 751
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.970640513581899e-06,
+      "loss": 0.14,
+      "step": 752
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.970598690951674e-06,
+      "loss": 1.7635,
+      "step": 753
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9705568683214486e-06,
+      "loss": 1.3237,
+      "step": 754
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9705150456912232e-06,
+      "loss": 0.0378,
+      "step": 755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.970473223060998e-06,
+      "loss": 0.0308,
+      "step": 756
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9704314004307728e-06,
+      "loss": 0.0947,
+      "step": 757
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9703895778005478e-06,
+      "loss": 0.0523,
+      "step": 758
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9703477551703224e-06,
+      "loss": 0.1156,
+      "step": 759
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9703059325400974e-06,
+      "loss": 0.1105,
+      "step": 760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9702641099098724e-06,
+      "loss": 1.9895,
+      "step": 761
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.970222287279647e-06,
+      "loss": 0.3784,
+      "step": 762
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.970180464649422e-06,
+      "loss": 0.3488,
+      "step": 763
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9701386420191965e-06,
+      "loss": 1.5724,
+      "step": 764
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9700968193889715e-06,
+      "loss": 0.1694,
+      "step": 765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.970054996758746e-06,
+      "loss": 0.0613,
+      "step": 766
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9700131741285206e-06,
+      "loss": 0.0365,
+      "step": 767
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9699713514982956e-06,
+      "loss": 0.2734,
+      "step": 768
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9699295288680702e-06,
+      "loss": 0.0638,
+      "step": 769
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9698877062378452e-06,
+      "loss": 0.0481,
+      "step": 770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96984588360762e-06,
+      "loss": 1.4291,
+      "step": 771
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969804060977395e-06,
+      "loss": 0.9801,
+      "step": 772
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9697622383471694e-06,
+      "loss": 0.035,
+      "step": 773
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9697204157169444e-06,
+      "loss": 0.0283,
+      "step": 774
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9696785930867194e-06,
+      "loss": 0.0903,
+      "step": 775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969636770456494e-06,
+      "loss": 0.0798,
+      "step": 776
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969594947826269e-06,
+      "loss": 0.0298,
+      "step": 777
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9695531251960435e-06,
+      "loss": 0.0392,
+      "step": 778
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9695113025658185e-06,
+      "loss": 0.1061,
+      "step": 779
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969469479935593e-06,
+      "loss": 0.08,
+      "step": 780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9694276573053677e-06,
+      "loss": 0.1489,
+      "step": 781
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9693858346751427e-06,
+      "loss": 0.0329,
+      "step": 782
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9693440120449172e-06,
+      "loss": 0.0288,
+      "step": 783
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9693021894146922e-06,
+      "loss": 0.1914,
+      "step": 784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969260366784467e-06,
+      "loss": 0.1701,
+      "step": 785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969218544154242e-06,
+      "loss": 2.3634,
+      "step": 786
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9691767215240164e-06,
+      "loss": 0.029,
+      "step": 787
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9691348988937914e-06,
+      "loss": 0.4271,
+      "step": 788
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969093076263566e-06,
+      "loss": 0.0263,
+      "step": 789
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969051253633341e-06,
+      "loss": 0.1321,
+      "step": 790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969009431003116e-06,
+      "loss": 0.1314,
+      "step": 791
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9689676083728905e-06,
+      "loss": 0.1263,
+      "step": 792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9689257857426655e-06,
+      "loss": 0.4084,
+      "step": 793
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96888396311244e-06,
+      "loss": 0.0271,
+      "step": 794
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9688421404822147e-06,
+      "loss": 1.6419,
+      "step": 795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9688003178519897e-06,
+      "loss": 0.141,
+      "step": 796
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9687584952217642e-06,
+      "loss": 0.0362,
+      "step": 797
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9687166725915392e-06,
+      "loss": 0.0351,
+      "step": 798
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968674849961314e-06,
+      "loss": 2.1435,
+      "step": 799
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968633027331089e-06,
+      "loss": 1.1006,
+      "step": 800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9685912047008634e-06,
+      "loss": 0.064,
+      "step": 801
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9685493820706384e-06,
+      "loss": 0.024,
+      "step": 802
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968507559440413e-06,
+      "loss": 1.3972,
+      "step": 803
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968465736810188e-06,
+      "loss": 0.0235,
+      "step": 804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968423914179963e-06,
+      "loss": 0.2989,
+      "step": 805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9683820915497375e-06,
+      "loss": 0.0335,
+      "step": 806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9683402689195125e-06,
+      "loss": 0.0244,
+      "step": 807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968298446289287e-06,
+      "loss": 1.3934,
+      "step": 808
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9682566236590617e-06,
+      "loss": 0.0418,
+      "step": 809
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9682148010288367e-06,
+      "loss": 0.0239,
+      "step": 810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9681729783986112e-06,
+      "loss": 0.6572,
+      "step": 811
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9681311557683862e-06,
+      "loss": 0.4568,
+      "step": 812
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968089333138161e-06,
+      "loss": 1.67,
+      "step": 813
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968047510507936e-06,
+      "loss": 0.0362,
+      "step": 814
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9680056878777104e-06,
+      "loss": 0.2407,
+      "step": 815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9679638652474854e-06,
+      "loss": 1.6487,
+      "step": 816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96792204261726e-06,
+      "loss": 0.0251,
+      "step": 817
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967880219987035e-06,
+      "loss": 0.196,
+      "step": 818
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96783839735681e-06,
+      "loss": 0.0415,
+      "step": 819
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9677965747265845e-06,
+      "loss": 1.3261,
+      "step": 820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9677547520963595e-06,
+      "loss": 0.0269,
+      "step": 821
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967712929466134e-06,
+      "loss": 0.0523,
+      "step": 822
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9676711068359087e-06,
+      "loss": 0.0463,
+      "step": 823
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9676292842056837e-06,
+      "loss": 0.0222,
+      "step": 824
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9675874615754583e-06,
+      "loss": 0.0293,
+      "step": 825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9675456389452333e-06,
+      "loss": 0.3081,
+      "step": 826
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967503816315008e-06,
+      "loss": 0.1007,
+      "step": 827
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967461993684783e-06,
+      "loss": 0.1022,
+      "step": 828
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9674201710545574e-06,
+      "loss": 0.0176,
+      "step": 829
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9673783484243324e-06,
+      "loss": 0.0318,
+      "step": 830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967336525794107e-06,
+      "loss": 0.0247,
+      "step": 831
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967294703163882e-06,
+      "loss": 0.0227,
+      "step": 832
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9672528805336565e-06,
+      "loss": 0.0839,
+      "step": 833
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9672110579034315e-06,
+      "loss": 0.0519,
+      "step": 834
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967169235273206e-06,
+      "loss": 0.0323,
+      "step": 835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967127412642981e-06,
+      "loss": 1.5281,
+      "step": 836
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9670855900127557e-06,
+      "loss": 1.5346,
+      "step": 837
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9670437673825307e-06,
+      "loss": 0.0305,
+      "step": 838
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9670019447523053e-06,
+      "loss": 0.0474,
+      "step": 839
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9669601221220803e-06,
+      "loss": 0.3066,
+      "step": 840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966918299491855e-06,
+      "loss": 0.7795,
+      "step": 841
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96687647686163e-06,
+      "loss": 0.0459,
+      "step": 842
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9668346542314044e-06,
+      "loss": 0.0411,
+      "step": 843
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9667928316011794e-06,
+      "loss": 0.1336,
+      "step": 844
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966751008970954e-06,
+      "loss": 0.0202,
+      "step": 845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966709186340729e-06,
+      "loss": 0.1012,
+      "step": 846
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9666673637105036e-06,
+      "loss": 2.431,
+      "step": 847
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9666255410802786e-06,
+      "loss": 1.1029,
+      "step": 848
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966583718450053e-06,
+      "loss": 0.0194,
+      "step": 849
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966541895819828e-06,
+      "loss": 0.0214,
+      "step": 850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9665000731896027e-06,
+      "loss": 0.2864,
+      "step": 851
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9664582505593777e-06,
+      "loss": 0.0317,
+      "step": 852
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9664164279291523e-06,
+      "loss": 0.6342,
+      "step": 853
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9663746052989273e-06,
+      "loss": 0.2238,
+      "step": 854
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966332782668702e-06,
+      "loss": 1.5463,
+      "step": 855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966290960038477e-06,
+      "loss": 0.0309,
+      "step": 856
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9662491374082514e-06,
+      "loss": 0.0637,
+      "step": 857
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9662073147780264e-06,
+      "loss": 0.1626,
+      "step": 858
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966165492147801e-06,
+      "loss": 0.1686,
+      "step": 859
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966123669517576e-06,
+      "loss": 1.117,
+      "step": 860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9660818468873506e-06,
+      "loss": 0.1182,
+      "step": 861
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966040024257125e-06,
+      "loss": 0.0215,
+      "step": 862
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9659982016269e-06,
+      "loss": 1.4502,
+      "step": 863
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965956378996675e-06,
+      "loss": 0.1553,
+      "step": 864
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9659145563664497e-06,
+      "loss": 0.0558,
+      "step": 865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9658727337362247e-06,
+      "loss": 0.1415,
+      "step": 866
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9658309111059993e-06,
+      "loss": 2.0228,
+      "step": 867
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9657890884757743e-06,
+      "loss": 1.9146,
+      "step": 868
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965747265845549e-06,
+      "loss": 0.117,
+      "step": 869
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965705443215324e-06,
+      "loss": 0.039,
+      "step": 870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9656636205850984e-06,
+      "loss": 0.8313,
+      "step": 871
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9656217979548734e-06,
+      "loss": 0.0299,
+      "step": 872
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965579975324648e-06,
+      "loss": 0.3556,
+      "step": 873
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965538152694423e-06,
+      "loss": 0.4893,
+      "step": 874
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9654963300641976e-06,
+      "loss": 0.0217,
+      "step": 875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965454507433972e-06,
+      "loss": 0.2387,
+      "step": 876
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965412684803747e-06,
+      "loss": 0.8926,
+      "step": 877
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.965370862173522e-06,
+      "loss": 0.0193,
+      "step": 878
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9653290395432967e-06,
+      "loss": 0.1186,
+      "step": 879
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9652872169130717e-06,
+      "loss": 0.0222,
+      "step": 880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9652453942828463e-06,
+      "loss": 0.0801,
+      "step": 881
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9652035716526213e-06,
+      "loss": 0.5012,
+      "step": 882
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.965161749022396e-06,
+      "loss": 0.0228,
+      "step": 883
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.965119926392171e-06,
+      "loss": 0.0296,
+      "step": 884
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9650781037619454e-06,
+      "loss": 2.1425,
+      "step": 885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9650362811317204e-06,
+      "loss": 0.7486,
+      "step": 886
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964994458501495e-06,
+      "loss": 0.025,
+      "step": 887
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96495263587127e-06,
+      "loss": 1.0759,
+      "step": 888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9649108132410446e-06,
+      "loss": 0.0314,
+      "step": 889
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964868990610819e-06,
+      "loss": 0.0222,
+      "step": 890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964827167980594e-06,
+      "loss": 0.0224,
+      "step": 891
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964785345350369e-06,
+      "loss": 1.2986,
+      "step": 892
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9647435227201437e-06,
+      "loss": 2.3662,
+      "step": 893
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9647017000899187e-06,
+      "loss": 0.0294,
+      "step": 894
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9646598774596933e-06,
+      "loss": 0.0272,
+      "step": 895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9646180548294683e-06,
+      "loss": 0.2117,
+      "step": 896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964576232199243e-06,
+      "loss": 0.04,
+      "step": 897
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964534409569018e-06,
+      "loss": 0.5546,
+      "step": 898
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9644925869387924e-06,
+      "loss": 0.0409,
+      "step": 899
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9644507643085674e-06,
+      "loss": 2.6504,
+      "step": 900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964408941678342e-06,
+      "loss": 0.0329,
+      "step": 901
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964367119048117e-06,
+      "loss": 0.0257,
+      "step": 902
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9643252964178916e-06,
+      "loss": 0.0199,
+      "step": 903
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964283473787666e-06,
+      "loss": 1.862,
+      "step": 904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.964241651157441e-06,
+      "loss": 0.2457,
+      "step": 905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9641998285272157e-06,
+      "loss": 0.0684,
+      "step": 906
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9641580058969907e-06,
+      "loss": 0.028,
+      "step": 907
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9641161832667657e-06,
+      "loss": 0.0325,
+      "step": 908
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9640743606365403e-06,
+      "loss": 0.9851,
+      "step": 909
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9640325380063153e-06,
+      "loss": 0.019,
+      "step": 910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96399071537609e-06,
+      "loss": 0.0351,
+      "step": 911
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963948892745865e-06,
+      "loss": 0.0202,
+      "step": 912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9639070701156395e-06,
+      "loss": 0.0603,
+      "step": 913
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9638652474854145e-06,
+      "loss": 0.8688,
+      "step": 914
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963823424855189e-06,
+      "loss": 1.9119,
+      "step": 915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9637816022249636e-06,
+      "loss": 0.0294,
+      "step": 916
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9637397795947386e-06,
+      "loss": 0.0507,
+      "step": 917
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963697956964513e-06,
+      "loss": 0.0331,
+      "step": 918
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963656134334288e-06,
+      "loss": 1.0551,
+      "step": 919
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9636143117040627e-06,
+      "loss": 2.1313,
+      "step": 920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9635724890738377e-06,
+      "loss": 0.0913,
+      "step": 921
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9635306664436127e-06,
+      "loss": 0.0214,
+      "step": 922
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9634888438133873e-06,
+      "loss": 0.3247,
+      "step": 923
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9634470211831623e-06,
+      "loss": 1.6262,
+      "step": 924
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963405198552937e-06,
+      "loss": 0.4027,
+      "step": 925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963363375922712e-06,
+      "loss": 0.0528,
+      "step": 926
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9633215532924865e-06,
+      "loss": 0.1467,
+      "step": 927
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9632797306622615e-06,
+      "loss": 0.1026,
+      "step": 928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963237908032036e-06,
+      "loss": 0.0156,
+      "step": 929
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9631960854018106e-06,
+      "loss": 0.0396,
+      "step": 930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9631542627715856e-06,
+      "loss": 0.0176,
+      "step": 931
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96311244014136e-06,
+      "loss": 1.2622,
+      "step": 932
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963070617511135e-06,
+      "loss": 0.0224,
+      "step": 933
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9630287948809098e-06,
+      "loss": 0.075,
+      "step": 934
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9629869722506848e-06,
+      "loss": 1.6879,
+      "step": 935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9629451496204598e-06,
+      "loss": 0.0302,
+      "step": 936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9629033269902343e-06,
+      "loss": 0.0195,
+      "step": 937
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9628615043600093e-06,
+      "loss": 1.6392,
+      "step": 938
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962819681729784e-06,
+      "loss": 0.0307,
+      "step": 939
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962777859099559e-06,
+      "loss": 1.0031,
+      "step": 940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9627360364693335e-06,
+      "loss": 0.3294,
+      "step": 941
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9626942138391085e-06,
+      "loss": 0.4531,
+      "step": 942
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962652391208883e-06,
+      "loss": 1.2205,
+      "step": 943
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9626105685786576e-06,
+      "loss": 0.0355,
+      "step": 944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9625687459484326e-06,
+      "loss": 0.0405,
+      "step": 945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962526923318207e-06,
+      "loss": 0.1033,
+      "step": 946
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962485100687982e-06,
+      "loss": 1.9566,
+      "step": 947
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9624432780577568e-06,
+      "loss": 0.2139,
+      "step": 948
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9624014554275318e-06,
+      "loss": 0.0246,
+      "step": 949
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9623596327973063e-06,
+      "loss": 0.427,
+      "step": 950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9623178101670813e-06,
+      "loss": 0.0154,
+      "step": 951
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9622759875368563e-06,
+      "loss": 0.0211,
+      "step": 952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962234164906631e-06,
+      "loss": 0.0723,
+      "step": 953
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962192342276406e-06,
+      "loss": 0.0186,
+      "step": 954
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9621505196461805e-06,
+      "loss": 0.1712,
+      "step": 955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9621086970159555e-06,
+      "loss": 0.0133,
+      "step": 956
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96206687438573e-06,
+      "loss": 0.1506,
+      "step": 957
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9620250517555046e-06,
+      "loss": 1.9656,
+      "step": 958
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9619832291252796e-06,
+      "loss": 0.055,
+      "step": 959
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961941406495054e-06,
+      "loss": 0.1895,
+      "step": 960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961899583864829e-06,
+      "loss": 0.0178,
+      "step": 961
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9618577612346038e-06,
+      "loss": 0.0131,
+      "step": 962
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9618159386043788e-06,
+      "loss": 0.2739,
+      "step": 963
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9617741159741533e-06,
+      "loss": 1.5569,
+      "step": 964
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9617322933439283e-06,
+      "loss": 0.0188,
+      "step": 965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9616904707137033e-06,
+      "loss": 0.019,
+      "step": 966
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961648648083478e-06,
+      "loss": 1.7032,
+      "step": 967
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961606825453253e-06,
+      "loss": 0.025,
+      "step": 968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9615650028230275e-06,
+      "loss": 0.0154,
+      "step": 969
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9615231801928025e-06,
+      "loss": 0.019,
+      "step": 970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961481357562577e-06,
+      "loss": 1.5439,
+      "step": 971
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9614395349323516e-06,
+      "loss": 0.1296,
+      "step": 972
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9613977123021266e-06,
+      "loss": 0.0172,
+      "step": 973
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961355889671901e-06,
+      "loss": 0.0334,
+      "step": 974
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961314067041676e-06,
+      "loss": 0.0224,
+      "step": 975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9612722444114508e-06,
+      "loss": 0.2636,
+      "step": 976
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9612304217812258e-06,
+      "loss": 0.0269,
+      "step": 977
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9611885991510004e-06,
+      "loss": 0.1121,
+      "step": 978
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9611467765207754e-06,
+      "loss": 1.6847,
+      "step": 979
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9611049538905503e-06,
+      "loss": 0.0245,
+      "step": 980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961063131260325e-06,
+      "loss": 0.0158,
+      "step": 981
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9610213086301e-06,
+      "loss": 0.1576,
+      "step": 982
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9609794859998745e-06,
+      "loss": 0.2329,
+      "step": 983
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960937663369649e-06,
+      "loss": 0.0151,
+      "step": 984
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960895840739424e-06,
+      "loss": 0.0118,
+      "step": 985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9608540181091986e-06,
+      "loss": 0.0218,
+      "step": 986
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9608121954789736e-06,
+      "loss": 1.1238,
+      "step": 987
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9607703728487482e-06,
+      "loss": 0.2865,
+      "step": 988
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9607285502185232e-06,
+      "loss": 1.5517,
+      "step": 989
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960686727588298e-06,
+      "loss": 0.0118,
+      "step": 990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9606449049580728e-06,
+      "loss": 0.0118,
+      "step": 991
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9606030823278474e-06,
+      "loss": 0.0167,
+      "step": 992
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9605612596976224e-06,
+      "loss": 0.0122,
+      "step": 993
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960519437067397e-06,
+      "loss": 0.0194,
+      "step": 994
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960477614437172e-06,
+      "loss": 0.0121,
+      "step": 995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960435791806947e-06,
+      "loss": 0.2644,
+      "step": 996
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9603939691767215e-06,
+      "loss": 0.01,
+      "step": 997
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960352146546496e-06,
+      "loss": 0.0169,
+      "step": 998
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960310323916271e-06,
+      "loss": 0.0148,
+      "step": 999
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9602685012860457e-06,
+      "loss": 0.0164,
+      "step": 1000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9602266786558206e-06,
+      "loss": 0.2948,
+      "step": 1001
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9601848560255952e-06,
+      "loss": 0.0158,
+      "step": 1002
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9601430333953702e-06,
+      "loss": 0.0162,
+      "step": 1003
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960101210765145e-06,
+      "loss": 2.1151,
+      "step": 1004
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96005938813492e-06,
+      "loss": 0.0142,
+      "step": 1005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9600175655046944e-06,
+      "loss": 0.011,
+      "step": 1006
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9599757428744694e-06,
+      "loss": 0.0114,
+      "step": 1007
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959933920244244e-06,
+      "loss": 0.0273,
+      "step": 1008
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959892097614019e-06,
+      "loss": 0.2126,
+      "step": 1009
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959850274983794e-06,
+      "loss": 0.0285,
+      "step": 1010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9598084523535685e-06,
+      "loss": 0.0113,
+      "step": 1011
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959766629723343e-06,
+      "loss": 2.2772,
+      "step": 1012
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959724807093118e-06,
+      "loss": 0.0752,
+      "step": 1013
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9596829844628927e-06,
+      "loss": 0.0146,
+      "step": 1014
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9596411618326677e-06,
+      "loss": 0.0187,
+      "step": 1015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9595993392024422e-06,
+      "loss": 0.0298,
+      "step": 1016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9595575165722172e-06,
+      "loss": 0.0103,
+      "step": 1017
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959515693941992e-06,
+      "loss": 0.1492,
+      "step": 1018
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959473871311767e-06,
+      "loss": 0.0227,
+      "step": 1019
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9594320486815414e-06,
+      "loss": 0.0253,
+      "step": 1020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9593902260513164e-06,
+      "loss": 0.0727,
+      "step": 1021
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959348403421091e-06,
+      "loss": 0.5857,
+      "step": 1022
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959306580790866e-06,
+      "loss": 0.0544,
+      "step": 1023
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959264758160641e-06,
+      "loss": 0.0115,
+      "step": 1024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9592229355304155e-06,
+      "loss": 0.2181,
+      "step": 1025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.95918111290019e-06,
+      "loss": 0.2568,
+      "step": 1026
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959139290269965e-06,
+      "loss": 0.0863,
+      "step": 1027
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9590974676397397e-06,
+      "loss": 2.1571,
+      "step": 1028
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9590556450095147e-06,
+      "loss": 0.0361,
+      "step": 1029
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9590138223792892e-06,
+      "loss": 0.0229,
+      "step": 1030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9589719997490642e-06,
+      "loss": 0.0205,
+      "step": 1031
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958930177118839e-06,
+      "loss": 0.0102,
+      "step": 1032
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958888354488614e-06,
+      "loss": 0.5255,
+      "step": 1033
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9588465318583884e-06,
+      "loss": 2.4861,
+      "step": 1034
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9588047092281634e-06,
+      "loss": 2.6569,
+      "step": 1035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958762886597938e-06,
+      "loss": 0.1551,
+      "step": 1036
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958721063967713e-06,
+      "loss": 0.0156,
+      "step": 1037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9586792413374875e-06,
+      "loss": 0.0434,
+      "step": 1038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9586374187072625e-06,
+      "loss": 0.0221,
+      "step": 1039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.958595596077037e-06,
+      "loss": 0.0217,
+      "step": 1040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.958553773446812e-06,
+      "loss": 0.0089,
+      "step": 1041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9585119508165867e-06,
+      "loss": 0.6925,
+      "step": 1042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9584701281863617e-06,
+      "loss": 0.1142,
+      "step": 1043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9584283055561363e-06,
+      "loss": 0.0177,
+      "step": 1044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9583864829259112e-06,
+      "loss": 0.0126,
+      "step": 1045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.958344660295686e-06,
+      "loss": 0.0097,
+      "step": 1046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.958302837665461e-06,
+      "loss": 0.0157,
+      "step": 1047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9582610150352354e-06,
+      "loss": 0.0863,
+      "step": 1048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9582191924050104e-06,
+      "loss": 0.3716,
+      "step": 1049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.958177369774785e-06,
+      "loss": 0.018,
+      "step": 1050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9581355471445595e-06,
+      "loss": 0.0168,
+      "step": 1051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9580937245143345e-06,
+      "loss": 2.553,
+      "step": 1052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9580519018841095e-06,
+      "loss": 0.0243,
+      "step": 1053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.958010079253884e-06,
+      "loss": 0.1269,
+      "step": 1054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957968256623659e-06,
+      "loss": 1.477,
+      "step": 1055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9579264339934337e-06,
+      "loss": 0.0138,
+      "step": 1056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9578846113632087e-06,
+      "loss": 0.0741,
+      "step": 1057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9578427887329833e-06,
+      "loss": 0.3423,
+      "step": 1058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9578009661027583e-06,
+      "loss": 0.0095,
+      "step": 1059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957759143472533e-06,
+      "loss": 0.2042,
+      "step": 1060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957717320842308e-06,
+      "loss": 1.7061,
+      "step": 1061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9576754982120824e-06,
+      "loss": 0.0093,
+      "step": 1062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9576336755818574e-06,
+      "loss": 0.0089,
+      "step": 1063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957591852951632e-06,
+      "loss": 0.0118,
+      "step": 1064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9575500303214066e-06,
+      "loss": 0.0204,
+      "step": 1065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9575082076911815e-06,
+      "loss": 0.0245,
+      "step": 1066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957466385060956e-06,
+      "loss": 1.2117,
+      "step": 1067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957424562430731e-06,
+      "loss": 0.6927,
+      "step": 1068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957382739800506e-06,
+      "loss": 0.0483,
+      "step": 1069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9573409171702807e-06,
+      "loss": 0.1687,
+      "step": 1070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9572990945400557e-06,
+      "loss": 0.0201,
+      "step": 1071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9572572719098303e-06,
+      "loss": 0.085,
+      "step": 1072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9572154492796053e-06,
+      "loss": 0.0096,
+      "step": 1073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95717362664938e-06,
+      "loss": 0.0145,
+      "step": 1074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957131804019155e-06,
+      "loss": 0.0134,
+      "step": 1075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9570899813889294e-06,
+      "loss": 0.2803,
+      "step": 1076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9570481587587044e-06,
+      "loss": 0.0131,
+      "step": 1077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.957006336128479e-06,
+      "loss": 1.5264,
+      "step": 1078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9569645134982536e-06,
+      "loss": 3.2552,
+      "step": 1079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9569226908680286e-06,
+      "loss": 0.0147,
+      "step": 1080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956880868237803e-06,
+      "loss": 0.2347,
+      "step": 1081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956839045607578e-06,
+      "loss": 0.0141,
+      "step": 1082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956797222977353e-06,
+      "loss": 0.3866,
+      "step": 1083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9567554003471277e-06,
+      "loss": 0.0125,
+      "step": 1084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9567135777169027e-06,
+      "loss": 2.2441,
+      "step": 1085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9566717550866773e-06,
+      "loss": 0.0286,
+      "step": 1086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9566299324564523e-06,
+      "loss": 0.0142,
+      "step": 1087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956588109826227e-06,
+      "loss": 0.023,
+      "step": 1088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956546287196002e-06,
+      "loss": 0.0161,
+      "step": 1089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9565044645657764e-06,
+      "loss": 0.0299,
+      "step": 1090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9564626419355514e-06,
+      "loss": 0.0151,
+      "step": 1091
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956420819305326e-06,
+      "loss": 0.3636,
+      "step": 1092
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9563789966751006e-06,
+      "loss": 0.0124,
+      "step": 1093
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9563371740448756e-06,
+      "loss": 0.584,
+      "step": 1094
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95629535141465e-06,
+      "loss": 0.0137,
+      "step": 1095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956253528784425e-06,
+      "loss": 0.3522,
+      "step": 1096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9562117061542e-06,
+      "loss": 0.0141,
+      "step": 1097
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9561698835239747e-06,
+      "loss": 0.0236,
+      "step": 1098
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9561280608937497e-06,
+      "loss": 0.0594,
+      "step": 1099
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9560862382635243e-06,
+      "loss": 0.0109,
+      "step": 1100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9560444156332993e-06,
+      "loss": 0.1666,
+      "step": 1101
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956002593003074e-06,
+      "loss": 0.0156,
+      "step": 1102
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955960770372849e-06,
+      "loss": 0.0182,
+      "step": 1103
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9559189477426234e-06,
+      "loss": 1.9622,
+      "step": 1104
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9558771251123984e-06,
+      "loss": 0.0128,
+      "step": 1105
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955835302482173e-06,
+      "loss": 0.013,
+      "step": 1106
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9557934798519476e-06,
+      "loss": 0.0146,
+      "step": 1107
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9557516572217226e-06,
+      "loss": 0.0103,
+      "step": 1108
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955709834591497e-06,
+      "loss": 0.0112,
+      "step": 1109
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955668011961272e-06,
+      "loss": 0.0228,
+      "step": 1110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9556261893310467e-06,
+      "loss": 0.015,
+      "step": 1111
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9555843667008217e-06,
+      "loss": 0.0184,
+      "step": 1112
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9555425440705967e-06,
+      "loss": 0.0173,
+      "step": 1113
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9555007214403713e-06,
+      "loss": 0.0163,
+      "step": 1114
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9554588988101463e-06,
+      "loss": 0.0453,
+      "step": 1115
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955417076179921e-06,
+      "loss": 2.0042,
+      "step": 1116
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955375253549696e-06,
+      "loss": 0.0203,
+      "step": 1117
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9553334309194704e-06,
+      "loss": 2.2064,
+      "step": 1118
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955291608289245e-06,
+      "loss": 0.0108,
+      "step": 1119
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95524978565902e-06,
+      "loss": 0.1398,
+      "step": 1120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9552079630287946e-06,
+      "loss": 1.6529,
+      "step": 1121
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9551661403985696e-06,
+      "loss": 1.6545,
+      "step": 1122
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955124317768344e-06,
+      "loss": 0.1758,
+      "step": 1123
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955082495138119e-06,
+      "loss": 0.9292,
+      "step": 1124
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9550406725078937e-06,
+      "loss": 0.1816,
+      "step": 1125
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9549988498776687e-06,
+      "loss": 0.2627,
+      "step": 1126
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9549570272474437e-06,
+      "loss": 0.0204,
+      "step": 1127
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9549152046172183e-06,
+      "loss": 0.1366,
+      "step": 1128
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9548733819869933e-06,
+      "loss": 0.9806,
+      "step": 1129
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954831559356768e-06,
+      "loss": 0.4512,
+      "step": 1130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954789736726543e-06,
+      "loss": 0.0071,
+      "step": 1131
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9547479140963174e-06,
+      "loss": 0.0118,
+      "step": 1132
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954706091466092e-06,
+      "loss": 2.3966,
+      "step": 1133
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954664268835867e-06,
+      "loss": 0.0582,
+      "step": 1134
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9546224462056416e-06,
+      "loss": 0.0157,
+      "step": 1135
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9545806235754166e-06,
+      "loss": 0.0125,
+      "step": 1136
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954538800945191e-06,
+      "loss": 0.017,
+      "step": 1137
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954496978314966e-06,
+      "loss": 0.0097,
+      "step": 1138
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9544551556847407e-06,
+      "loss": 0.2267,
+      "step": 1139
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9544133330545157e-06,
+      "loss": 0.0083,
+      "step": 1140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9543715104242907e-06,
+      "loss": 0.1135,
+      "step": 1141
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9543296877940653e-06,
+      "loss": 0.0096,
+      "step": 1142
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9542878651638403e-06,
+      "loss": 0.0231,
+      "step": 1143
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954246042533615e-06,
+      "loss": 0.0118,
+      "step": 1144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95420421990339e-06,
+      "loss": 0.0071,
+      "step": 1145
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9541623972731645e-06,
+      "loss": 0.0075,
+      "step": 1146
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954120574642939e-06,
+      "loss": 0.0258,
+      "step": 1147
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954078752012714e-06,
+      "loss": 2.5157,
+      "step": 1148
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9540369293824886e-06,
+      "loss": 0.2857,
+      "step": 1149
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9539951067522636e-06,
+      "loss": 0.0095,
+      "step": 1150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953953284122038e-06,
+      "loss": 0.1868,
+      "step": 1151
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953911461491813e-06,
+      "loss": 0.0081,
+      "step": 1152
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9538696388615877e-06,
+      "loss": 0.0123,
+      "step": 1153
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9538278162313627e-06,
+      "loss": 0.0173,
+      "step": 1154
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9537859936011373e-06,
+      "loss": 0.0101,
+      "step": 1155
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9537441709709123e-06,
+      "loss": 0.0112,
+      "step": 1156
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9537023483406873e-06,
+      "loss": 0.3522,
+      "step": 1157
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953660525710462e-06,
+      "loss": 0.0077,
+      "step": 1158
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953618703080237e-06,
+      "loss": 0.9213,
+      "step": 1159
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9535768804500115e-06,
+      "loss": 1.8957,
+      "step": 1160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953535057819786e-06,
+      "loss": 0.017,
+      "step": 1161
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953493235189561e-06,
+      "loss": 0.1136,
+      "step": 1162
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9534514125593356e-06,
+      "loss": 0.12,
+      "step": 1163
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9534095899291106e-06,
+      "loss": 0.0354,
+      "step": 1164
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953367767298885e-06,
+      "loss": 0.0603,
+      "step": 1165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95332594466866e-06,
+      "loss": 0.0175,
+      "step": 1166
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9532841220384348e-06,
+      "loss": 0.0105,
+      "step": 1167
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9532422994082098e-06,
+      "loss": 0.0096,
+      "step": 1168
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9532004767779843e-06,
+      "loss": 0.0145,
+      "step": 1169
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9531586541477593e-06,
+      "loss": 0.0116,
+      "step": 1170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9531168315175343e-06,
+      "loss": 0.0122,
+      "step": 1171
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953075008887309e-06,
+      "loss": 0.0104,
+      "step": 1172
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953033186257084e-06,
+      "loss": 0.4655,
+      "step": 1173
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9529913636268585e-06,
+      "loss": 0.0068,
+      "step": 1174
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952949540996633e-06,
+      "loss": 0.0089,
+      "step": 1175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952907718366408e-06,
+      "loss": 0.0074,
+      "step": 1176
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9528658957361826e-06,
+      "loss": 1.7369,
+      "step": 1177
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9528240731059576e-06,
+      "loss": 0.0179,
+      "step": 1178
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952782250475732e-06,
+      "loss": 0.1648,
+      "step": 1179
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952740427845507e-06,
+      "loss": 0.029,
+      "step": 1180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9526986052152818e-06,
+      "loss": 0.0093,
+      "step": 1181
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9526567825850568e-06,
+      "loss": 0.0286,
+      "step": 1182
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9526149599548313e-06,
+      "loss": 0.2882,
+      "step": 1183
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9525731373246063e-06,
+      "loss": 0.0113,
+      "step": 1184
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9525313146943813e-06,
+      "loss": 0.0093,
+      "step": 1185
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952489492064156e-06,
+      "loss": 0.0788,
+      "step": 1186
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9524476694339305e-06,
+      "loss": 0.0089,
+      "step": 1187
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9524058468037055e-06,
+      "loss": 0.0142,
+      "step": 1188
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95236402417348e-06,
+      "loss": 0.0085,
+      "step": 1189
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952322201543255e-06,
+      "loss": 0.0095,
+      "step": 1190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9522803789130296e-06,
+      "loss": 0.3679,
+      "step": 1191
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9522385562828046e-06,
+      "loss": 0.1174,
+      "step": 1192
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952196733652579e-06,
+      "loss": 0.3615,
+      "step": 1193
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952154911022354e-06,
+      "loss": 0.0082,
+      "step": 1194
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9521130883921288e-06,
+      "loss": 2.5721,
+      "step": 1195
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9520712657619038e-06,
+      "loss": 1.4125,
+      "step": 1196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9520294431316783e-06,
+      "loss": 0.3379,
+      "step": 1197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9519876205014533e-06,
+      "loss": 0.0083,
+      "step": 1198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951945797871228e-06,
+      "loss": 2.7726,
+      "step": 1199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951903975241003e-06,
+      "loss": 0.0084,
+      "step": 1200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9518621526107775e-06,
+      "loss": 0.179,
+      "step": 1201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9518203299805525e-06,
+      "loss": 0.011,
+      "step": 1202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951778507350327e-06,
+      "loss": 0.3995,
+      "step": 1203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951736684720102e-06,
+      "loss": 0.3986,
+      "step": 1204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9516948620898766e-06,
+      "loss": 0.0211,
+      "step": 1205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9516530394596516e-06,
+      "loss": 0.0112,
+      "step": 1206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951611216829426e-06,
+      "loss": 0.0537,
+      "step": 1207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951569394199201e-06,
+      "loss": 0.8235,
+      "step": 1208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9515275715689758e-06,
+      "loss": 2.0223,
+      "step": 1209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9514857489387508e-06,
+      "loss": 0.0154,
+      "step": 1210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9514439263085254e-06,
+      "loss": 0.2768,
+      "step": 1211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9514021036783004e-06,
+      "loss": 2.0057,
+      "step": 1212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951360281048075e-06,
+      "loss": 0.0089,
+      "step": 1213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9513184584178495e-06,
+      "loss": 0.0067,
+      "step": 1214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9512766357876245e-06,
+      "loss": 0.4285,
+      "step": 1215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9512348131573995e-06,
+      "loss": 0.0055,
+      "step": 1216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951192990527174e-06,
+      "loss": 1.9143,
+      "step": 1217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.951151167896949e-06,
+      "loss": 0.0144,
+      "step": 1218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9511093452667236e-06,
+      "loss": 2.0227,
+      "step": 1219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9510675226364986e-06,
+      "loss": 0.0108,
+      "step": 1220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9510257000062732e-06,
+      "loss": 0.119,
+      "step": 1221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9509838773760482e-06,
+      "loss": 0.0114,
+      "step": 1222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.950942054745823e-06,
+      "loss": 0.0205,
+      "step": 1223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.950900232115598e-06,
+      "loss": 0.0112,
+      "step": 1224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9508584094853724e-06,
+      "loss": 0.0081,
+      "step": 1225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9508165868551474e-06,
+      "loss": 1.2944,
+      "step": 1226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.950774764224922e-06,
+      "loss": 1.0797,
+      "step": 1227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9507329415946965e-06,
+      "loss": 0.0089,
+      "step": 1228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9506911189644715e-06,
+      "loss": 0.0118,
+      "step": 1229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9506492963342465e-06,
+      "loss": 0.3396,
+      "step": 1230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.950607473704021e-06,
+      "loss": 0.0065,
+      "step": 1231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.950565651073796e-06,
+      "loss": 0.0128,
+      "step": 1232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9505238284435707e-06,
+      "loss": 0.0152,
+      "step": 1233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9504820058133457e-06,
+      "loss": 1.931,
+      "step": 1234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9504401831831202e-06,
+      "loss": 0.0057,
+      "step": 1235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9503983605528952e-06,
+      "loss": 2.6745,
+      "step": 1236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.95035653792267e-06,
+      "loss": 0.4207,
+      "step": 1237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.950314715292445e-06,
+      "loss": 0.219,
+      "step": 1238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9502728926622194e-06,
+      "loss": 1.6831,
+      "step": 1239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9502310700319944e-06,
+      "loss": 0.0134,
+      "step": 1240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.950189247401769e-06,
+      "loss": 1.6038,
+      "step": 1241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9501474247715435e-06,
+      "loss": 1.6973,
+      "step": 1242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9501056021413185e-06,
+      "loss": 0.0147,
+      "step": 1243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9500637795110935e-06,
+      "loss": 0.0112,
+      "step": 1244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.950021956880868e-06,
+      "loss": 2.5849,
+      "step": 1245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949980134250643e-06,
+      "loss": 0.0108,
+      "step": 1246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9499383116204177e-06,
+      "loss": 0.0115,
+      "step": 1247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9498964889901927e-06,
+      "loss": 0.0091,
+      "step": 1248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9498546663599672e-06,
+      "loss": 0.015,
+      "step": 1249
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9498128437297422e-06,
+      "loss": 0.0233,
+      "step": 1250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949771021099517e-06,
+      "loss": 0.0072,
+      "step": 1251
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949729198469292e-06,
+      "loss": 0.0084,
+      "step": 1252
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9496873758390664e-06,
+      "loss": 0.0066,
+      "step": 1253
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949645553208841e-06,
+      "loss": 0.0088,
+      "step": 1254
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949603730578616e-06,
+      "loss": 0.0215,
+      "step": 1255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9495619079483905e-06,
+      "loss": 0.017,
+      "step": 1256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9495200853181655e-06,
+      "loss": 0.01,
+      "step": 1257
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9494782626879405e-06,
+      "loss": 0.0077,
+      "step": 1258
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949436440057715e-06,
+      "loss": 1.0765,
+      "step": 1259
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94939461742749e-06,
+      "loss": 0.0089,
+      "step": 1260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9493527947972647e-06,
+      "loss": 0.4734,
+      "step": 1261
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9493109721670397e-06,
+      "loss": 1.5231,
+      "step": 1262
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9492691495368142e-06,
+      "loss": 0.0083,
+      "step": 1263
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9492273269065892e-06,
+      "loss": 0.0066,
+      "step": 1264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949185504276364e-06,
+      "loss": 0.226,
+      "step": 1265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949143681646139e-06,
+      "loss": 0.0141,
+      "step": 1266
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9491018590159134e-06,
+      "loss": 0.0477,
+      "step": 1267
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949060036385688e-06,
+      "loss": 0.6357,
+      "step": 1268
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949018213755463e-06,
+      "loss": 0.2964,
+      "step": 1269
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9489763911252375e-06,
+      "loss": 1.4577,
+      "step": 1270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9489345684950125e-06,
+      "loss": 0.0067,
+      "step": 1271
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948892745864787e-06,
+      "loss": 0.0093,
+      "step": 1272
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948850923234562e-06,
+      "loss": 2.2898,
+      "step": 1273
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948809100604337e-06,
+      "loss": 1.9366,
+      "step": 1274
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9487672779741117e-06,
+      "loss": 0.1838,
+      "step": 1275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9487254553438867e-06,
+      "loss": 0.0326,
+      "step": 1276
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9486836327136613e-06,
+      "loss": 0.3617,
+      "step": 1277
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9486418100834363e-06,
+      "loss": 0.0067,
+      "step": 1278
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948599987453211e-06,
+      "loss": 1.3354,
+      "step": 1279
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948558164822986e-06,
+      "loss": 0.011,
+      "step": 1280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9485163421927604e-06,
+      "loss": 0.0121,
+      "step": 1281
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948474519562535e-06,
+      "loss": 1.6401,
+      "step": 1282
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94843269693231e-06,
+      "loss": 0.0081,
+      "step": 1283
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9483908743020845e-06,
+      "loss": 0.0091,
+      "step": 1284
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9483490516718595e-06,
+      "loss": 0.3321,
+      "step": 1285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948307229041634e-06,
+      "loss": 0.02,
+      "step": 1286
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948265406411409e-06,
+      "loss": 0.3085,
+      "step": 1287
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948223583781184e-06,
+      "loss": 1.0149,
+      "step": 1288
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9481817611509587e-06,
+      "loss": 0.0091,
+      "step": 1289
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9481399385207337e-06,
+      "loss": 0.0093,
+      "step": 1290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9480981158905083e-06,
+      "loss": 0.3912,
+      "step": 1291
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9480562932602833e-06,
+      "loss": 0.0077,
+      "step": 1292
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948014470630058e-06,
+      "loss": 0.0072,
+      "step": 1293
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947972647999833e-06,
+      "loss": 0.0069,
+      "step": 1294
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9479308253696074e-06,
+      "loss": 1.7212,
+      "step": 1295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947889002739382e-06,
+      "loss": 1.8413,
+      "step": 1296
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947847180109157e-06,
+      "loss": 2.9012,
+      "step": 1297
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9478053574789316e-06,
+      "loss": 0.0088,
+      "step": 1298
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9477635348487066e-06,
+      "loss": 0.0066,
+      "step": 1299
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947721712218481e-06,
+      "loss": 3.0073,
+      "step": 1300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947679889588256e-06,
+      "loss": 0.0125,
+      "step": 1301
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947638066958031e-06,
+      "loss": 0.0821,
+      "step": 1302
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9475962443278057e-06,
+      "loss": 0.0192,
+      "step": 1303
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9475544216975807e-06,
+      "loss": 0.0706,
+      "step": 1304
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9475125990673553e-06,
+      "loss": 0.0085,
+      "step": 1305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9474707764371303e-06,
+      "loss": 0.0357,
+      "step": 1306
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947428953806905e-06,
+      "loss": 0.1665,
+      "step": 1307
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94738713117668e-06,
+      "loss": 0.009,
+      "step": 1308
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9473453085464544e-06,
+      "loss": 0.0098,
+      "step": 1309
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947303485916229e-06,
+      "loss": 0.1757,
+      "step": 1310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947261663286004e-06,
+      "loss": 0.006,
+      "step": 1311
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9472198406557786e-06,
+      "loss": 0.1852,
+      "step": 1312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9471780180255536e-06,
+      "loss": 0.0101,
+      "step": 1313
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947136195395328e-06,
+      "loss": 0.0074,
+      "step": 1314
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947094372765103e-06,
+      "loss": 0.1364,
+      "step": 1315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9470525501348777e-06,
+      "loss": 0.0052,
+      "step": 1316
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9470107275046527e-06,
+      "loss": 0.0111,
+      "step": 1317
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9469689048744277e-06,
+      "loss": 0.0071,
+      "step": 1318
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9469270822442023e-06,
+      "loss": 0.2847,
+      "step": 1319
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9468852596139773e-06,
+      "loss": 0.0049,
+      "step": 1320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946843436983752e-06,
+      "loss": 0.0145,
+      "step": 1321
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9468016143535264e-06,
+      "loss": 0.0048,
+      "step": 1322
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9467597917233014e-06,
+      "loss": 2.7188,
+      "step": 1323
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946717969093076e-06,
+      "loss": 0.3814,
+      "step": 1324
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946676146462851e-06,
+      "loss": 1.697,
+      "step": 1325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9466343238326256e-06,
+      "loss": 1.988,
+      "step": 1326
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9465925012024006e-06,
+      "loss": 1.4148,
+      "step": 1327
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946550678572175e-06,
+      "loss": 1.8283,
+      "step": 1328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94650885594195e-06,
+      "loss": 0.0506,
+      "step": 1329
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9464670333117247e-06,
+      "loss": 0.2775,
+      "step": 1330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9464252106814997e-06,
+      "loss": 0.0184,
+      "step": 1331
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9463833880512747e-06,
+      "loss": 0.0071,
+      "step": 1332
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9463415654210493e-06,
+      "loss": 0.0096,
+      "step": 1333
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9462997427908243e-06,
+      "loss": 0.0167,
+      "step": 1334
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946257920160599e-06,
+      "loss": 0.718,
+      "step": 1335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9462160975303734e-06,
+      "loss": 0.0061,
+      "step": 1336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9461742749001484e-06,
+      "loss": 0.0081,
+      "step": 1337
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946132452269923e-06,
+      "loss": 1.6196,
+      "step": 1338
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946090629639698e-06,
+      "loss": 0.0074,
+      "step": 1339
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9460488070094726e-06,
+      "loss": 0.0085,
+      "step": 1340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9460069843792476e-06,
+      "loss": 2.2761,
+      "step": 1341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945965161749022e-06,
+      "loss": 0.0058,
+      "step": 1342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945923339118797e-06,
+      "loss": 0.0067,
+      "step": 1343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9458815164885717e-06,
+      "loss": 0.0093,
+      "step": 1344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9458396938583467e-06,
+      "loss": 0.009,
+      "step": 1345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9457978712281217e-06,
+      "loss": 0.0069,
+      "step": 1346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9457560485978963e-06,
+      "loss": 1.1138,
+      "step": 1347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9457142259676713e-06,
+      "loss": 0.0151,
+      "step": 1348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945672403337446e-06,
+      "loss": 0.1485,
+      "step": 1349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9456305807072204e-06,
+      "loss": 0.1477,
+      "step": 1350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9455887580769954e-06,
+      "loss": 0.0082,
+      "step": 1351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94554693544677e-06,
+      "loss": 0.0575,
+      "step": 1352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945505112816545e-06,
+      "loss": 0.0735,
+      "step": 1353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9454632901863196e-06,
+      "loss": 0.2453,
+      "step": 1354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9454214675560946e-06,
+      "loss": 0.076,
+      "step": 1355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945379644925869e-06,
+      "loss": 0.0126,
+      "step": 1356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.945337822295644e-06,
+      "loss": 0.0099,
+      "step": 1357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9452959996654187e-06,
+      "loss": 0.0081,
+      "step": 1358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9452541770351937e-06,
+      "loss": 0.0134,
+      "step": 1359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9452123544049683e-06,
+      "loss": 0.0079,
+      "step": 1360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9451705317747433e-06,
+      "loss": 0.2229,
+      "step": 1361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9451287091445183e-06,
+      "loss": 0.0986,
+      "step": 1362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.945086886514293e-06,
+      "loss": 1.1439,
+      "step": 1363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9450450638840675e-06,
+      "loss": 0.0143,
+      "step": 1364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9450032412538425e-06,
+      "loss": 0.0069,
+      "step": 1365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.944961418623617e-06,
+      "loss": 0.0062,
+      "step": 1366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.944919595993392e-06,
+      "loss": 0.0059,
+      "step": 1367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9448777733631666e-06,
+      "loss": 0.0073,
+      "step": 1368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9448359507329416e-06,
+      "loss": 0.5598,
+      "step": 1369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.944794128102716e-06,
+      "loss": 0.0199,
+      "step": 1370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.944752305472491e-06,
+      "loss": 0.0052,
+      "step": 1371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9447104828422657e-06,
+      "loss": 0.0048,
+      "step": 1372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9446686602120407e-06,
+      "loss": 0.0357,
+      "step": 1373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9446268375818153e-06,
+      "loss": 2.2764,
+      "step": 1374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9445850149515903e-06,
+      "loss": 2.6915,
+      "step": 1375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9445431923213653e-06,
+      "loss": 1.3379,
+      "step": 1376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.94450136969114e-06,
+      "loss": 0.0058,
+      "step": 1377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9444595470609145e-06,
+      "loss": 0.241,
+      "step": 1378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9444177244306895e-06,
+      "loss": 0.4367,
+      "step": 1379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.944375901800464e-06,
+      "loss": 0.0064,
+      "step": 1380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.944334079170239e-06,
+      "loss": 0.0182,
+      "step": 1381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9442922565400136e-06,
+      "loss": 0.0098,
+      "step": 1382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9442504339097886e-06,
+      "loss": 1.2899,
+      "step": 1383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.944208611279563e-06,
+      "loss": 0.0064,
+      "step": 1384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.944166788649338e-06,
+      "loss": 0.0122,
+      "step": 1385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9441249660191128e-06,
+      "loss": 0.0076,
+      "step": 1386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9440831433888877e-06,
+      "loss": 0.1688,
+      "step": 1387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9440413207586623e-06,
+      "loss": 1.6427,
+      "step": 1388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943999498128437e-06,
+      "loss": 3.2285,
+      "step": 1389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943957675498212e-06,
+      "loss": 0.0063,
+      "step": 1390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943915852867987e-06,
+      "loss": 0.0058,
+      "step": 1391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9438740302377615e-06,
+      "loss": 0.0062,
+      "step": 1392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9438322076075365e-06,
+      "loss": 2.4357,
+      "step": 1393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943790384977311e-06,
+      "loss": 0.0055,
+      "step": 1394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943748562347086e-06,
+      "loss": 0.0047,
+      "step": 1395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9437067397168606e-06,
+      "loss": 1.0988,
+      "step": 1396
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9436649170866356e-06,
+      "loss": 0.0195,
+      "step": 1397
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.94362309445641e-06,
+      "loss": 0.3329,
+      "step": 1398
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943581271826185e-06,
+      "loss": 2.5323,
+      "step": 1399
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9435394491959598e-06,
+      "loss": 0.0077,
+      "step": 1400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9434976265657348e-06,
+      "loss": 1.7447,
+      "step": 1401
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9434558039355093e-06,
+      "loss": 0.0054,
+      "step": 1402
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943413981305284e-06,
+      "loss": 0.0109,
+      "step": 1403
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943372158675059e-06,
+      "loss": 0.0122,
+      "step": 1404
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943330336044834e-06,
+      "loss": 0.0071,
+      "step": 1405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9432885134146085e-06,
+      "loss": 0.0169,
+      "step": 1406
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9432466907843835e-06,
+      "loss": 2.3584,
+      "step": 1407
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943204868154158e-06,
+      "loss": 0.0048,
+      "step": 1408
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943163045523933e-06,
+      "loss": 0.0057,
+      "step": 1409
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9431212228937076e-06,
+      "loss": 0.0087,
+      "step": 1410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9430794002634826e-06,
+      "loss": 0.025,
+      "step": 1411
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943037577633257e-06,
+      "loss": 0.0093,
+      "step": 1412
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942995755003032e-06,
+      "loss": 0.0047,
+      "step": 1413
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9429539323728068e-06,
+      "loss": 0.0064,
+      "step": 1414
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9429121097425818e-06,
+      "loss": 1.5165,
+      "step": 1415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9428702871123563e-06,
+      "loss": 2.9417,
+      "step": 1416
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942828464482131e-06,
+      "loss": 0.362,
+      "step": 1417
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942786641851906e-06,
+      "loss": 0.0932,
+      "step": 1418
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9427448192216805e-06,
+      "loss": 0.0052,
+      "step": 1419
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9427029965914555e-06,
+      "loss": 0.0055,
+      "step": 1420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9426611739612305e-06,
+      "loss": 0.0124,
+      "step": 1421
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942619351331005e-06,
+      "loss": 0.0059,
+      "step": 1422
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.94257752870078e-06,
+      "loss": 0.0075,
+      "step": 1423
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9425357060705546e-06,
+      "loss": 0.5342,
+      "step": 1424
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9424938834403296e-06,
+      "loss": 0.0051,
+      "step": 1425
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942452060810104e-06,
+      "loss": 0.006,
+      "step": 1426
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942410238179879e-06,
+      "loss": 0.0045,
+      "step": 1427
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9423684155496538e-06,
+      "loss": 0.0074,
+      "step": 1428
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9423265929194288e-06,
+      "loss": 2.4508,
+      "step": 1429
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9422847702892033e-06,
+      "loss": 0.0126,
+      "step": 1430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942242947658978e-06,
+      "loss": 0.1363,
+      "step": 1431
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942201125028753e-06,
+      "loss": 0.0102,
+      "step": 1432
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9421593023985275e-06,
+      "loss": 0.0068,
+      "step": 1433
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9421174797683025e-06,
+      "loss": 0.0065,
+      "step": 1434
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9420756571380775e-06,
+      "loss": 0.0155,
+      "step": 1435
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942033834507852e-06,
+      "loss": 1.3592,
+      "step": 1436
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941992011877627e-06,
+      "loss": 0.0371,
+      "step": 1437
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9419501892474016e-06,
+      "loss": 0.01,
+      "step": 1438
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9419083666171766e-06,
+      "loss": 0.0512,
+      "step": 1439
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9418665439869512e-06,
+      "loss": 0.007,
+      "step": 1440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941824721356726e-06,
+      "loss": 0.2101,
+      "step": 1441
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9417828987265008e-06,
+      "loss": 0.0046,
+      "step": 1442
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9417410760962758e-06,
+      "loss": 0.0091,
+      "step": 1443
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9416992534660504e-06,
+      "loss": 0.0058,
+      "step": 1444
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941657430835825e-06,
+      "loss": 0.0231,
+      "step": 1445
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9416156082056e-06,
+      "loss": 0.0136,
+      "step": 1446
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9415737855753745e-06,
+      "loss": 0.3111,
+      "step": 1447
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9415319629451495e-06,
+      "loss": 0.0046,
+      "step": 1448
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9414901403149245e-06,
+      "loss": 2.0892,
+      "step": 1449
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941448317684699e-06,
+      "loss": 0.0187,
+      "step": 1450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941406495054474e-06,
+      "loss": 0.0055,
+      "step": 1451
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9413646724242486e-06,
+      "loss": 1.1077,
+      "step": 1452
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9413228497940236e-06,
+      "loss": 1.8893,
+      "step": 1453
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9412810271637982e-06,
+      "loss": 0.0119,
+      "step": 1454
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9412392045335732e-06,
+      "loss": 0.0057,
+      "step": 1455
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941197381903348e-06,
+      "loss": 0.0047,
+      "step": 1456
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9411555592731224e-06,
+      "loss": 2.4997,
+      "step": 1457
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9411137366428974e-06,
+      "loss": 1.9,
+      "step": 1458
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941071914012672e-06,
+      "loss": 3.0626,
+      "step": 1459
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941030091382447e-06,
+      "loss": 1.1103,
+      "step": 1460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9409882687522215e-06,
+      "loss": 0.0056,
+      "step": 1461
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9409464461219965e-06,
+      "loss": 0.7339,
+      "step": 1462
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9409046234917715e-06,
+      "loss": 0.0116,
+      "step": 1463
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940862800861546e-06,
+      "loss": 0.0064,
+      "step": 1464
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940820978231321e-06,
+      "loss": 0.0083,
+      "step": 1465
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9407791556010957e-06,
+      "loss": 0.2039,
+      "step": 1466
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9407373329708707e-06,
+      "loss": 0.0049,
+      "step": 1467
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9406955103406452e-06,
+      "loss": 0.7216,
+      "step": 1468
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9406536877104202e-06,
+      "loss": 0.0116,
+      "step": 1469
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940611865080195e-06,
+      "loss": 0.0131,
+      "step": 1470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9405700424499694e-06,
+      "loss": 0.0055,
+      "step": 1471
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9405282198197444e-06,
+      "loss": 0.0071,
+      "step": 1472
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940486397189519e-06,
+      "loss": 0.0101,
+      "step": 1473
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940444574559294e-06,
+      "loss": 0.0102,
+      "step": 1474
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9404027519290685e-06,
+      "loss": 0.1095,
+      "step": 1475
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9403609292988435e-06,
+      "loss": 0.0237,
+      "step": 1476
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940319106668618e-06,
+      "loss": 0.0051,
+      "step": 1477
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940277284038393e-06,
+      "loss": 0.0044,
+      "step": 1478
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940235461408168e-06,
+      "loss": 0.0054,
+      "step": 1479
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9401936387779427e-06,
+      "loss": 0.0207,
+      "step": 1480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9401518161477177e-06,
+      "loss": 0.0113,
+      "step": 1481
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9401099935174922e-06,
+      "loss": 0.6602,
+      "step": 1482
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9400681708872672e-06,
+      "loss": 0.0126,
+      "step": 1483
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940026348257042e-06,
+      "loss": 0.0149,
+      "step": 1484
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9399845256268164e-06,
+      "loss": 0.1142,
+      "step": 1485
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9399427029965914e-06,
+      "loss": 0.0046,
+      "step": 1486
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939900880366366e-06,
+      "loss": 1.8568,
+      "step": 1487
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939859057736141e-06,
+      "loss": 0.0063,
+      "step": 1488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9398172351059155e-06,
+      "loss": 0.0036,
+      "step": 1489
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9397754124756905e-06,
+      "loss": 0.0276,
+      "step": 1490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939733589845465e-06,
+      "loss": 0.0598,
+      "step": 1491
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.93969176721524e-06,
+      "loss": 0.0086,
+      "step": 1492
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939649944585015e-06,
+      "loss": 0.0046,
+      "step": 1493
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9396081219547897e-06,
+      "loss": 0.0045,
+      "step": 1494
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9395662993245647e-06,
+      "loss": 0.0042,
+      "step": 1495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9395244766943392e-06,
+      "loss": 0.0045,
+      "step": 1496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9394826540641142e-06,
+      "loss": 2.9431,
+      "step": 1497
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939440831433889e-06,
+      "loss": 1.8087,
+      "step": 1498
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9393990088036634e-06,
+      "loss": 0.0039,
+      "step": 1499
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9393571861734384e-06,
+      "loss": 0.0093,
+      "step": 1500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939315363543213e-06,
+      "loss": 0.0082,
+      "step": 1501
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939273540912988e-06,
+      "loss": 0.006,
+      "step": 1502
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9392317182827625e-06,
+      "loss": 0.0052,
+      "step": 1503
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9391898956525375e-06,
+      "loss": 0.0048,
+      "step": 1504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939148073022312e-06,
+      "loss": 0.0119,
+      "step": 1505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939106250392087e-06,
+      "loss": 0.005,
+      "step": 1506
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939064427761862e-06,
+      "loss": 0.0034,
+      "step": 1507
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9390226051316367e-06,
+      "loss": 0.137,
+      "step": 1508
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9389807825014117e-06,
+      "loss": 0.0344,
+      "step": 1509
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9389389598711863e-06,
+      "loss": 2.5344,
+      "step": 1510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9388971372409613e-06,
+      "loss": 0.0041,
+      "step": 1511
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.938855314610736e-06,
+      "loss": 2.0821,
+      "step": 1512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9388134919805104e-06,
+      "loss": 0.0039,
+      "step": 1513
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9387716693502854e-06,
+      "loss": 0.0041,
+      "step": 1514
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.93872984672006e-06,
+      "loss": 0.0193,
+      "step": 1515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.938688024089835e-06,
+      "loss": 0.004,
+      "step": 1516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9386462014596095e-06,
+      "loss": 0.0181,
+      "step": 1517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9386043788293845e-06,
+      "loss": 0.0046,
+      "step": 1518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.938562556199159e-06,
+      "loss": 0.0059,
+      "step": 1519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.938520733568934e-06,
+      "loss": 0.4552,
+      "step": 1520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9384789109387087e-06,
+      "loss": 0.0052,
+      "step": 1521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9384370883084837e-06,
+      "loss": 0.0051,
+      "step": 1522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9383952656782587e-06,
+      "loss": 0.0049,
+      "step": 1523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9383534430480333e-06,
+      "loss": 1.7219,
+      "step": 1524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.938311620417808e-06,
+      "loss": 0.0071,
+      "step": 1525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.938269797787583e-06,
+      "loss": 2.2063,
+      "step": 1526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9382279751573574e-06,
+      "loss": 0.0056,
+      "step": 1527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9381861525271324e-06,
+      "loss": 0.0178,
+      "step": 1528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.938144329896907e-06,
+      "loss": 0.0042,
+      "step": 1529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.938102507266682e-06,
+      "loss": 0.0058,
+      "step": 1530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9380606846364566e-06,
+      "loss": 0.0074,
+      "step": 1531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9380188620062316e-06,
+      "loss": 0.0037,
+      "step": 1532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937977039376006e-06,
+      "loss": 0.0079,
+      "step": 1533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937935216745781e-06,
+      "loss": 1.6849,
+      "step": 1534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9378933941155557e-06,
+      "loss": 0.1659,
+      "step": 1535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9378515714853307e-06,
+      "loss": 3.236,
+      "step": 1536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9378097488551057e-06,
+      "loss": 1.4068,
+      "step": 1537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9377679262248803e-06,
+      "loss": 0.0054,
+      "step": 1538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937726103594655e-06,
+      "loss": 0.0107,
+      "step": 1539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93768428096443e-06,
+      "loss": 0.006,
+      "step": 1540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9376424583342044e-06,
+      "loss": 0.2978,
+      "step": 1541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9376006357039794e-06,
+      "loss": 0.0059,
+      "step": 1542
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937558813073754e-06,
+      "loss": 0.0071,
+      "step": 1543
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937516990443529e-06,
+      "loss": 0.0042,
+      "step": 1544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9374751678133036e-06,
+      "loss": 0.0056,
+      "step": 1545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9374333451830786e-06,
+      "loss": 0.1138,
+      "step": 1546
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937391522552853e-06,
+      "loss": 0.0239,
+      "step": 1547
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937349699922628e-06,
+      "loss": 0.0427,
+      "step": 1548
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9373078772924027e-06,
+      "loss": 1.7571,
+      "step": 1549
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9372660546621777e-06,
+      "loss": 0.004,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9372242320319527e-06,
+      "loss": 0.872,
+      "step": 1551
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9371824094017273e-06,
+      "loss": 0.0058,
+      "step": 1552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937140586771502e-06,
+      "loss": 0.0051,
+      "step": 1553
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.937098764141277e-06,
+      "loss": 0.007,
+      "step": 1554
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9370569415110514e-06,
+      "loss": 1.5389,
+      "step": 1555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9370151188808264e-06,
+      "loss": 0.0043,
+      "step": 1556
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936973296250601e-06,
+      "loss": 0.0071,
+      "step": 1557
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936931473620376e-06,
+      "loss": 0.0092,
+      "step": 1558
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9368896509901506e-06,
+      "loss": 1.4325,
+      "step": 1559
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9368478283599256e-06,
+      "loss": 0.0082,
+      "step": 1560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9368060057297e-06,
+      "loss": 0.007,
+      "step": 1561
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936764183099475e-06,
+      "loss": 0.1324,
+      "step": 1562
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9367223604692497e-06,
+      "loss": 1.5908,
+      "step": 1563
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9366805378390247e-06,
+      "loss": 0.0051,
+      "step": 1564
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9366387152087993e-06,
+      "loss": 0.0051,
+      "step": 1565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9365968925785743e-06,
+      "loss": 0.378,
+      "step": 1566
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936555069948349e-06,
+      "loss": 0.2905,
+      "step": 1567
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936513247318124e-06,
+      "loss": 0.0045,
+      "step": 1568
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9364714246878984e-06,
+      "loss": 0.1013,
+      "step": 1569
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9364296020576734e-06,
+      "loss": 0.0052,
+      "step": 1570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936387779427448e-06,
+      "loss": 3.2982,
+      "step": 1571
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936345956797223e-06,
+      "loss": 1.6049,
+      "step": 1572
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9363041341669976e-06,
+      "loss": 0.1102,
+      "step": 1573
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9362623115367726e-06,
+      "loss": 0.4397,
+      "step": 1574
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936220488906547e-06,
+      "loss": 0.5994,
+      "step": 1575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936178666276322e-06,
+      "loss": 0.006,
+      "step": 1576
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9361368436460967e-06,
+      "loss": 0.0058,
+      "step": 1577
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9360950210158717e-06,
+      "loss": 0.0055,
+      "step": 1578
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9360531983856463e-06,
+      "loss": 0.0631,
+      "step": 1579
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936011375755421e-06,
+      "loss": 0.0298,
+      "step": 1580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935969553125196e-06,
+      "loss": 0.0055,
+      "step": 1581
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935927730494971e-06,
+      "loss": 0.0039,
+      "step": 1582
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9358859078647454e-06,
+      "loss": 0.0042,
+      "step": 1583
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9358440852345204e-06,
+      "loss": 0.0126,
+      "step": 1584
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935802262604295e-06,
+      "loss": 0.0048,
+      "step": 1585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93576043997407e-06,
+      "loss": 0.0041,
+      "step": 1586
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9357186173438446e-06,
+      "loss": 0.3738,
+      "step": 1587
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9356767947136196e-06,
+      "loss": 0.0065,
+      "step": 1588
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935634972083394e-06,
+      "loss": 0.0567,
+      "step": 1589
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935593149453169e-06,
+      "loss": 0.1945,
+      "step": 1590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9355513268229437e-06,
+      "loss": 0.6363,
+      "step": 1591
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9355095041927183e-06,
+      "loss": 0.3094,
+      "step": 1592
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9354676815624933e-06,
+      "loss": 1.665,
+      "step": 1593
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935425858932268e-06,
+      "loss": 0.1346,
+      "step": 1594
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935384036302043e-06,
+      "loss": 0.004,
+      "step": 1595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935342213671818e-06,
+      "loss": 0.0036,
+      "step": 1596
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9353003910415925e-06,
+      "loss": 0.0044,
+      "step": 1597
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9352585684113675e-06,
+      "loss": 0.0041,
+      "step": 1598
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935216745781142e-06,
+      "loss": 1.8154,
+      "step": 1599
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935174923150917e-06,
+      "loss": 2.5609,
+      "step": 1600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9351331005206916e-06,
+      "loss": 1.2234,
+      "step": 1601
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9350912778904666e-06,
+      "loss": 0.1848,
+      "step": 1602
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935049455260241e-06,
+      "loss": 0.0117,
+      "step": 1603
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935007632630016e-06,
+      "loss": 0.0037,
+      "step": 1604
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9349658099997907e-06,
+      "loss": 0.0109,
+      "step": 1605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9349239873695653e-06,
+      "loss": 0.7501,
+      "step": 1606
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9348821647393403e-06,
+      "loss": 0.0048,
+      "step": 1607
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934840342109115e-06,
+      "loss": 0.5153,
+      "step": 1608
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93479851947889e-06,
+      "loss": 0.0038,
+      "step": 1609
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934756696848665e-06,
+      "loss": 0.005,
+      "step": 1610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9347148742184395e-06,
+      "loss": 2.4949,
+      "step": 1611
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9346730515882145e-06,
+      "loss": 0.0086,
+      "step": 1612
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934631228957989e-06,
+      "loss": 0.2295,
+      "step": 1613
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934589406327764e-06,
+      "loss": 0.0062,
+      "step": 1614
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9345475836975386e-06,
+      "loss": 0.0058,
+      "step": 1615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9345057610673136e-06,
+      "loss": 0.0064,
+      "step": 1616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934463938437088e-06,
+      "loss": 0.0047,
+      "step": 1617
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934422115806863e-06,
+      "loss": 0.0052,
+      "step": 1618
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9343802931766378e-06,
+      "loss": 1.6747,
+      "step": 1619
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9343384705464123e-06,
+      "loss": 0.005,
+      "step": 1620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9342966479161873e-06,
+      "loss": 2.3059,
+      "step": 1621
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934254825285962e-06,
+      "loss": 0.0043,
+      "step": 1622
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934213002655737e-06,
+      "loss": 0.2283,
+      "step": 1623
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9341711800255115e-06,
+      "loss": 0.004,
+      "step": 1624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9341293573952865e-06,
+      "loss": 1.6447,
+      "step": 1625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9340875347650615e-06,
+      "loss": 0.0033,
+      "step": 1626
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934045712134836e-06,
+      "loss": 0.2176,
+      "step": 1627
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934003889504611e-06,
+      "loss": 0.0077,
+      "step": 1628
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9339620668743856e-06,
+      "loss": 0.0055,
+      "step": 1629
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9339202442441606e-06,
+      "loss": 0.1296,
+      "step": 1630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933878421613935e-06,
+      "loss": 0.0048,
+      "step": 1631
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93383659898371e-06,
+      "loss": 0.1862,
+      "step": 1632
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9337947763534848e-06,
+      "loss": 0.0046,
+      "step": 1633
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9337529537232593e-06,
+      "loss": 0.5094,
+      "step": 1634
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9337111310930343e-06,
+      "loss": 0.0173,
+      "step": 1635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933669308462809e-06,
+      "loss": 0.0044,
+      "step": 1636
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933627485832584e-06,
+      "loss": 2.25,
+      "step": 1637
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9335856632023585e-06,
+      "loss": 0.3112,
+      "step": 1638
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9335438405721335e-06,
+      "loss": 0.1017,
+      "step": 1639
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9335020179419085e-06,
+      "loss": 0.5142,
+      "step": 1640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933460195311683e-06,
+      "loss": 1.0219,
+      "step": 1641
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933418372681458e-06,
+      "loss": 0.0172,
+      "step": 1642
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9333765500512326e-06,
+      "loss": 0.0053,
+      "step": 1643
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9333347274210076e-06,
+      "loss": 0.0047,
+      "step": 1644
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933292904790782e-06,
+      "loss": 0.0062,
+      "step": 1645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933251082160557e-06,
+      "loss": 0.0046,
+      "step": 1646
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9332092595303318e-06,
+      "loss": 0.0048,
+      "step": 1647
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9331674369001063e-06,
+      "loss": 0.0035,
+      "step": 1648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9331256142698813e-06,
+      "loss": 0.0042,
+      "step": 1649
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933083791639656e-06,
+      "loss": 0.2461,
+      "step": 1650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933041969009431e-06,
+      "loss": 0.0047,
+      "step": 1651
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9330001463792055e-06,
+      "loss": 0.6261,
+      "step": 1652
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9329583237489805e-06,
+      "loss": 0.0069,
+      "step": 1653
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9329165011187555e-06,
+      "loss": 0.0063,
+      "step": 1654
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93287467848853e-06,
+      "loss": 0.0053,
+      "step": 1655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932832855858305e-06,
+      "loss": 0.0035,
+      "step": 1656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9327910332280796e-06,
+      "loss": 0.3417,
+      "step": 1657
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9327492105978546e-06,
+      "loss": 1.6334,
+      "step": 1658
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932707387967629e-06,
+      "loss": 0.814,
+      "step": 1659
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9326655653374038e-06,
+      "loss": 0.0077,
+      "step": 1660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9326237427071788e-06,
+      "loss": 0.0041,
+      "step": 1661
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9325819200769534e-06,
+      "loss": 0.2611,
+      "step": 1662
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9325400974467284e-06,
+      "loss": 0.3224,
+      "step": 1663
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932498274816503e-06,
+      "loss": 0.0583,
+      "step": 1664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932456452186278e-06,
+      "loss": 0.0061,
+      "step": 1665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9324146295560525e-06,
+      "loss": 0.0091,
+      "step": 1666
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9323728069258275e-06,
+      "loss": 0.005,
+      "step": 1667
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9323309842956025e-06,
+      "loss": 0.0069,
+      "step": 1668
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932289161665377e-06,
+      "loss": 0.004,
+      "step": 1669
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932247339035152e-06,
+      "loss": 0.0161,
+      "step": 1670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9322055164049266e-06,
+      "loss": 1.8752,
+      "step": 1671
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9321636937747016e-06,
+      "loss": 0.0375,
+      "step": 1672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9321218711444762e-06,
+      "loss": 0.0056,
+      "step": 1673
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932080048514251e-06,
+      "loss": 0.0135,
+      "step": 1674
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932038225884026e-06,
+      "loss": 0.0093,
+      "step": 1675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9319964032538004e-06,
+      "loss": 0.2154,
+      "step": 1676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9319545806235754e-06,
+      "loss": 1.6984,
+      "step": 1677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.93191275799335e-06,
+      "loss": 1.4275,
+      "step": 1678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931870935363125e-06,
+      "loss": 1.4765,
+      "step": 1679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9318291127328995e-06,
+      "loss": 0.003,
+      "step": 1680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9317872901026745e-06,
+      "loss": 0.0034,
+      "step": 1681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931745467472449e-06,
+      "loss": 0.0095,
+      "step": 1682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931703644842224e-06,
+      "loss": 0.0036,
+      "step": 1683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931661822211999e-06,
+      "loss": 2.0973,
+      "step": 1684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9316199995817737e-06,
+      "loss": 1.713,
+      "step": 1685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9315781769515487e-06,
+      "loss": 0.0061,
+      "step": 1686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9315363543213232e-06,
+      "loss": 0.01,
+      "step": 1687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931494531691098e-06,
+      "loss": 0.0727,
+      "step": 1688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931452709060873e-06,
+      "loss": 0.0051,
+      "step": 1689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9314108864306474e-06,
+      "loss": 0.0099,
+      "step": 1690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9313690638004224e-06,
+      "loss": 0.0086,
+      "step": 1691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931327241170197e-06,
+      "loss": 0.3668,
+      "step": 1692
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931285418539972e-06,
+      "loss": 0.0844,
+      "step": 1693
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9312435959097465e-06,
+      "loss": 0.0061,
+      "step": 1694
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9312017732795215e-06,
+      "loss": 1.4017,
+      "step": 1695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931159950649296e-06,
+      "loss": 0.0042,
+      "step": 1696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931118128019071e-06,
+      "loss": 2.4339,
+      "step": 1697
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.931076305388846e-06,
+      "loss": 0.007,
+      "step": 1698
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9310344827586207e-06,
+      "loss": 1.2224,
+      "step": 1699
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9309926601283957e-06,
+      "loss": 0.0062,
+      "step": 1700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9309508374981702e-06,
+      "loss": 0.005,
+      "step": 1701
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930909014867945e-06,
+      "loss": 0.0045,
+      "step": 1702
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.93086719223772e-06,
+      "loss": 0.3745,
+      "step": 1703
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9308253696074944e-06,
+      "loss": 0.0044,
+      "step": 1704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9307835469772694e-06,
+      "loss": 0.2173,
+      "step": 1705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930741724347044e-06,
+      "loss": 2.7228,
+      "step": 1706
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930699901716819e-06,
+      "loss": 0.0054,
+      "step": 1707
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9306580790865935e-06,
+      "loss": 0.087,
+      "step": 1708
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9306162564563685e-06,
+      "loss": 0.0041,
+      "step": 1709
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930574433826143e-06,
+      "loss": 0.2286,
+      "step": 1710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930532611195918e-06,
+      "loss": 2.0611,
+      "step": 1711
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930490788565693e-06,
+      "loss": 0.0041,
+      "step": 1712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9304489659354677e-06,
+      "loss": 0.0063,
+      "step": 1713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9304071433052427e-06,
+      "loss": 0.0077,
+      "step": 1714
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9303653206750172e-06,
+      "loss": 0.0995,
+      "step": 1715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930323498044792e-06,
+      "loss": 3.6598,
+      "step": 1716
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930281675414567e-06,
+      "loss": 2.0059,
+      "step": 1717
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9302398527843414e-06,
+      "loss": 0.0252,
+      "step": 1718
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9301980301541164e-06,
+      "loss": 0.0042,
+      "step": 1719
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930156207523891e-06,
+      "loss": 0.0096,
+      "step": 1720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.930114384893666e-06,
+      "loss": 0.1615,
+      "step": 1721
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9300725622634405e-06,
+      "loss": 0.0063,
+      "step": 1722
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9300307396332155e-06,
+      "loss": 0.004,
+      "step": 1723
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92998891700299e-06,
+      "loss": 0.3251,
+      "step": 1724
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929947094372765e-06,
+      "loss": 0.004,
+      "step": 1725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9299052717425397e-06,
+      "loss": 0.0308,
+      "step": 1726
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9298634491123147e-06,
+      "loss": 0.005,
+      "step": 1727
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9298216264820893e-06,
+      "loss": 1.8078,
+      "step": 1728
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9297798038518643e-06,
+      "loss": 0.0266,
+      "step": 1729
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929737981221639e-06,
+      "loss": 0.0094,
+      "step": 1730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929696158591414e-06,
+      "loss": 0.0046,
+      "step": 1731
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9296543359611884e-06,
+      "loss": 0.007,
+      "step": 1732
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9296125133309634e-06,
+      "loss": 0.5544,
+      "step": 1733
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929570690700738e-06,
+      "loss": 0.793,
+      "step": 1734
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929528868070513e-06,
+      "loss": 0.0047,
+      "step": 1735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9294870454402875e-06,
+      "loss": 0.0038,
+      "step": 1736
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9294452228100625e-06,
+      "loss": 0.0062,
+      "step": 1737
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929403400179837e-06,
+      "loss": 0.0159,
+      "step": 1738
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929361577549612e-06,
+      "loss": 1.7069,
+      "step": 1739
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9293197549193867e-06,
+      "loss": 0.4586,
+      "step": 1740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9292779322891613e-06,
+      "loss": 0.0058,
+      "step": 1741
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9292361096589363e-06,
+      "loss": 0.0212,
+      "step": 1742
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9291942870287113e-06,
+      "loss": 0.0039,
+      "step": 1743
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929152464398486e-06,
+      "loss": 0.0032,
+      "step": 1744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929110641768261e-06,
+      "loss": 0.1408,
+      "step": 1745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9290688191380354e-06,
+      "loss": 0.0041,
+      "step": 1746
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9290269965078104e-06,
+      "loss": 0.007,
+      "step": 1747
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928985173877585e-06,
+      "loss": 0.3304,
+      "step": 1748
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92894335124736e-06,
+      "loss": 0.0052,
+      "step": 1749
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9289015286171346e-06,
+      "loss": 0.0074,
+      "step": 1750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9288597059869095e-06,
+      "loss": 0.0056,
+      "step": 1751
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928817883356684e-06,
+      "loss": 0.004,
+      "step": 1752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928776060726459e-06,
+      "loss": 0.8225,
+      "step": 1753
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9287342380962337e-06,
+      "loss": 0.0028,
+      "step": 1754
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9286924154660083e-06,
+      "loss": 0.0034,
+      "step": 1755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9286505928357833e-06,
+      "loss": 0.0029,
+      "step": 1756
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9286087702055583e-06,
+      "loss": 0.0037,
+      "step": 1757
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928566947575333e-06,
+      "loss": 0.0058,
+      "step": 1758
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928525124945108e-06,
+      "loss": 0.3322,
+      "step": 1759
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9284833023148824e-06,
+      "loss": 0.062,
+      "step": 1760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9284414796846574e-06,
+      "loss": 0.0052,
+      "step": 1761
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928399657054432e-06,
+      "loss": 0.0137,
+      "step": 1762
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928357834424207e-06,
+      "loss": 0.0136,
+      "step": 1763
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9283160117939816e-06,
+      "loss": 0.004,
+      "step": 1764
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9282741891637566e-06,
+      "loss": 0.0064,
+      "step": 1765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928232366533531e-06,
+      "loss": 0.0033,
+      "step": 1766
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928190543903306e-06,
+      "loss": 0.5754,
+      "step": 1767
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9281487212730807e-06,
+      "loss": 1.9911,
+      "step": 1768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9281068986428553e-06,
+      "loss": 0.005,
+      "step": 1769
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9280650760126303e-06,
+      "loss": 0.0059,
+      "step": 1770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9280232533824053e-06,
+      "loss": 0.0162,
+      "step": 1771
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92798143075218e-06,
+      "loss": 0.0038,
+      "step": 1772
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927939608121955e-06,
+      "loss": 0.0607,
+      "step": 1773
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9278977854917294e-06,
+      "loss": 0.3681,
+      "step": 1774
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9278559628615044e-06,
+      "loss": 3.6661,
+      "step": 1775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927814140231279e-06,
+      "loss": 0.0047,
+      "step": 1776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927772317601054e-06,
+      "loss": 0.381,
+      "step": 1777
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9277304949708286e-06,
+      "loss": 0.0036,
+      "step": 1778
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9276886723406036e-06,
+      "loss": 0.2115,
+      "step": 1779
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927646849710378e-06,
+      "loss": 1.6571,
+      "step": 1780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927605027080153e-06,
+      "loss": 0.0043,
+      "step": 1781
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9275632044499277e-06,
+      "loss": 0.064,
+      "step": 1782
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9275213818197023e-06,
+      "loss": 2.72,
+      "step": 1783
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9274795591894773e-06,
+      "loss": 0.008,
+      "step": 1784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927437736559252e-06,
+      "loss": 0.3399,
+      "step": 1785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927395913929027e-06,
+      "loss": 0.0038,
+      "step": 1786
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927354091298802e-06,
+      "loss": 0.104,
+      "step": 1787
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9273122686685764e-06,
+      "loss": 0.0066,
+      "step": 1788
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9272704460383514e-06,
+      "loss": 0.0043,
+      "step": 1789
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927228623408126e-06,
+      "loss": 0.007,
+      "step": 1790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927186800777901e-06,
+      "loss": 1.47,
+      "step": 1791
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9271449781476756e-06,
+      "loss": 0.0037,
+      "step": 1792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9271031555174506e-06,
+      "loss": 0.0037,
+      "step": 1793
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927061332887225e-06,
+      "loss": 2.4726,
+      "step": 1794
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9270195102569997e-06,
+      "loss": 0.0043,
+      "step": 1795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9269776876267747e-06,
+      "loss": 0.3081,
+      "step": 1796
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9269358649965493e-06,
+      "loss": 0.0039,
+      "step": 1797
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9268940423663243e-06,
+      "loss": 0.0038,
+      "step": 1798
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926852219736099e-06,
+      "loss": 0.0122,
+      "step": 1799
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926810397105874e-06,
+      "loss": 2.2706,
+      "step": 1800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926768574475649e-06,
+      "loss": 0.0038,
+      "step": 1801
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9267267518454234e-06,
+      "loss": 0.0064,
+      "step": 1802
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9266849292151984e-06,
+      "loss": 0.1058,
+      "step": 1803
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926643106584973e-06,
+      "loss": 0.0053,
+      "step": 1804
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926601283954748e-06,
+      "loss": 3.2463,
+      "step": 1805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9265594613245226e-06,
+      "loss": 0.0034,
+      "step": 1806
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9265176386942976e-06,
+      "loss": 0.005,
+      "step": 1807
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926475816064072e-06,
+      "loss": 1.9213,
+      "step": 1808
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9264339934338467e-06,
+      "loss": 1.0849,
+      "step": 1809
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9263921708036217e-06,
+      "loss": 0.0035,
+      "step": 1810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9263503481733963e-06,
+      "loss": 0.0034,
+      "step": 1811
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9263085255431713e-06,
+      "loss": 1.2402,
+      "step": 1812
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926266702912946e-06,
+      "loss": 0.0341,
+      "step": 1813
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926224880282721e-06,
+      "loss": 0.0162,
+      "step": 1814
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926183057652496e-06,
+      "loss": 0.0042,
+      "step": 1815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9261412350222704e-06,
+      "loss": 0.0037,
+      "step": 1816
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9260994123920454e-06,
+      "loss": 1.0083,
+      "step": 1817
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92605758976182e-06,
+      "loss": 2.6736,
+      "step": 1818
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926015767131595e-06,
+      "loss": 0.0033,
+      "step": 1819
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9259739445013696e-06,
+      "loss": 0.0037,
+      "step": 1820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9259321218711446e-06,
+      "loss": 0.007,
+      "step": 1821
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925890299240919e-06,
+      "loss": 0.0031,
+      "step": 1822
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9258484766106937e-06,
+      "loss": 0.0304,
+      "step": 1823
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9258066539804687e-06,
+      "loss": 0.005,
+      "step": 1824
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9257648313502433e-06,
+      "loss": 0.185,
+      "step": 1825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9257230087200183e-06,
+      "loss": 0.4723,
+      "step": 1826
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925681186089793e-06,
+      "loss": 0.535,
+      "step": 1827
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925639363459568e-06,
+      "loss": 2.2735,
+      "step": 1828
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9255975408293425e-06,
+      "loss": 0.0674,
+      "step": 1829
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9255557181991175e-06,
+      "loss": 1.6025,
+      "step": 1830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9255138955688925e-06,
+      "loss": 0.0055,
+      "step": 1831
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925472072938667e-06,
+      "loss": 0.0063,
+      "step": 1832
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925430250308442e-06,
+      "loss": 0.0068,
+      "step": 1833
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9253884276782166e-06,
+      "loss": 0.004,
+      "step": 1834
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9253466050479916e-06,
+      "loss": 0.0037,
+      "step": 1835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925304782417766e-06,
+      "loss": 0.0051,
+      "step": 1836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9252629597875408e-06,
+      "loss": 0.009,
+      "step": 1837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9252211371573157e-06,
+      "loss": 0.0063,
+      "step": 1838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9251793145270903e-06,
+      "loss": 0.0057,
+      "step": 1839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9251374918968653e-06,
+      "loss": 0.0448,
+      "step": 1840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92509566926664e-06,
+      "loss": 0.0037,
+      "step": 1841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.925053846636415e-06,
+      "loss": 0.0042,
+      "step": 1842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9250120240061895e-06,
+      "loss": 0.0038,
+      "step": 1843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9249702013759645e-06,
+      "loss": 0.0265,
+      "step": 1844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9249283787457395e-06,
+      "loss": 0.0035,
+      "step": 1845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924886556115514e-06,
+      "loss": 0.011,
+      "step": 1846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924844733485289e-06,
+      "loss": 2.3678,
+      "step": 1847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9248029108550636e-06,
+      "loss": 2.1149,
+      "step": 1848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9247610882248386e-06,
+      "loss": 0.0053,
+      "step": 1849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924719265594613e-06,
+      "loss": 0.1818,
+      "step": 1850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9246774429643878e-06,
+      "loss": 2.0326,
+      "step": 1851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9246356203341628e-06,
+      "loss": 0.1663,
+      "step": 1852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9245937977039373e-06,
+      "loss": 0.7054,
+      "step": 1853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9245519750737123e-06,
+      "loss": 0.0108,
+      "step": 1854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924510152443487e-06,
+      "loss": 0.0043,
+      "step": 1855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924468329813262e-06,
+      "loss": 0.0038,
+      "step": 1856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9244265071830365e-06,
+      "loss": 0.0035,
+      "step": 1857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9243846845528115e-06,
+      "loss": 0.003,
+      "step": 1858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9243428619225865e-06,
+      "loss": 0.0057,
+      "step": 1859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924301039292361e-06,
+      "loss": 0.0048,
+      "step": 1860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924259216662136e-06,
+      "loss": 0.0038,
+      "step": 1861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9242173940319106e-06,
+      "loss": 0.0862,
+      "step": 1862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.924175571401685e-06,
+      "loss": 0.3708,
+      "step": 1863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92413374877146e-06,
+      "loss": 0.089,
+      "step": 1864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9240919261412348e-06,
+      "loss": 0.0039,
+      "step": 1865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9240501035110098e-06,
+      "loss": 0.0155,
+      "step": 1866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9240082808807843e-06,
+      "loss": 0.0077,
+      "step": 1867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9239664582505593e-06,
+      "loss": 0.0058,
+      "step": 1868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923924635620334e-06,
+      "loss": 0.0064,
+      "step": 1869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923882812990109e-06,
+      "loss": 0.0037,
+      "step": 1870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9238409903598835e-06,
+      "loss": 0.0049,
+      "step": 1871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9237991677296585e-06,
+      "loss": 0.0068,
+      "step": 1872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9237573450994335e-06,
+      "loss": 0.013,
+      "step": 1873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923715522469208e-06,
+      "loss": 0.0027,
+      "step": 1874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923673699838983e-06,
+      "loss": 0.0048,
+      "step": 1875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9236318772087576e-06,
+      "loss": 0.0051,
+      "step": 1876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923590054578532e-06,
+      "loss": 0.0031,
+      "step": 1877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923548231948307e-06,
+      "loss": 0.0034,
+      "step": 1878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9235064093180818e-06,
+      "loss": 3.2526,
+      "step": 1879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9234645866878568e-06,
+      "loss": 0.0029,
+      "step": 1880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9234227640576313e-06,
+      "loss": 0.0041,
+      "step": 1881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9233809414274063e-06,
+      "loss": 0.0031,
+      "step": 1882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923339118797181e-06,
+      "loss": 1.0265,
+      "step": 1883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923297296166956e-06,
+      "loss": 0.0109,
+      "step": 1884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9232554735367305e-06,
+      "loss": 0.0077,
+      "step": 1885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9232136509065055e-06,
+      "loss": 0.0068,
+      "step": 1886
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92317182827628e-06,
+      "loss": 0.0341,
+      "step": 1887
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923130005646055e-06,
+      "loss": 0.0035,
+      "step": 1888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92308818301583e-06,
+      "loss": 0.0059,
+      "step": 1889
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9230463603856046e-06,
+      "loss": 0.0044,
+      "step": 1890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9230045377553792e-06,
+      "loss": 3.1969,
+      "step": 1891
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922962715125154e-06,
+      "loss": 0.0144,
+      "step": 1892
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9229208924949288e-06,
+      "loss": 0.0071,
+      "step": 1893
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9228790698647038e-06,
+      "loss": 0.0327,
+      "step": 1894
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9228372472344784e-06,
+      "loss": 0.0033,
+      "step": 1895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9227954246042534e-06,
+      "loss": 0.0069,
+      "step": 1896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922753601974028e-06,
+      "loss": 2.4644,
+      "step": 1897
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922711779343803e-06,
+      "loss": 0.0038,
+      "step": 1898
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9226699567135775e-06,
+      "loss": 0.0147,
+      "step": 1899
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9226281340833525e-06,
+      "loss": 0.0238,
+      "step": 1900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922586311453127e-06,
+      "loss": 0.0028,
+      "step": 1901
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922544488822902e-06,
+      "loss": 0.9473,
+      "step": 1902
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922502666192677e-06,
+      "loss": 0.0028,
+      "step": 1903
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9224608435624516e-06,
+      "loss": 0.003,
+      "step": 1904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9224190209322262e-06,
+      "loss": 0.0033,
+      "step": 1905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9223771983020012e-06,
+      "loss": 0.0035,
+      "step": 1906
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922335375671776e-06,
+      "loss": 3.0567,
+      "step": 1907
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922293553041551e-06,
+      "loss": 0.0038,
+      "step": 1908
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9222517304113254e-06,
+      "loss": 0.4174,
+      "step": 1909
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9222099077811004e-06,
+      "loss": 0.0064,
+      "step": 1910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922168085150875e-06,
+      "loss": 0.0042,
+      "step": 1911
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92212626252065e-06,
+      "loss": 0.218,
+      "step": 1912
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9220844398904245e-06,
+      "loss": 0.0068,
+      "step": 1913
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9220426172601995e-06,
+      "loss": 0.0063,
+      "step": 1914
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922000794629974e-06,
+      "loss": 0.0065,
+      "step": 1915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921958971999749e-06,
+      "loss": 0.0175,
+      "step": 1916
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921917149369524e-06,
+      "loss": 0.0027,
+      "step": 1917
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9218753267392987e-06,
+      "loss": 0.5999,
+      "step": 1918
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9218335041090732e-06,
+      "loss": 0.0043,
+      "step": 1919
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9217916814788482e-06,
+      "loss": 0.4871,
+      "step": 1920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921749858848623e-06,
+      "loss": 0.6999,
+      "step": 1921
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921708036218398e-06,
+      "loss": 0.0906,
+      "step": 1922
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9216662135881724e-06,
+      "loss": 0.0036,
+      "step": 1923
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9216243909579474e-06,
+      "loss": 3.2755,
+      "step": 1924
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921582568327722e-06,
+      "loss": 0.0062,
+      "step": 1925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921540745697497e-06,
+      "loss": 0.0073,
+      "step": 1926
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9214989230672715e-06,
+      "loss": 0.3031,
+      "step": 1927
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9214571004370465e-06,
+      "loss": 0.0031,
+      "step": 1928
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921415277806821e-06,
+      "loss": 2.9151,
+      "step": 1929
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921373455176596e-06,
+      "loss": 2.1951,
+      "step": 1930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9213316325463707e-06,
+      "loss": 0.0043,
+      "step": 1931
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9212898099161457e-06,
+      "loss": 1.5156,
+      "step": 1932
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9212479872859202e-06,
+      "loss": 0.0962,
+      "step": 1933
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9212061646556952e-06,
+      "loss": 0.0047,
+      "step": 1934
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92116434202547e-06,
+      "loss": 0.0154,
+      "step": 1935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921122519395245e-06,
+      "loss": 0.0106,
+      "step": 1936
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9210806967650194e-06,
+      "loss": 0.1281,
+      "step": 1937
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9210388741347944e-06,
+      "loss": 2.7307,
+      "step": 1938
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920997051504569e-06,
+      "loss": 0.0034,
+      "step": 1939
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920955228874344e-06,
+      "loss": 0.4357,
+      "step": 1940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9209134062441185e-06,
+      "loss": 1.0936,
+      "step": 1941
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9208715836138935e-06,
+      "loss": 0.0033,
+      "step": 1942
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920829760983668e-06,
+      "loss": 0.004,
+      "step": 1943
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9207879383534427e-06,
+      "loss": 0.0049,
+      "step": 1944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9207461157232177e-06,
+      "loss": 0.0065,
+      "step": 1945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9207042930929922e-06,
+      "loss": 1.7149,
+      "step": 1946
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9206624704627672e-06,
+      "loss": 0.0855,
+      "step": 1947
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9206206478325422e-06,
+      "loss": 0.1523,
+      "step": 1948
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920578825202317e-06,
+      "loss": 0.0205,
+      "step": 1949
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920537002572092e-06,
+      "loss": 0.0387,
+      "step": 1950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9204951799418664e-06,
+      "loss": 0.0067,
+      "step": 1951
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9204533573116414e-06,
+      "loss": 0.3185,
+      "step": 1952
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920411534681416e-06,
+      "loss": 0.003,
+      "step": 1953
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920369712051191e-06,
+      "loss": 0.0066,
+      "step": 1954
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9203278894209655e-06,
+      "loss": 1.6268,
+      "step": 1955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9202860667907405e-06,
+      "loss": 0.6705,
+      "step": 1956
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920244244160515e-06,
+      "loss": 0.0093,
+      "step": 1957
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9202024215302897e-06,
+      "loss": 0.0137,
+      "step": 1958
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9201605989000647e-06,
+      "loss": 0.3683,
+      "step": 1959
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9201187762698393e-06,
+      "loss": 0.0094,
+      "step": 1960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9200769536396143e-06,
+      "loss": 0.4098,
+      "step": 1961
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9200351310093893e-06,
+      "loss": 0.0059,
+      "step": 1962
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919993308379164e-06,
+      "loss": 0.0059,
+      "step": 1963
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919951485748939e-06,
+      "loss": 0.003,
+      "step": 1964
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9199096631187134e-06,
+      "loss": 2.1809,
+      "step": 1965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9198678404884884e-06,
+      "loss": 0.0029,
+      "step": 1966
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919826017858263e-06,
+      "loss": 1.6433,
+      "step": 1967
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919784195228038e-06,
+      "loss": 0.0039,
+      "step": 1968
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9197423725978125e-06,
+      "loss": 1.2248,
+      "step": 1969
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9197005499675875e-06,
+      "loss": 0.0179,
+      "step": 1970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919658727337362e-06,
+      "loss": 0.1496,
+      "step": 1971
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9196169047071367e-06,
+      "loss": 0.0043,
+      "step": 1972
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9195750820769117e-06,
+      "loss": 0.0028,
+      "step": 1973
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9195332594466863e-06,
+      "loss": 0.0151,
+      "step": 1974
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9194914368164613e-06,
+      "loss": 0.0039,
+      "step": 1975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9194496141862363e-06,
+      "loss": 0.0031,
+      "step": 1976
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919407791556011e-06,
+      "loss": 0.0028,
+      "step": 1977
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919365968925786e-06,
+      "loss": 0.0031,
+      "step": 1978
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9193241462955604e-06,
+      "loss": 0.0054,
+      "step": 1979
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9192823236653354e-06,
+      "loss": 0.0047,
+      "step": 1980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.91924050103511e-06,
+      "loss": 0.0077,
+      "step": 1981
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919198678404885e-06,
+      "loss": 2.2539,
+      "step": 1982
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9191568557746596e-06,
+      "loss": 0.0041,
+      "step": 1983
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9191150331444346e-06,
+      "loss": 2.6581,
+      "step": 1984
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919073210514209e-06,
+      "loss": 0.003,
+      "step": 1985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9190313878839837e-06,
+      "loss": 1.1201,
+      "step": 1986
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9189895652537587e-06,
+      "loss": 0.0077,
+      "step": 1987
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9189477426235333e-06,
+      "loss": 0.0027,
+      "step": 1988
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9189059199933083e-06,
+      "loss": 1.8715,
+      "step": 1989
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918864097363083e-06,
+      "loss": 0.2704,
+      "step": 1990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918822274732858e-06,
+      "loss": 0.0051,
+      "step": 1991
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918780452102633e-06,
+      "loss": 0.0041,
+      "step": 1992
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9187386294724074e-06,
+      "loss": 1.5984,
+      "step": 1993
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9186968068421824e-06,
+      "loss": 1.7397,
+      "step": 1994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918654984211957e-06,
+      "loss": 0.0732,
+      "step": 1995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918613161581732e-06,
+      "loss": 0.0031,
+      "step": 1996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9185713389515066e-06,
+      "loss": 0.0041,
+      "step": 1997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918529516321281e-06,
+      "loss": 0.0041,
+      "step": 1998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918487693691056e-06,
+      "loss": 0.0038,
+      "step": 1999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9184458710608307e-06,
+      "loss": 0.005,
+      "step": 2000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9184040484306057e-06,
+      "loss": 0.0036,
+      "step": 2001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9183622258003803e-06,
+      "loss": 1.8543,
+      "step": 2002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9183204031701553e-06,
+      "loss": 1.4129,
+      "step": 2003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91827858053993e-06,
+      "loss": 2.0413,
+      "step": 2004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918236757909705e-06,
+      "loss": 0.1493,
+      "step": 2005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91819493527948e-06,
+      "loss": 0.0032,
+      "step": 2006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9181531126492544e-06,
+      "loss": 0.011,
+      "step": 2007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9181112900190294e-06,
+      "loss": 0.3188,
+      "step": 2008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918069467388804e-06,
+      "loss": 0.006,
+      "step": 2009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.918027644758579e-06,
+      "loss": 0.0295,
+      "step": 2010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9179858221283536e-06,
+      "loss": 0.0065,
+      "step": 2011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917943999498128e-06,
+      "loss": 0.0034,
+      "step": 2012
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917902176867903e-06,
+      "loss": 0.0067,
+      "step": 2013
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9178603542376777e-06,
+      "loss": 0.0029,
+      "step": 2014
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9178185316074527e-06,
+      "loss": 0.0066,
+      "step": 2015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9177767089772273e-06,
+      "loss": 0.0034,
+      "step": 2016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9177348863470023e-06,
+      "loss": 1.3149,
+      "step": 2017
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917693063716777e-06,
+      "loss": 0.0034,
+      "step": 2018
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917651241086552e-06,
+      "loss": 0.004,
+      "step": 2019
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917609418456327e-06,
+      "loss": 0.0041,
+      "step": 2020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9175675958261014e-06,
+      "loss": 0.411,
+      "step": 2021
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9175257731958764e-06,
+      "loss": 0.0035,
+      "step": 2022
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917483950565651e-06,
+      "loss": 0.3987,
+      "step": 2023
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917442127935426e-06,
+      "loss": 0.1386,
+      "step": 2024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9174003053052006e-06,
+      "loss": 1.7386,
+      "step": 2025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917358482674975e-06,
+      "loss": 0.0045,
+      "step": 2026
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91731666004475e-06,
+      "loss": 0.0037,
+      "step": 2027
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9172748374145247e-06,
+      "loss": 0.0043,
+      "step": 2028
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9172330147842997e-06,
+      "loss": 0.0054,
+      "step": 2029
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9171911921540743e-06,
+      "loss": 0.0191,
+      "step": 2030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9171493695238493e-06,
+      "loss": 0.0088,
+      "step": 2031
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917107546893624e-06,
+      "loss": 0.0051,
+      "step": 2032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.917065724263399e-06,
+      "loss": 0.0756,
+      "step": 2033
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9170239016331734e-06,
+      "loss": 0.0254,
+      "step": 2034
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9169820790029484e-06,
+      "loss": 0.0043,
+      "step": 2035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9169402563727234e-06,
+      "loss": 0.0035,
+      "step": 2036
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916898433742498e-06,
+      "loss": 0.0026,
+      "step": 2037
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916856611112273e-06,
+      "loss": 0.9341,
+      "step": 2038
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9168147884820476e-06,
+      "loss": 1.7782,
+      "step": 2039
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916772965851822e-06,
+      "loss": 0.0026,
+      "step": 2040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916731143221597e-06,
+      "loss": 0.003,
+      "step": 2041
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9166893205913717e-06,
+      "loss": 0.0061,
+      "step": 2042
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9166474979611467e-06,
+      "loss": 0.0213,
+      "step": 2043
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9166056753309213e-06,
+      "loss": 0.0026,
+      "step": 2044
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9165638527006963e-06,
+      "loss": 0.2648,
+      "step": 2045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916522030070471e-06,
+      "loss": 0.0681,
+      "step": 2046
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916480207440246e-06,
+      "loss": 0.0176,
+      "step": 2047
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9164383848100205e-06,
+      "loss": 0.0043,
+      "step": 2048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9163965621797955e-06,
+      "loss": 0.0064,
+      "step": 2049
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9163547395495705e-06,
+      "loss": 0.0188,
+      "step": 2050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916312916919345e-06,
+      "loss": 2.8799,
+      "step": 2051
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91627109428912e-06,
+      "loss": 0.0038,
+      "step": 2052
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9162292716588946e-06,
+      "loss": 2.1194,
+      "step": 2053
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916187449028669e-06,
+      "loss": 0.0023,
+      "step": 2054
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916145626398444e-06,
+      "loss": 0.0033,
+      "step": 2055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9161038037682187e-06,
+      "loss": 0.0022,
+      "step": 2056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9160619811379937e-06,
+      "loss": 0.0027,
+      "step": 2057
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9160201585077683e-06,
+      "loss": 0.196,
+      "step": 2058
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9159783358775433e-06,
+      "loss": 0.0058,
+      "step": 2059
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915936513247318e-06,
+      "loss": 0.0039,
+      "step": 2060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915894690617093e-06,
+      "loss": 0.0022,
+      "step": 2061
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9158528679868675e-06,
+      "loss": 0.0026,
+      "step": 2062
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9158110453566425e-06,
+      "loss": 0.0026,
+      "step": 2063
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9157692227264175e-06,
+      "loss": 0.0025,
+      "step": 2064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915727400096192e-06,
+      "loss": 3.4285,
+      "step": 2065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9156855774659666e-06,
+      "loss": 2.896,
+      "step": 2066
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9156437548357416e-06,
+      "loss": 0.0024,
+      "step": 2067
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915601932205516e-06,
+      "loss": 0.0032,
+      "step": 2068
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915560109575291e-06,
+      "loss": 0.0024,
+      "step": 2069
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9155182869450658e-06,
+      "loss": 0.0022,
+      "step": 2070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9154764643148408e-06,
+      "loss": 0.003,
+      "step": 2071
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9154346416846153e-06,
+      "loss": 1.1212,
+      "step": 2072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9153928190543903e-06,
+      "loss": 1.4255,
+      "step": 2073
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915350996424165e-06,
+      "loss": 0.0044,
+      "step": 2074
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91530917379394e-06,
+      "loss": 0.0028,
+      "step": 2075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9152673511637145e-06,
+      "loss": 0.0033,
+      "step": 2076
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9152255285334895e-06,
+      "loss": 0.0022,
+      "step": 2077
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9151837059032645e-06,
+      "loss": 0.0022,
+      "step": 2078
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915141883273039e-06,
+      "loss": 0.0274,
+      "step": 2079
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9151000606428136e-06,
+      "loss": 0.0039,
+      "step": 2080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9150582380125886e-06,
+      "loss": 1.5589,
+      "step": 2081
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915016415382363e-06,
+      "loss": 0.0041,
+      "step": 2082
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914974592752138e-06,
+      "loss": 0.0025,
+      "step": 2083
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9149327701219128e-06,
+      "loss": 1.6571,
+      "step": 2084
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9148909474916878e-06,
+      "loss": 2.9224,
+      "step": 2085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9148491248614623e-06,
+      "loss": 2.0453,
+      "step": 2086
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9148073022312373e-06,
+      "loss": 0.0089,
+      "step": 2087
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914765479601012e-06,
+      "loss": 0.007,
+      "step": 2088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914723656970787e-06,
+      "loss": 0.0028,
+      "step": 2089
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9146818343405615e-06,
+      "loss": 0.005,
+      "step": 2090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9146400117103365e-06,
+      "loss": 0.1593,
+      "step": 2091
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914598189080111e-06,
+      "loss": 0.0029,
+      "step": 2092
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914556366449886e-06,
+      "loss": 0.0042,
+      "step": 2093
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9145145438196606e-06,
+      "loss": 0.6558,
+      "step": 2094
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9144727211894356e-06,
+      "loss": 0.5673,
+      "step": 2095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91443089855921e-06,
+      "loss": 0.0047,
+      "step": 2096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914389075928985e-06,
+      "loss": 0.0039,
+      "step": 2097
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9143472532987598e-06,
+      "loss": 0.0027,
+      "step": 2098
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9143054306685348e-06,
+      "loss": 0.0037,
+      "step": 2099
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9142636080383093e-06,
+      "loss": 0.0319,
+      "step": 2100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9142217854080843e-06,
+      "loss": 0.0083,
+      "step": 2101
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914179962777859e-06,
+      "loss": 0.0029,
+      "step": 2102
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914138140147634e-06,
+      "loss": 0.003,
+      "step": 2103
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9140963175174085e-06,
+      "loss": 0.3076,
+      "step": 2104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9140544948871835e-06,
+      "loss": 0.0066,
+      "step": 2105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914012672256958e-06,
+      "loss": 0.2275,
+      "step": 2106
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9139708496267326e-06,
+      "loss": 0.0023,
+      "step": 2107
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9139290269965076e-06,
+      "loss": 0.5967,
+      "step": 2108
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9138872043662826e-06,
+      "loss": 0.0029,
+      "step": 2109
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913845381736057e-06,
+      "loss": 0.0047,
+      "step": 2110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913803559105832e-06,
+      "loss": 0.2971,
+      "step": 2111
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9137617364756068e-06,
+      "loss": 0.0187,
+      "step": 2112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9137199138453818e-06,
+      "loss": 0.9163,
+      "step": 2113
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9136780912151564e-06,
+      "loss": 0.0167,
+      "step": 2114
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9136362685849314e-06,
+      "loss": 0.0312,
+      "step": 2115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913594445954706e-06,
+      "loss": 0.1877,
+      "step": 2116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913552623324481e-06,
+      "loss": 0.2485,
+      "step": 2117
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9135108006942555e-06,
+      "loss": 0.0042,
+      "step": 2118
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9134689780640305e-06,
+      "loss": 0.0032,
+      "step": 2119
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913427155433805e-06,
+      "loss": 0.0249,
+      "step": 2120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9133853328035796e-06,
+      "loss": 2.5426,
+      "step": 2121
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9133435101733546e-06,
+      "loss": 0.0063,
+      "step": 2122
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9133016875431296e-06,
+      "loss": 0.0052,
+      "step": 2123
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9132598649129042e-06,
+      "loss": 0.0049,
+      "step": 2124
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9132180422826792e-06,
+      "loss": 0.0096,
+      "step": 2125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9131762196524538e-06,
+      "loss": 1.4701,
+      "step": 2126
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9131343970222288e-06,
+      "loss": 0.2262,
+      "step": 2127
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9130925743920034e-06,
+      "loss": 0.0285,
+      "step": 2128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9130507517617784e-06,
+      "loss": 0.0054,
+      "step": 2129
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913008929131553e-06,
+      "loss": 0.0039,
+      "step": 2130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912967106501328e-06,
+      "loss": 0.0028,
+      "step": 2131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9129252838711025e-06,
+      "loss": 0.4375,
+      "step": 2132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9128834612408775e-06,
+      "loss": 0.003,
+      "step": 2133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912841638610652e-06,
+      "loss": 0.2826,
+      "step": 2134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9127998159804267e-06,
+      "loss": 1.0273,
+      "step": 2135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9127579933502017e-06,
+      "loss": 0.0037,
+      "step": 2136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9127161707199766e-06,
+      "loss": 0.0033,
+      "step": 2137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9126743480897512e-06,
+      "loss": 1.2495,
+      "step": 2138
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9126325254595262e-06,
+      "loss": 0.0029,
+      "step": 2139
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912590702829301e-06,
+      "loss": 0.0024,
+      "step": 2140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912548880199076e-06,
+      "loss": 0.0023,
+      "step": 2141
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9125070575688504e-06,
+      "loss": 0.0028,
+      "step": 2142
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9124652349386254e-06,
+      "loss": 0.0226,
+      "step": 2143
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9124234123084e-06,
+      "loss": 0.0048,
+      "step": 2144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912381589678175e-06,
+      "loss": 0.0021,
+      "step": 2145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9123397670479495e-06,
+      "loss": 0.0024,
+      "step": 2146
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912297944417724e-06,
+      "loss": 0.0835,
+      "step": 2147
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912256121787499e-06,
+      "loss": 0.0376,
+      "step": 2148
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9122142991572737e-06,
+      "loss": 0.0032,
+      "step": 2149
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9121724765270487e-06,
+      "loss": 0.1063,
+      "step": 2150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9121306538968232e-06,
+      "loss": 0.0211,
+      "step": 2151
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9120888312665982e-06,
+      "loss": 0.2468,
+      "step": 2152
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9120470086363732e-06,
+      "loss": 0.002,
+      "step": 2153
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912005186006148e-06,
+      "loss": 0.0082,
+      "step": 2154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911963363375923e-06,
+      "loss": 0.003,
+      "step": 2155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9119215407456974e-06,
+      "loss": 0.0028,
+      "step": 2156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9118797181154724e-06,
+      "loss": 0.0036,
+      "step": 2157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911837895485247e-06,
+      "loss": 2.0989,
+      "step": 2158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911796072855022e-06,
+      "loss": 0.0026,
+      "step": 2159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9117542502247965e-06,
+      "loss": 2.5388,
+      "step": 2160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911712427594571e-06,
+      "loss": 0.0047,
+      "step": 2161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911670604964346e-06,
+      "loss": 0.004,
+      "step": 2162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9116287823341207e-06,
+      "loss": 1.5195,
+      "step": 2163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9115869597038957e-06,
+      "loss": 0.0066,
+      "step": 2164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9115451370736702e-06,
+      "loss": 0.005,
+      "step": 2165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9115033144434452e-06,
+      "loss": 0.0123,
+      "step": 2166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9114614918132202e-06,
+      "loss": 0.0034,
+      "step": 2167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911419669182995e-06,
+      "loss": 0.002,
+      "step": 2168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.91137784655277e-06,
+      "loss": 0.0032,
+      "step": 2169
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9113360239225444e-06,
+      "loss": 0.0057,
+      "step": 2170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9112942012923194e-06,
+      "loss": 1.0465,
+      "step": 2171
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911252378662094e-06,
+      "loss": 0.8998,
+      "step": 2172
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911210556031869e-06,
+      "loss": 0.0023,
+      "step": 2173
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9111687334016435e-06,
+      "loss": 0.0062,
+      "step": 2174
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911126910771418e-06,
+      "loss": 0.0021,
+      "step": 2175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.911085088141193e-06,
+      "loss": 0.1842,
+      "step": 2176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9110432655109677e-06,
+      "loss": 0.2732,
+      "step": 2177
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9110014428807427e-06,
+      "loss": 1.323,
+      "step": 2178
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9109596202505173e-06,
+      "loss": 0.0024,
+      "step": 2179
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9109177976202922e-06,
+      "loss": 0.0027,
+      "step": 2180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9108759749900672e-06,
+      "loss": 0.0022,
+      "step": 2181
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910834152359842e-06,
+      "loss": 0.5615,
+      "step": 2182
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910792329729617e-06,
+      "loss": 0.0024,
+      "step": 2183
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9107505070993914e-06,
+      "loss": 0.0026,
+      "step": 2184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9107086844691664e-06,
+      "loss": 0.0028,
+      "step": 2185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910666861838941e-06,
+      "loss": 1.7536,
+      "step": 2186
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910625039208716e-06,
+      "loss": 0.0029,
+      "step": 2187
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9105832165784905e-06,
+      "loss": 0.6245,
+      "step": 2188
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910541393948265e-06,
+      "loss": 0.0022,
+      "step": 2189
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.91049957131804e-06,
+      "loss": 0.0024,
+      "step": 2190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9104577486878147e-06,
+      "loss": 0.7006,
+      "step": 2191
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9104159260575897e-06,
+      "loss": 0.0547,
+      "step": 2192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9103741034273643e-06,
+      "loss": 0.006,
+      "step": 2193
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9103322807971393e-06,
+      "loss": 0.0025,
+      "step": 2194
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910290458166914e-06,
+      "loss": 0.0024,
+      "step": 2195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910248635536689e-06,
+      "loss": 0.0031,
+      "step": 2196
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910206812906464e-06,
+      "loss": 0.0353,
+      "step": 2197
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9101649902762384e-06,
+      "loss": 0.0026,
+      "step": 2198
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9101231676460134e-06,
+      "loss": 0.0025,
+      "step": 2199
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.910081345015788e-06,
+      "loss": 0.003,
+      "step": 2200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9100395223855626e-06,
+      "loss": 0.0022,
+      "step": 2201
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9099976997553375e-06,
+      "loss": 0.0022,
+      "step": 2202
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909955877125112e-06,
+      "loss": 0.0019,
+      "step": 2203
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909914054494887e-06,
+      "loss": 3.4697,
+      "step": 2204
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9098722318646617e-06,
+      "loss": 0.0023,
+      "step": 2205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9098304092344367e-06,
+      "loss": 0.003,
+      "step": 2206
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9097885866042113e-06,
+      "loss": 3.4146,
+      "step": 2207
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9097467639739863e-06,
+      "loss": 0.0022,
+      "step": 2208
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909704941343761e-06,
+      "loss": 2.0692,
+      "step": 2209
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909663118713536e-06,
+      "loss": 2.03,
+      "step": 2210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909621296083311e-06,
+      "loss": 0.0035,
+      "step": 2211
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9095794734530854e-06,
+      "loss": 0.0027,
+      "step": 2212
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9095376508228604e-06,
+      "loss": 2.1728,
+      "step": 2213
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909495828192635e-06,
+      "loss": 0.0436,
+      "step": 2214
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9094540055624096e-06,
+      "loss": 0.0377,
+      "step": 2215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9094121829321846e-06,
+      "loss": 0.6651,
+      "step": 2216
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909370360301959e-06,
+      "loss": 1.3527,
+      "step": 2217
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909328537671734e-06,
+      "loss": 0.0021,
+      "step": 2218
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9092867150415087e-06,
+      "loss": 0.0034,
+      "step": 2219
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9092448924112837e-06,
+      "loss": 0.0029,
+      "step": 2220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9092030697810583e-06,
+      "loss": 0.5258,
+      "step": 2221
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9091612471508333e-06,
+      "loss": 0.0026,
+      "step": 2222
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909119424520608e-06,
+      "loss": 0.3672,
+      "step": 2223
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909077601890383e-06,
+      "loss": 0.6188,
+      "step": 2224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909035779260158e-06,
+      "loss": 0.0403,
+      "step": 2225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9089939566299324e-06,
+      "loss": 0.4751,
+      "step": 2226
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9089521339997074e-06,
+      "loss": 3.5022,
+      "step": 2227
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908910311369482e-06,
+      "loss": 0.0167,
+      "step": 2228
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9088684887392566e-06,
+      "loss": 0.655,
+      "step": 2229
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9088266661090316e-06,
+      "loss": 0.0284,
+      "step": 2230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908784843478806e-06,
+      "loss": 2.5372,
+      "step": 2231
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908743020848581e-06,
+      "loss": 0.0032,
+      "step": 2232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9087011982183557e-06,
+      "loss": 0.0063,
+      "step": 2233
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9086593755881307e-06,
+      "loss": 0.005,
+      "step": 2234
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9086175529579053e-06,
+      "loss": 0.0032,
+      "step": 2235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9085757303276803e-06,
+      "loss": 0.1449,
+      "step": 2236
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908533907697455e-06,
+      "loss": 0.0031,
+      "step": 2237
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90849208506723e-06,
+      "loss": 0.0035,
+      "step": 2238
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908450262437005e-06,
+      "loss": 0.0039,
+      "step": 2239
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9084084398067794e-06,
+      "loss": 0.0168,
+      "step": 2240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9083666171765544e-06,
+      "loss": 0.0033,
+      "step": 2241
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908324794546329e-06,
+      "loss": 1.7958,
+      "step": 2242
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9082829719161036e-06,
+      "loss": 0.0099,
+      "step": 2243
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9082411492858786e-06,
+      "loss": 0.4554,
+      "step": 2244
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908199326655653e-06,
+      "loss": 1.7912,
+      "step": 2245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908157504025428e-06,
+      "loss": 0.0541,
+      "step": 2246
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9081156813952027e-06,
+      "loss": 3.57,
+      "step": 2247
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9080738587649777e-06,
+      "loss": 3.9757,
+      "step": 2248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9080320361347523e-06,
+      "loss": 2.7341,
+      "step": 2249
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9079902135045273e-06,
+      "loss": 1.7585,
+      "step": 2250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907948390874302e-06,
+      "loss": 0.9929,
+      "step": 2251
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907906568244077e-06,
+      "loss": 0.003,
+      "step": 2252
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9078647456138514e-06,
+      "loss": 0.0087,
+      "step": 2253
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9078229229836264e-06,
+      "loss": 0.0026,
+      "step": 2254
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9077811003534014e-06,
+      "loss": 2.2172,
+      "step": 2255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907739277723176e-06,
+      "loss": 0.004,
+      "step": 2256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076974550929506e-06,
+      "loss": 0.0026,
+      "step": 2257
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076556324627256e-06,
+      "loss": 0.0244,
+      "step": 2258
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076138098325e-06,
+      "loss": 0.0075,
+      "step": 2259
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907571987202275e-06,
+      "loss": 0.2254,
+      "step": 2260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9075301645720497e-06,
+      "loss": 0.0068,
+      "step": 2261
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9074883419418247e-06,
+      "loss": 0.0371,
+      "step": 2262
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9074465193115993e-06,
+      "loss": 0.0027,
+      "step": 2263
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9074046966813743e-06,
+      "loss": 0.289,
+      "step": 2264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907362874051149e-06,
+      "loss": 0.1097,
+      "step": 2265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9073210514209237e-06,
+      "loss": 0.0023,
+      "step": 2266
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9072792287906984e-06,
+      "loss": 0.5346,
+      "step": 2267
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9072374061604732e-06,
+      "loss": 0.0032,
+      "step": 2268
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9071955835302482e-06,
+      "loss": 0.0369,
+      "step": 2269
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907153760900023e-06,
+      "loss": 2.0222,
+      "step": 2270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9071119382697978e-06,
+      "loss": 0.0027,
+      "step": 2271
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9070701156395726e-06,
+      "loss": 0.3688,
+      "step": 2272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9070282930093474e-06,
+      "loss": 0.1364,
+      "step": 2273
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9069864703791222e-06,
+      "loss": 0.005,
+      "step": 2274
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906944647748897e-06,
+      "loss": 0.0033,
+      "step": 2275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9069028251186715e-06,
+      "loss": 2.9731,
+      "step": 2276
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9068610024884463e-06,
+      "loss": 0.0043,
+      "step": 2277
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906819179858221e-06,
+      "loss": 0.4733,
+      "step": 2278
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9067773572279959e-06,
+      "loss": 0.0023,
+      "step": 2279
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9067355345977707e-06,
+      "loss": 0.0181,
+      "step": 2280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9066937119675455e-06,
+      "loss": 0.0028,
+      "step": 2281
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9066518893373202e-06,
+      "loss": 0.003,
+      "step": 2282
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9066100667070952e-06,
+      "loss": 0.2036,
+      "step": 2283
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90656824407687e-06,
+      "loss": 0.2004,
+      "step": 2284
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9065264214466448e-06,
+      "loss": 0.0025,
+      "step": 2285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9064845988164196e-06,
+      "loss": 0.0023,
+      "step": 2286
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9064427761861944e-06,
+      "loss": 0.0036,
+      "step": 2287
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9064009535559692e-06,
+      "loss": 1.6651,
+      "step": 2288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9063591309257437e-06,
+      "loss": 0.0033,
+      "step": 2289
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9063173082955185e-06,
+      "loss": 0.0027,
+      "step": 2290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9062754856652933e-06,
+      "loss": 0.0785,
+      "step": 2291
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9062336630350681e-06,
+      "loss": 0.0086,
+      "step": 2292
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9061918404048429e-06,
+      "loss": 0.0033,
+      "step": 2293
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9061500177746177e-06,
+      "loss": 0.0031,
+      "step": 2294
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9061081951443925e-06,
+      "loss": 0.0027,
+      "step": 2295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9060663725141673e-06,
+      "loss": 0.0021,
+      "step": 2296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906024549883942e-06,
+      "loss": 0.4284,
+      "step": 2297
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905982727253717e-06,
+      "loss": 0.8065,
+      "step": 2298
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9059409046234918e-06,
+      "loss": 0.0025,
+      "step": 2299
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9058990819932666e-06,
+      "loss": 2.6049,
+      "step": 2300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9058572593630414e-06,
+      "loss": 0.0024,
+      "step": 2301
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9058154367328162e-06,
+      "loss": 0.0031,
+      "step": 2302
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9057736141025908e-06,
+      "loss": 2.9935,
+      "step": 2303
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9057317914723655e-06,
+      "loss": 0.004,
+      "step": 2304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9056899688421403e-06,
+      "loss": 0.0079,
+      "step": 2305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9056481462119151e-06,
+      "loss": 1.3834,
+      "step": 2306
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90560632358169e-06,
+      "loss": 0.8057,
+      "step": 2307
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9055645009514647e-06,
+      "loss": 0.0028,
+      "step": 2308
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9055226783212395e-06,
+      "loss": 0.003,
+      "step": 2309
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9054808556910143e-06,
+      "loss": 0.0184,
+      "step": 2310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905439033060789e-06,
+      "loss": 0.0024,
+      "step": 2311
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9053972104305638e-06,
+      "loss": 0.0027,
+      "step": 2312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9053553878003388e-06,
+      "loss": 2.8722,
+      "step": 2313
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9053135651701136e-06,
+      "loss": 0.5182,
+      "step": 2314
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9052717425398884e-06,
+      "loss": 2.1097,
+      "step": 2315
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9052299199096632e-06,
+      "loss": 0.0027,
+      "step": 2316
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9051880972794378e-06,
+      "loss": 0.0121,
+      "step": 2317
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9051462746492126e-06,
+      "loss": 0.0043,
+      "step": 2318
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9051044520189873e-06,
+      "loss": 0.0053,
+      "step": 2319
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9050626293887621e-06,
+      "loss": 1.903,
+      "step": 2320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.905020806758537e-06,
+      "loss": 0.0066,
+      "step": 2321
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9049789841283117e-06,
+      "loss": 0.0031,
+      "step": 2322
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9049371614980865e-06,
+      "loss": 1.5259,
+      "step": 2323
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9048953388678613e-06,
+      "loss": 0.0031,
+      "step": 2324
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904853516237636e-06,
+      "loss": 0.146,
+      "step": 2325
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9048116936074108e-06,
+      "loss": 0.0044,
+      "step": 2326
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9047698709771858e-06,
+      "loss": 0.0043,
+      "step": 2327
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9047280483469606e-06,
+      "loss": 0.003,
+      "step": 2328
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9046862257167354e-06,
+      "loss": 0.0132,
+      "step": 2329
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90464440308651e-06,
+      "loss": 0.0031,
+      "step": 2330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9046025804562848e-06,
+      "loss": 0.0054,
+      "step": 2331
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9045607578260596e-06,
+      "loss": 0.0027,
+      "step": 2332
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9045189351958343e-06,
+      "loss": 0.0049,
+      "step": 2333
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9044771125656091e-06,
+      "loss": 0.2147,
+      "step": 2334
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904435289935384e-06,
+      "loss": 0.0031,
+      "step": 2335
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9043934673051587e-06,
+      "loss": 1.2136,
+      "step": 2336
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9043516446749335e-06,
+      "loss": 0.0075,
+      "step": 2337
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9043098220447083e-06,
+      "loss": 0.0038,
+      "step": 2338
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904267999414483e-06,
+      "loss": 0.0032,
+      "step": 2339
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9042261767842579e-06,
+      "loss": 0.075,
+      "step": 2340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9041843541540326e-06,
+      "loss": 0.3747,
+      "step": 2341
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9041425315238076e-06,
+      "loss": 0.0195,
+      "step": 2342
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9041007088935824e-06,
+      "loss": 1.2753,
+      "step": 2343
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904058886263357e-06,
+      "loss": 0.0033,
+      "step": 2344
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9040170636331318e-06,
+      "loss": 0.0057,
+      "step": 2345
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9039752410029066e-06,
+      "loss": 0.0054,
+      "step": 2346
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9039334183726814e-06,
+      "loss": 0.0023,
+      "step": 2347
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9038915957424561e-06,
+      "loss": 1.3208,
+      "step": 2348
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903849773112231e-06,
+      "loss": 0.6936,
+      "step": 2349
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9038079504820057e-06,
+      "loss": 0.6141,
+      "step": 2350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9037661278517805e-06,
+      "loss": 0.003,
+      "step": 2351
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9037243052215553e-06,
+      "loss": 0.3045,
+      "step": 2352
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90368248259133e-06,
+      "loss": 0.0023,
+      "step": 2353
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9036406599611049e-06,
+      "loss": 0.0024,
+      "step": 2354
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9035988373308796e-06,
+      "loss": 0.0028,
+      "step": 2355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9035570147006544e-06,
+      "loss": 3.2914,
+      "step": 2356
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9035151920704292e-06,
+      "loss": 0.1635,
+      "step": 2357
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903473369440204e-06,
+      "loss": 0.0036,
+      "step": 2358
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9034315468099788e-06,
+      "loss": 0.0031,
+      "step": 2359
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9033897241797536e-06,
+      "loss": 0.0032,
+      "step": 2360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9033479015495284e-06,
+      "loss": 2.4095,
+      "step": 2361
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9033060789193031e-06,
+      "loss": 0.0025,
+      "step": 2362
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903264256289078e-06,
+      "loss": 0.0023,
+      "step": 2363
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9032224336588527e-06,
+      "loss": 0.0033,
+      "step": 2364
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9031806110286275e-06,
+      "loss": 0.1912,
+      "step": 2365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9031387883984023e-06,
+      "loss": 0.0041,
+      "step": 2366
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903096965768177e-06,
+      "loss": 3.0832,
+      "step": 2367
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9030551431379519e-06,
+      "loss": 0.1382,
+      "step": 2368
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9030133205077267e-06,
+      "loss": 0.2052,
+      "step": 2369
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9029714978775012e-06,
+      "loss": 0.0379,
+      "step": 2370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9029296752472762e-06,
+      "loss": 0.5783,
+      "step": 2371
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902887852617051e-06,
+      "loss": 0.0034,
+      "step": 2372
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9028460299868258e-06,
+      "loss": 0.0044,
+      "step": 2373
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9028042073566006e-06,
+      "loss": 0.0084,
+      "step": 2374
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9027623847263754e-06,
+      "loss": 0.0026,
+      "step": 2375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9027205620961502e-06,
+      "loss": 0.0172,
+      "step": 2376
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902678739465925e-06,
+      "loss": 0.0042,
+      "step": 2377
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9026369168356997e-06,
+      "loss": 0.0024,
+      "step": 2378
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9025950942054745e-06,
+      "loss": 1.9718,
+      "step": 2379
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9025532715752493e-06,
+      "loss": 0.0039,
+      "step": 2380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902511448945024e-06,
+      "loss": 0.0032,
+      "step": 2381
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9024696263147989e-06,
+      "loss": 0.1277,
+      "step": 2382
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9024278036845737e-06,
+      "loss": 0.0068,
+      "step": 2383
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9023859810543482e-06,
+      "loss": 0.4037,
+      "step": 2384
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902344158424123e-06,
+      "loss": 0.0831,
+      "step": 2385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902302335793898e-06,
+      "loss": 0.016,
+      "step": 2386
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9022605131636728e-06,
+      "loss": 0.0034,
+      "step": 2387
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9022186905334476e-06,
+      "loss": 0.0049,
+      "step": 2388
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9021768679032224e-06,
+      "loss": 0.0028,
+      "step": 2389
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9021350452729972e-06,
+      "loss": 1.9949,
+      "step": 2390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902093222642772e-06,
+      "loss": 0.0191,
+      "step": 2391
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9020514000125467e-06,
+      "loss": 0.0039,
+      "step": 2392
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9020095773823215e-06,
+      "loss": 1.7381,
+      "step": 2393
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9019677547520963e-06,
+      "loss": 1.3358,
+      "step": 2394
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901925932121871e-06,
+      "loss": 0.0048,
+      "step": 2395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9018841094916459e-06,
+      "loss": 0.0048,
+      "step": 2396
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9018422868614205e-06,
+      "loss": 0.0031,
+      "step": 2397
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9018004642311952e-06,
+      "loss": 0.0037,
+      "step": 2398
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90175864160097e-06,
+      "loss": 1.2562,
+      "step": 2399
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9017168189707448e-06,
+      "loss": 0.0025,
+      "step": 2400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9016749963405198e-06,
+      "loss": 0.961,
+      "step": 2401
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9016331737102946e-06,
+      "loss": 2.4297,
+      "step": 2402
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9015913510800694e-06,
+      "loss": 0.0329,
+      "step": 2403
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9015495284498442e-06,
+      "loss": 0.0331,
+      "step": 2404
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901507705819619e-06,
+      "loss": 0.0361,
+      "step": 2405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9014658831893937e-06,
+      "loss": 0.0038,
+      "step": 2406
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9014240605591685e-06,
+      "loss": 0.0031,
+      "step": 2407
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9013822379289433e-06,
+      "loss": 0.0035,
+      "step": 2408
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9013404152987181e-06,
+      "loss": 0.003,
+      "step": 2409
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9012985926684929e-06,
+      "loss": 0.0044,
+      "step": 2410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9012567700382675e-06,
+      "loss": 2.6577,
+      "step": 2411
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9012149474080423e-06,
+      "loss": 0.0067,
+      "step": 2412
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901173124777817e-06,
+      "loss": 0.0036,
+      "step": 2413
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9011313021475918e-06,
+      "loss": 0.2527,
+      "step": 2414
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9010894795173668e-06,
+      "loss": 0.0027,
+      "step": 2415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9010476568871416e-06,
+      "loss": 0.0032,
+      "step": 2416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9010058342569164e-06,
+      "loss": 0.0037,
+      "step": 2417
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9009640116266912e-06,
+      "loss": 0.0044,
+      "step": 2418
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900922188996466e-06,
+      "loss": 0.0063,
+      "step": 2419
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9008803663662408e-06,
+      "loss": 0.0048,
+      "step": 2420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9008385437360155e-06,
+      "loss": 1.9746,
+      "step": 2421
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9007967211057903e-06,
+      "loss": 0.913,
+      "step": 2422
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9007548984755651e-06,
+      "loss": 2.6661,
+      "step": 2423
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90071307584534e-06,
+      "loss": 0.0057,
+      "step": 2424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9006712532151145e-06,
+      "loss": 0.0026,
+      "step": 2425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9006294305848893e-06,
+      "loss": 0.0036,
+      "step": 2426
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900587607954664e-06,
+      "loss": 0.0487,
+      "step": 2427
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9005457853244388e-06,
+      "loss": 0.0038,
+      "step": 2428
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9005039626942136e-06,
+      "loss": 0.2643,
+      "step": 2429
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9004621400639886e-06,
+      "loss": 0.0174,
+      "step": 2430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9004203174337634e-06,
+      "loss": 0.0055,
+      "step": 2431
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9003784948035382e-06,
+      "loss": 1.6732,
+      "step": 2432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900336672173313e-06,
+      "loss": 0.0038,
+      "step": 2433
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9002948495430878e-06,
+      "loss": 0.0036,
+      "step": 2434
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9002530269128626e-06,
+      "loss": 0.0035,
+      "step": 2435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9002112042826373e-06,
+      "loss": 1.967,
+      "step": 2436
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9001693816524121e-06,
+      "loss": 0.0326,
+      "step": 2437
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9001275590221867e-06,
+      "loss": 2.2149,
+      "step": 2438
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9000857363919615e-06,
+      "loss": 0.0137,
+      "step": 2439
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9000439137617363e-06,
+      "loss": 0.0092,
+      "step": 2440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900002091131511e-06,
+      "loss": 0.6766,
+      "step": 2441
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8999602685012858e-06,
+      "loss": 0.0162,
+      "step": 2442
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8999184458710606e-06,
+      "loss": 0.2733,
+      "step": 2443
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8998766232408356e-06,
+      "loss": 0.0034,
+      "step": 2444
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8998348006106104e-06,
+      "loss": 2.3857,
+      "step": 2445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8997929779803852e-06,
+      "loss": 0.0034,
+      "step": 2446
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89975115535016e-06,
+      "loss": 0.0045,
+      "step": 2447
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8997093327199348e-06,
+      "loss": 0.0031,
+      "step": 2448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8996675100897096e-06,
+      "loss": 0.3332,
+      "step": 2449
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8996256874594843e-06,
+      "loss": 2.2162,
+      "step": 2450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8995838648292591e-06,
+      "loss": 0.0034,
+      "step": 2451
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8995420421990337e-06,
+      "loss": 0.058,
+      "step": 2452
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8995002195688085e-06,
+      "loss": 0.0022,
+      "step": 2453
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8994583969385833e-06,
+      "loss": 0.8209,
+      "step": 2454
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899416574308358e-06,
+      "loss": 0.6023,
+      "step": 2455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8993747516781329e-06,
+      "loss": 0.1524,
+      "step": 2456
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8993329290479076e-06,
+      "loss": 0.0034,
+      "step": 2457
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8992911064176824e-06,
+      "loss": 0.0033,
+      "step": 2458
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8992492837874574e-06,
+      "loss": 0.0057,
+      "step": 2459
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8992074611572322e-06,
+      "loss": 0.2074,
+      "step": 2460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899165638527007e-06,
+      "loss": 0.0035,
+      "step": 2461
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8991238158967818e-06,
+      "loss": 0.0036,
+      "step": 2462
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8990819932665566e-06,
+      "loss": 0.1709,
+      "step": 2463
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8990401706363314e-06,
+      "loss": 0.2444,
+      "step": 2464
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898998348006106e-06,
+      "loss": 0.0543,
+      "step": 2465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8989565253758807e-06,
+      "loss": 1.838,
+      "step": 2466
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8989147027456555e-06,
+      "loss": 0.0257,
+      "step": 2467
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8988728801154303e-06,
+      "loss": 0.0035,
+      "step": 2468
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898831057485205e-06,
+      "loss": 2.1902,
+      "step": 2469
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8987892348549799e-06,
+      "loss": 0.3003,
+      "step": 2470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8987474122247546e-06,
+      "loss": 0.003,
+      "step": 2471
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8987055895945294e-06,
+      "loss": 2.6463,
+      "step": 2472
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8986637669643042e-06,
+      "loss": 0.0084,
+      "step": 2473
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8986219443340792e-06,
+      "loss": 0.0305,
+      "step": 2474
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898580121703854e-06,
+      "loss": 0.0039,
+      "step": 2475
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8985382990736288e-06,
+      "loss": 0.0466,
+      "step": 2476
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8984964764434036e-06,
+      "loss": 0.0064,
+      "step": 2477
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8984546538131784e-06,
+      "loss": 0.0031,
+      "step": 2478
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898412831182953e-06,
+      "loss": 2.0568,
+      "step": 2479
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8983710085527277e-06,
+      "loss": 0.0067,
+      "step": 2480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8983291859225025e-06,
+      "loss": 0.0047,
+      "step": 2481
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8982873632922773e-06,
+      "loss": 0.9172,
+      "step": 2482
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898245540662052e-06,
+      "loss": 1.9402,
+      "step": 2483
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8982037180318269e-06,
+      "loss": 0.008,
+      "step": 2484
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8981618954016017e-06,
+      "loss": 1.237,
+      "step": 2485
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8981200727713764e-06,
+      "loss": 0.009,
+      "step": 2486
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8980782501411512e-06,
+      "loss": 0.2092,
+      "step": 2487
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8980364275109262e-06,
+      "loss": 0.0027,
+      "step": 2488
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897994604880701e-06,
+      "loss": 0.0137,
+      "step": 2489
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8979527822504758e-06,
+      "loss": 0.5444,
+      "step": 2490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8979109596202506e-06,
+      "loss": 0.0028,
+      "step": 2491
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8978691369900252e-06,
+      "loss": 0.0313,
+      "step": 2492
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8978273143598e-06,
+      "loss": 0.0193,
+      "step": 2493
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8977854917295747e-06,
+      "loss": 1.1242,
+      "step": 2494
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8977436690993495e-06,
+      "loss": 0.0025,
+      "step": 2495
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8977018464691243e-06,
+      "loss": 0.0022,
+      "step": 2496
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897660023838899e-06,
+      "loss": 0.0251,
+      "step": 2497
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8976182012086739e-06,
+      "loss": 0.0059,
+      "step": 2498
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8975763785784487e-06,
+      "loss": 0.0223,
+      "step": 2499
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8975345559482235e-06,
+      "loss": 0.2235,
+      "step": 2500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8974927333179982e-06,
+      "loss": 0.0046,
+      "step": 2501
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897450910687773e-06,
+      "loss": 0.0027,
+      "step": 2502
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897409088057548e-06,
+      "loss": 2.747,
+      "step": 2503
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8973672654273228e-06,
+      "loss": 0.0057,
+      "step": 2504
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8973254427970976e-06,
+      "loss": 0.0056,
+      "step": 2505
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8972836201668722e-06,
+      "loss": 0.1036,
+      "step": 2506
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897241797536647e-06,
+      "loss": 0.0042,
+      "step": 2507
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8971999749064217e-06,
+      "loss": 0.5172,
+      "step": 2508
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8971581522761965e-06,
+      "loss": 0.0128,
+      "step": 2509
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8971163296459713e-06,
+      "loss": 1.0198,
+      "step": 2510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897074507015746e-06,
+      "loss": 0.6104,
+      "step": 2511
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8970326843855209e-06,
+      "loss": 0.1817,
+      "step": 2512
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8969908617552957e-06,
+      "loss": 0.0261,
+      "step": 2513
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8969490391250705e-06,
+      "loss": 0.0024,
+      "step": 2514
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8969072164948452e-06,
+      "loss": 1.0736,
+      "step": 2515
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89686539386462e-06,
+      "loss": 0.5501,
+      "step": 2516
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8968235712343948e-06,
+      "loss": 1.875,
+      "step": 2517
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8967817486041698e-06,
+      "loss": 0.0069,
+      "step": 2518
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8967399259739446e-06,
+      "loss": 0.0027,
+      "step": 2519
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8966981033437192e-06,
+      "loss": 0.0032,
+      "step": 2520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896656280713494e-06,
+      "loss": 0.0031,
+      "step": 2521
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8966144580832688e-06,
+      "loss": 0.0111,
+      "step": 2522
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8965726354530435e-06,
+      "loss": 0.0037,
+      "step": 2523
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8965308128228183e-06,
+      "loss": 0.0025,
+      "step": 2524
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8964889901925931e-06,
+      "loss": 3.2009,
+      "step": 2525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896447167562368e-06,
+      "loss": 0.0024,
+      "step": 2526
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8964053449321427e-06,
+      "loss": 0.002,
+      "step": 2527
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8963635223019175e-06,
+      "loss": 1.9157,
+      "step": 2528
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8963216996716923e-06,
+      "loss": 0.3747,
+      "step": 2529
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896279877041467e-06,
+      "loss": 1.7403,
+      "step": 2530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8962380544112418e-06,
+      "loss": 0.0027,
+      "step": 2531
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8961962317810168e-06,
+      "loss": 0.0033,
+      "step": 2532
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8961544091507914e-06,
+      "loss": 0.0034,
+      "step": 2533
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8961125865205662e-06,
+      "loss": 0.0563,
+      "step": 2534
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896070763890341e-06,
+      "loss": 0.0031,
+      "step": 2535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8960289412601158e-06,
+      "loss": 0.0042,
+      "step": 2536
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8959871186298905e-06,
+      "loss": 0.0094,
+      "step": 2537
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8959452959996653e-06,
+      "loss": 0.0039,
+      "step": 2538
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8959034733694401e-06,
+      "loss": 0.0086,
+      "step": 2539
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895861650739215e-06,
+      "loss": 2.6834,
+      "step": 2540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8958198281089897e-06,
+      "loss": 0.0067,
+      "step": 2541
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8957780054787645e-06,
+      "loss": 0.0029,
+      "step": 2542
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8957361828485393e-06,
+      "loss": 0.0086,
+      "step": 2543
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895694360218314e-06,
+      "loss": 2.2662,
+      "step": 2544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8956525375880888e-06,
+      "loss": 0.0036,
+      "step": 2545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8956107149578634e-06,
+      "loss": 0.0039,
+      "step": 2546
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8955688923276384e-06,
+      "loss": 0.1475,
+      "step": 2547
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8955270696974132e-06,
+      "loss": 0.0045,
+      "step": 2548
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895485247067188e-06,
+      "loss": 0.0154,
+      "step": 2549
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8954434244369628e-06,
+      "loss": 3.5171,
+      "step": 2550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8954016018067376e-06,
+      "loss": 0.0067,
+      "step": 2551
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8953597791765123e-06,
+      "loss": 0.0032,
+      "step": 2552
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8953179565462871e-06,
+      "loss": 0.016,
+      "step": 2553
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895276133916062e-06,
+      "loss": 0.0022,
+      "step": 2554
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8952343112858367e-06,
+      "loss": 0.0268,
+      "step": 2555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8951924886556115e-06,
+      "loss": 0.5583,
+      "step": 2556
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8951506660253863e-06,
+      "loss": 0.747,
+      "step": 2557
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895108843395161e-06,
+      "loss": 0.0083,
+      "step": 2558
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8950670207649358e-06,
+      "loss": 0.0049,
+      "step": 2559
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8950251981347104e-06,
+      "loss": 0.0088,
+      "step": 2560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8949833755044852e-06,
+      "loss": 0.0046,
+      "step": 2561
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8949415528742602e-06,
+      "loss": 0.0312,
+      "step": 2562
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894899730244035e-06,
+      "loss": 0.1686,
+      "step": 2563
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8948579076138098e-06,
+      "loss": 0.0021,
+      "step": 2564
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8948160849835846e-06,
+      "loss": 0.007,
+      "step": 2565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8947742623533593e-06,
+      "loss": 0.6677,
+      "step": 2566
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8947324397231341e-06,
+      "loss": 2.5602,
+      "step": 2567
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894690617092909e-06,
+      "loss": 0.0052,
+      "step": 2568
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8946487944626837e-06,
+      "loss": 0.3206,
+      "step": 2569
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8946069718324585e-06,
+      "loss": 0.6215,
+      "step": 2570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8945651492022333e-06,
+      "loss": 0.0037,
+      "step": 2571
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894523326572008e-06,
+      "loss": 0.2617,
+      "step": 2572
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8944815039417826e-06,
+      "loss": 0.0669,
+      "step": 2573
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8944396813115574e-06,
+      "loss": 0.0031,
+      "step": 2574
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8943978586813322e-06,
+      "loss": 0.0025,
+      "step": 2575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8943560360511072e-06,
+      "loss": 0.7719,
+      "step": 2576
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894314213420882e-06,
+      "loss": 0.0025,
+      "step": 2577
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8942723907906568e-06,
+      "loss": 0.002,
+      "step": 2578
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8942305681604316e-06,
+      "loss": 0.0023,
+      "step": 2579
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8941887455302064e-06,
+      "loss": 0.0036,
+      "step": 2580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8941469228999811e-06,
+      "loss": 0.0262,
+      "step": 2581
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894105100269756e-06,
+      "loss": 0.002,
+      "step": 2582
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8940632776395307e-06,
+      "loss": 0.0333,
+      "step": 2583
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8940214550093055e-06,
+      "loss": 0.003,
+      "step": 2584
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8939796323790803e-06,
+      "loss": 0.0032,
+      "step": 2585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893937809748855e-06,
+      "loss": 1.5565,
+      "step": 2586
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8938959871186297e-06,
+      "loss": 0.5984,
+      "step": 2587
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8938541644884044e-06,
+      "loss": 1.6806,
+      "step": 2588
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8938123418581792e-06,
+      "loss": 0.0024,
+      "step": 2589
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893770519227954e-06,
+      "loss": 0.0041,
+      "step": 2590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893728696597729e-06,
+      "loss": 0.8844,
+      "step": 2591
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8936868739675038e-06,
+      "loss": 0.0033,
+      "step": 2592
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8936450513372786e-06,
+      "loss": 0.002,
+      "step": 2593
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8936032287070534e-06,
+      "loss": 0.0026,
+      "step": 2594
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8935614060768282e-06,
+      "loss": 0.1123,
+      "step": 2595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893519583446603e-06,
+      "loss": 0.0044,
+      "step": 2596
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8934777608163777e-06,
+      "loss": 0.0047,
+      "step": 2597
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8934359381861525e-06,
+      "loss": 0.2209,
+      "step": 2598
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8933941155559273e-06,
+      "loss": 0.0032,
+      "step": 2599
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8933522929257019e-06,
+      "loss": 0.0032,
+      "step": 2600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8933104702954767e-06,
+      "loss": 0.9361,
+      "step": 2601
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8932686476652514e-06,
+      "loss": 0.1513,
+      "step": 2602
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8932268250350262e-06,
+      "loss": 0.0053,
+      "step": 2603
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893185002404801e-06,
+      "loss": 1.2332,
+      "step": 2604
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8931431797745758e-06,
+      "loss": 2.6091,
+      "step": 2605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8931013571443508e-06,
+      "loss": 3.034,
+      "step": 2606
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8930595345141256e-06,
+      "loss": 0.0059,
+      "step": 2607
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8930177118839004e-06,
+      "loss": 0.8735,
+      "step": 2608
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8929758892536752e-06,
+      "loss": 0.002,
+      "step": 2609
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89293406662345e-06,
+      "loss": 0.0035,
+      "step": 2610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8928922439932247e-06,
+      "loss": 0.0056,
+      "step": 2611
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8928504213629995e-06,
+      "loss": 2.7549,
+      "step": 2612
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8928085987327743e-06,
+      "loss": 1.6412,
+      "step": 2613
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8927667761025489e-06,
+      "loss": 0.0612,
+      "step": 2614
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8927249534723237e-06,
+      "loss": 1.2575,
+      "step": 2615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8926831308420985e-06,
+      "loss": 0.0054,
+      "step": 2616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8926413082118732e-06,
+      "loss": 0.0079,
+      "step": 2617
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892599485581648e-06,
+      "loss": 0.0158,
+      "step": 2618
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8925576629514228e-06,
+      "loss": 0.0077,
+      "step": 2619
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8925158403211978e-06,
+      "loss": 0.6845,
+      "step": 2620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8924740176909726e-06,
+      "loss": 0.029,
+      "step": 2621
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8924321950607474e-06,
+      "loss": 0.0033,
+      "step": 2622
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8923903724305222e-06,
+      "loss": 0.0078,
+      "step": 2623
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892348549800297e-06,
+      "loss": 0.7117,
+      "step": 2624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8923067271700717e-06,
+      "loss": 0.0102,
+      "step": 2625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8922649045398465e-06,
+      "loss": 1.3211,
+      "step": 2626
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8922230819096213e-06,
+      "loss": 0.3145,
+      "step": 2627
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8921812592793959e-06,
+      "loss": 0.0027,
+      "step": 2628
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8921394366491707e-06,
+      "loss": 1.9053,
+      "step": 2629
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8920976140189455e-06,
+      "loss": 0.0037,
+      "step": 2630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8920557913887202e-06,
+      "loss": 0.0075,
+      "step": 2631
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892013968758495e-06,
+      "loss": 0.327,
+      "step": 2632
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919721461282698e-06,
+      "loss": 0.017,
+      "step": 2633
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919303234980446e-06,
+      "loss": 0.0082,
+      "step": 2634
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8918885008678196e-06,
+      "loss": 0.1161,
+      "step": 2635
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8918466782375944e-06,
+      "loss": 0.005,
+      "step": 2636
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8918048556073692e-06,
+      "loss": 0.0917,
+      "step": 2637
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891763032977144e-06,
+      "loss": 0.0231,
+      "step": 2638
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8917212103469188e-06,
+      "loss": 0.9189,
+      "step": 2639
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8916793877166935e-06,
+      "loss": 0.0023,
+      "step": 2640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8916375650864681e-06,
+      "loss": 0.4886,
+      "step": 2641
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891595742456243e-06,
+      "loss": 1.6576,
+      "step": 2642
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8915539198260177e-06,
+      "loss": 0.4745,
+      "step": 2643
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8915120971957925e-06,
+      "loss": 0.0125,
+      "step": 2644
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8914702745655673e-06,
+      "loss": 0.0028,
+      "step": 2645
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891428451935342e-06,
+      "loss": 0.0031,
+      "step": 2646
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8913866293051168e-06,
+      "loss": 0.0189,
+      "step": 2647
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8913448066748916e-06,
+      "loss": 1.3285,
+      "step": 2648
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8913029840446666e-06,
+      "loss": 0.0068,
+      "step": 2649
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8912611614144414e-06,
+      "loss": 0.0034,
+      "step": 2650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8912193387842162e-06,
+      "loss": 0.273,
+      "step": 2651
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891177516153991e-06,
+      "loss": 0.3185,
+      "step": 2652
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8911356935237658e-06,
+      "loss": 0.003,
+      "step": 2653
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8910938708935405e-06,
+      "loss": 0.0031,
+      "step": 2654
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8910520482633151e-06,
+      "loss": 0.004,
+      "step": 2655
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89101022563309e-06,
+      "loss": 0.0035,
+      "step": 2656
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8909684030028647e-06,
+      "loss": 0.0025,
+      "step": 2657
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8909265803726395e-06,
+      "loss": 0.0035,
+      "step": 2658
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8908847577424143e-06,
+      "loss": 0.0074,
+      "step": 2659
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890842935112189e-06,
+      "loss": 0.0031,
+      "step": 2660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8908011124819638e-06,
+      "loss": 0.0067,
+      "step": 2661
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8907592898517386e-06,
+      "loss": 0.6494,
+      "step": 2662
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8907174672215134e-06,
+      "loss": 0.003,
+      "step": 2663
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8906756445912884e-06,
+      "loss": 0.5034,
+      "step": 2664
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8906338219610632e-06,
+      "loss": 0.0404,
+      "step": 2665
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890591999330838e-06,
+      "loss": 0.1117,
+      "step": 2666
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8905501767006128e-06,
+      "loss": 3.0561,
+      "step": 2667
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8905083540703873e-06,
+      "loss": 0.1631,
+      "step": 2668
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8904665314401621e-06,
+      "loss": 0.002,
+      "step": 2669
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890424708809937e-06,
+      "loss": 0.3675,
+      "step": 2670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8903828861797117e-06,
+      "loss": 0.0028,
+      "step": 2671
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8903410635494865e-06,
+      "loss": 0.3464,
+      "step": 2672
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8902992409192613e-06,
+      "loss": 0.003,
+      "step": 2673
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890257418289036e-06,
+      "loss": 0.0034,
+      "step": 2674
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8902155956588108e-06,
+      "loss": 0.7484,
+      "step": 2675
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8901737730285856e-06,
+      "loss": 0.3392,
+      "step": 2676
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8901319503983604e-06,
+      "loss": 1.851,
+      "step": 2677
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8900901277681352e-06,
+      "loss": 0.0022,
+      "step": 2678
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8900483051379102e-06,
+      "loss": 0.191,
+      "step": 2679
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890006482507685e-06,
+      "loss": 0.0039,
+      "step": 2680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8899646598774598e-06,
+      "loss": 0.0036,
+      "step": 2681
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8899228372472344e-06,
+      "loss": 0.0073,
+      "step": 2682
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8898810146170091e-06,
+      "loss": 0.0119,
+      "step": 2683
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889839191986784e-06,
+      "loss": 0.2651,
+      "step": 2684
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8897973693565587e-06,
+      "loss": 0.1036,
+      "step": 2685
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8897555467263335e-06,
+      "loss": 0.0026,
+      "step": 2686
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8897137240961083e-06,
+      "loss": 2.2699,
+      "step": 2687
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889671901465883e-06,
+      "loss": 0.0029,
+      "step": 2688
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8896300788356579e-06,
+      "loss": 0.0043,
+      "step": 2689
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8895882562054326e-06,
+      "loss": 2.6625,
+      "step": 2690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8895464335752074e-06,
+      "loss": 0.0043,
+      "step": 2691
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8895046109449822e-06,
+      "loss": 0.2546,
+      "step": 2692
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8894627883147572e-06,
+      "loss": 0.0052,
+      "step": 2693
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889420965684532e-06,
+      "loss": 0.7401,
+      "step": 2694
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8893791430543066e-06,
+      "loss": 0.004,
+      "step": 2695
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8893373204240814e-06,
+      "loss": 1.3667,
+      "step": 2696
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8892954977938561e-06,
+      "loss": 0.0087,
+      "step": 2697
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889253675163631e-06,
+      "loss": 1.4385,
+      "step": 2698
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8892118525334057e-06,
+      "loss": 0.0035,
+      "step": 2699
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8891700299031805e-06,
+      "loss": 0.0037,
+      "step": 2700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8891282072729553e-06,
+      "loss": 0.2007,
+      "step": 2701
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88908638464273e-06,
+      "loss": 0.0041,
+      "step": 2702
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8890445620125049e-06,
+      "loss": 0.0029,
+      "step": 2703
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8890027393822797e-06,
+      "loss": 0.0026,
+      "step": 2704
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8889609167520544e-06,
+      "loss": 0.0053,
+      "step": 2705
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8889190941218292e-06,
+      "loss": 1.5731,
+      "step": 2706
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888877271491604e-06,
+      "loss": 0.0036,
+      "step": 2707
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888835448861379e-06,
+      "loss": 0.0042,
+      "step": 2708
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8887936262311536e-06,
+      "loss": 0.0044,
+      "step": 2709
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8887518036009284e-06,
+      "loss": 3.0888,
+      "step": 2710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8887099809707032e-06,
+      "loss": 0.0024,
+      "step": 2711
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888668158340478e-06,
+      "loss": 0.0028,
+      "step": 2712
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8886263357102527e-06,
+      "loss": 0.0027,
+      "step": 2713
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8885845130800275e-06,
+      "loss": 0.0102,
+      "step": 2714
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8885426904498023e-06,
+      "loss": 1.6873,
+      "step": 2715
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888500867819577e-06,
+      "loss": 0.0031,
+      "step": 2716
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8884590451893519e-06,
+      "loss": 0.0032,
+      "step": 2717
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8884172225591267e-06,
+      "loss": 0.0034,
+      "step": 2718
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8883753999289014e-06,
+      "loss": 0.0041,
+      "step": 2719
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8883335772986762e-06,
+      "loss": 0.2109,
+      "step": 2720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888291754668451e-06,
+      "loss": 0.0032,
+      "step": 2721
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8882499320382256e-06,
+      "loss": 0.0022,
+      "step": 2722
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8882081094080006e-06,
+      "loss": 0.0032,
+      "step": 2723
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8881662867777754e-06,
+      "loss": 0.0027,
+      "step": 2724
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8881244641475502e-06,
+      "loss": 0.0036,
+      "step": 2725
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888082641517325e-06,
+      "loss": 0.01,
+      "step": 2726
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8880408188870997e-06,
+      "loss": 3.0689,
+      "step": 2727
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8879989962568745e-06,
+      "loss": 0.0044,
+      "step": 2728
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8879571736266493e-06,
+      "loss": 2.2872,
+      "step": 2729
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887915350996424e-06,
+      "loss": 1.2697,
+      "step": 2730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8878735283661989e-06,
+      "loss": 0.0114,
+      "step": 2731
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8878317057359737e-06,
+      "loss": 0.0028,
+      "step": 2732
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8877898831057485e-06,
+      "loss": 0.0035,
+      "step": 2733
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8877480604755232e-06,
+      "loss": 0.0233,
+      "step": 2734
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887706237845298e-06,
+      "loss": 0.0775,
+      "step": 2735
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8876644152150726e-06,
+      "loss": 0.0031,
+      "step": 2736
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8876225925848476e-06,
+      "loss": 0.0027,
+      "step": 2737
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8875807699546224e-06,
+      "loss": 0.1724,
+      "step": 2738
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8875389473243972e-06,
+      "loss": 0.0033,
+      "step": 2739
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887497124694172e-06,
+      "loss": 0.0027,
+      "step": 2740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8874553020639467e-06,
+      "loss": 0.0043,
+      "step": 2741
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8874134794337215e-06,
+      "loss": 0.0071,
+      "step": 2742
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8873716568034963e-06,
+      "loss": 0.2091,
+      "step": 2743
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887329834173271e-06,
+      "loss": 0.0335,
+      "step": 2744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8872880115430459e-06,
+      "loss": 0.0051,
+      "step": 2745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8872461889128207e-06,
+      "loss": 0.0032,
+      "step": 2746
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8872043662825955e-06,
+      "loss": 0.0026,
+      "step": 2747
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8871625436523702e-06,
+      "loss": 0.0408,
+      "step": 2748
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8871207210221448e-06,
+      "loss": 0.0035,
+      "step": 2749
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8870788983919196e-06,
+      "loss": 0.0024,
+      "step": 2750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8870370757616944e-06,
+      "loss": 1.9198,
+      "step": 2751
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8869952531314694e-06,
+      "loss": 0.0024,
+      "step": 2752
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8869534305012442e-06,
+      "loss": 1.274,
+      "step": 2753
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886911607871019e-06,
+      "loss": 0.0064,
+      "step": 2754
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8868697852407938e-06,
+      "loss": 0.0083,
+      "step": 2755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8868279626105685e-06,
+      "loss": 0.0021,
+      "step": 2756
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8867861399803433e-06,
+      "loss": 0.0106,
+      "step": 2757
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8867443173501181e-06,
+      "loss": 0.0061,
+      "step": 2758
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886702494719893e-06,
+      "loss": 0.0019,
+      "step": 2759
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8866606720896677e-06,
+      "loss": 0.0552,
+      "step": 2760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8866188494594425e-06,
+      "loss": 0.0035,
+      "step": 2761
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8865770268292173e-06,
+      "loss": 0.0026,
+      "step": 2762
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8865352041989918e-06,
+      "loss": 0.0027,
+      "step": 2763
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8864933815687666e-06,
+      "loss": 0.0052,
+      "step": 2764
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8864515589385414e-06,
+      "loss": 0.0021,
+      "step": 2765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8864097363083162e-06,
+      "loss": 1.7018,
+      "step": 2766
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8863679136780912e-06,
+      "loss": 0.0083,
+      "step": 2767
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886326091047866e-06,
+      "loss": 0.0035,
+      "step": 2768
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8862842684176408e-06,
+      "loss": 3.1566,
+      "step": 2769
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8862424457874155e-06,
+      "loss": 0.0079,
+      "step": 2770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8862006231571903e-06,
+      "loss": 0.0031,
+      "step": 2771
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8861588005269651e-06,
+      "loss": 0.0044,
+      "step": 2772
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88611697789674e-06,
+      "loss": 2.7729,
+      "step": 2773
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8860751552665147e-06,
+      "loss": 0.0033,
+      "step": 2774
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8860333326362895e-06,
+      "loss": 1.7232,
+      "step": 2775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885991510006064e-06,
+      "loss": 0.4233,
+      "step": 2776
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8859496873758388e-06,
+      "loss": 0.0023,
+      "step": 2777
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8859078647456136e-06,
+      "loss": 0.0046,
+      "step": 2778
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8858660421153884e-06,
+      "loss": 0.0119,
+      "step": 2779
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8858242194851632e-06,
+      "loss": 0.0161,
+      "step": 2780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8857823968549382e-06,
+      "loss": 0.0058,
+      "step": 2781
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885740574224713e-06,
+      "loss": 0.0027,
+      "step": 2782
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8856987515944878e-06,
+      "loss": 0.004,
+      "step": 2783
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8856569289642626e-06,
+      "loss": 0.0026,
+      "step": 2784
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8856151063340373e-06,
+      "loss": 0.01,
+      "step": 2785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8855732837038121e-06,
+      "loss": 2.2567,
+      "step": 2786
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885531461073587e-06,
+      "loss": 0.0545,
+      "step": 2787
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8854896384433617e-06,
+      "loss": 0.0074,
+      "step": 2788
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8854478158131365e-06,
+      "loss": 0.0075,
+      "step": 2789
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885405993182911e-06,
+      "loss": 0.0023,
+      "step": 2790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8853641705526858e-06,
+      "loss": 0.8318,
+      "step": 2791
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8853223479224606e-06,
+      "loss": 0.0021,
+      "step": 2792
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8852805252922354e-06,
+      "loss": 0.0035,
+      "step": 2793
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8852387026620102e-06,
+      "loss": 0.003,
+      "step": 2794
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885196880031785e-06,
+      "loss": 0.0039,
+      "step": 2795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88515505740156e-06,
+      "loss": 0.0087,
+      "step": 2796
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851132347713348e-06,
+      "loss": 0.0021,
+      "step": 2797
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850714121411096e-06,
+      "loss": 0.0026,
+      "step": 2798
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850295895108844e-06,
+      "loss": 0.0019,
+      "step": 2799
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849877668806591e-06,
+      "loss": 0.0023,
+      "step": 2800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884945944250434e-06,
+      "loss": 0.0065,
+      "step": 2801
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849041216202087e-06,
+      "loss": 0.0121,
+      "step": 2802
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848622989899833e-06,
+      "loss": 0.0695,
+      "step": 2803
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884820476359758e-06,
+      "loss": 0.3518,
+      "step": 2804
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847786537295329e-06,
+      "loss": 0.0022,
+      "step": 2805
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847368310993076e-06,
+      "loss": 0.0048,
+      "step": 2806
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8846950084690824e-06,
+      "loss": 0.4765,
+      "step": 2807
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8846531858388572e-06,
+      "loss": 0.002,
+      "step": 2808
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884611363208632e-06,
+      "loss": 0.0027,
+      "step": 2809
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845695405784068e-06,
+      "loss": 0.0083,
+      "step": 2810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845277179481818e-06,
+      "loss": 0.0019,
+      "step": 2811
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844858953179566e-06,
+      "loss": 0.0022,
+      "step": 2812
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844440726877314e-06,
+      "loss": 0.0057,
+      "step": 2813
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844022500575061e-06,
+      "loss": 0.0073,
+      "step": 2814
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884360427427281e-06,
+      "loss": 0.3475,
+      "step": 2815
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843186047970557e-06,
+      "loss": 0.0033,
+      "step": 2816
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8842767821668303e-06,
+      "loss": 0.0022,
+      "step": 2817
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884234959536605e-06,
+      "loss": 0.0028,
+      "step": 2818
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841931369063799e-06,
+      "loss": 0.4956,
+      "step": 2819
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841513142761547e-06,
+      "loss": 0.0019,
+      "step": 2820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841094916459294e-06,
+      "loss": 0.0258,
+      "step": 2821
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840676690157042e-06,
+      "loss": 0.0051,
+      "step": 2822
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884025846385479e-06,
+      "loss": 0.0028,
+      "step": 2823
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839840237552538e-06,
+      "loss": 0.0141,
+      "step": 2824
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839422011250288e-06,
+      "loss": 0.0042,
+      "step": 2825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839003784948036e-06,
+      "loss": 0.0027,
+      "step": 2826
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838585558645784e-06,
+      "loss": 0.0035,
+      "step": 2827
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838167332343532e-06,
+      "loss": 0.0025,
+      "step": 2828
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883774910604128e-06,
+      "loss": 0.0918,
+      "step": 2829
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837330879739027e-06,
+      "loss": 0.1229,
+      "step": 2830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836912653436773e-06,
+      "loss": 0.0028,
+      "step": 2831
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883649442713452e-06,
+      "loss": 0.0025,
+      "step": 2832
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836076200832269e-06,
+      "loss": 0.4425,
+      "step": 2833
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8835657974530017e-06,
+      "loss": 0.003,
+      "step": 2834
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8835239748227764e-06,
+      "loss": 0.0025,
+      "step": 2835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834821521925512e-06,
+      "loss": 0.0017,
+      "step": 2836
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883440329562326e-06,
+      "loss": 0.0022,
+      "step": 2837
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833985069321008e-06,
+      "loss": 0.0023,
+      "step": 2838
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833566843018756e-06,
+      "loss": 0.0022,
+      "step": 2839
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833148616716506e-06,
+      "loss": 0.0042,
+      "step": 2840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832730390414254e-06,
+      "loss": 0.6705,
+      "step": 2841
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832312164112002e-06,
+      "loss": 0.002,
+      "step": 2842
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883189393780975e-06,
+      "loss": 0.0025,
+      "step": 2843
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831475711507495e-06,
+      "loss": 0.002,
+      "step": 2844
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831057485205243e-06,
+      "loss": 0.0021,
+      "step": 2845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883063925890299e-06,
+      "loss": 0.034,
+      "step": 2846
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830221032600739e-06,
+      "loss": 1.1782,
+      "step": 2847
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829802806298487e-06,
+      "loss": 0.0034,
+      "step": 2848
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829384579996235e-06,
+      "loss": 2.2992,
+      "step": 2849
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828966353693982e-06,
+      "loss": 2.0116,
+      "step": 2850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882854812739173e-06,
+      "loss": 0.0027,
+      "step": 2851
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828129901089478e-06,
+      "loss": 0.0023,
+      "step": 2852
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827711674787226e-06,
+      "loss": 0.0021,
+      "step": 2853
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827293448484976e-06,
+      "loss": 0.0613,
+      "step": 2854
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826875222182724e-06,
+      "loss": 0.0021,
+      "step": 2855
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826456995880472e-06,
+      "loss": 0.0021,
+      "step": 2856
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882603876957822e-06,
+      "loss": 0.0017,
+      "step": 2857
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825620543275965e-06,
+      "loss": 0.6467,
+      "step": 2858
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825202316973713e-06,
+      "loss": 0.0021,
+      "step": 2859
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824784090671461e-06,
+      "loss": 0.0019,
+      "step": 2860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824365864369209e-06,
+      "loss": 0.0014,
+      "step": 2861
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823947638066957e-06,
+      "loss": 0.1215,
+      "step": 2862
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823529411764705e-06,
+      "loss": 0.0029,
+      "step": 2863
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823111185462453e-06,
+      "loss": 1.6028,
+      "step": 2864
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88226929591602e-06,
+      "loss": 0.0038,
+      "step": 2865
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822274732857948e-06,
+      "loss": 0.0046,
+      "step": 2866
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821856506555696e-06,
+      "loss": 0.0028,
+      "step": 2867
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821438280253444e-06,
+      "loss": 0.0023,
+      "step": 2868
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821020053951194e-06,
+      "loss": 2.1913,
+      "step": 2869
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820601827648942e-06,
+      "loss": 0.0023,
+      "step": 2870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820183601346688e-06,
+      "loss": 0.0024,
+      "step": 2871
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819765375044435e-06,
+      "loss": 0.0036,
+      "step": 2872
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819347148742183e-06,
+      "loss": 0.0018,
+      "step": 2873
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818928922439931e-06,
+      "loss": 0.0022,
+      "step": 2874
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881851069613768e-06,
+      "loss": 0.3334,
+      "step": 2875
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818092469835427e-06,
+      "loss": 0.0023,
+      "step": 2876
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817674243533175e-06,
+      "loss": 0.0021,
+      "step": 2877
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817256017230923e-06,
+      "loss": 0.0166,
+      "step": 2878
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881683779092867e-06,
+      "loss": 0.002,
+      "step": 2879
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816419564626418e-06,
+      "loss": 0.0028,
+      "step": 2880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816001338324166e-06,
+      "loss": 3.2782,
+      "step": 2881
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815583112021914e-06,
+      "loss": 1.6369,
+      "step": 2882
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815164885719662e-06,
+      "loss": 0.0017,
+      "step": 2883
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814746659417412e-06,
+      "loss": 0.0039,
+      "step": 2884
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814328433115158e-06,
+      "loss": 0.002,
+      "step": 2885
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813910206812906e-06,
+      "loss": 0.0021,
+      "step": 2886
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813491980510653e-06,
+      "loss": 0.0027,
+      "step": 2887
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813073754208401e-06,
+      "loss": 0.0029,
+      "step": 2888
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881265552790615e-06,
+      "loss": 3.0779,
+      "step": 2889
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812237301603897e-06,
+      "loss": 3.1915,
+      "step": 2890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811819075301645e-06,
+      "loss": 0.0016,
+      "step": 2891
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811400848999393e-06,
+      "loss": 0.0016,
+      "step": 2892
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881098262269714e-06,
+      "loss": 0.003,
+      "step": 2893
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810564396394888e-06,
+      "loss": 0.191,
+      "step": 2894
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810146170092636e-06,
+      "loss": 0.247,
+      "step": 2895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809727943790384e-06,
+      "loss": 0.6277,
+      "step": 2896
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809309717488132e-06,
+      "loss": 0.0024,
+      "step": 2897
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880889149118588e-06,
+      "loss": 0.4442,
+      "step": 2898
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808473264883628e-06,
+      "loss": 0.0023,
+      "step": 2899
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808055038581376e-06,
+      "loss": 0.002,
+      "step": 2900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807636812279123e-06,
+      "loss": 0.5911,
+      "step": 2901
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807218585976871e-06,
+      "loss": 0.0044,
+      "step": 2902
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880680035967462e-06,
+      "loss": 0.0057,
+      "step": 2903
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806382133372367e-06,
+      "loss": 0.7733,
+      "step": 2904
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805963907070115e-06,
+      "loss": 0.0024,
+      "step": 2905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805545680767863e-06,
+      "loss": 0.0025,
+      "step": 2906
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880512745446561e-06,
+      "loss": 0.0032,
+      "step": 2907
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804709228163359e-06,
+      "loss": 0.0016,
+      "step": 2908
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804291001861106e-06,
+      "loss": 0.0025,
+      "step": 2909
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803872775558854e-06,
+      "loss": 0.0453,
+      "step": 2910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88034545492566e-06,
+      "loss": 0.0018,
+      "step": 2911
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803036322954348e-06,
+      "loss": 0.0104,
+      "step": 2912
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802618096652098e-06,
+      "loss": 0.0038,
+      "step": 2913
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802199870349846e-06,
+      "loss": 1.2514,
+      "step": 2914
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801781644047594e-06,
+      "loss": 0.0015,
+      "step": 2915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801363417745341e-06,
+      "loss": 0.0025,
+      "step": 2916
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880094519144309e-06,
+      "loss": 0.1725,
+      "step": 2917
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800526965140837e-06,
+      "loss": 0.0024,
+      "step": 2918
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800108738838585e-06,
+      "loss": 0.0029,
+      "step": 2919
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799690512536333e-06,
+      "loss": 0.0021,
+      "step": 2920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879927228623408e-06,
+      "loss": 0.4099,
+      "step": 2921
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798854059931829e-06,
+      "loss": 0.0027,
+      "step": 2922
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798435833629576e-06,
+      "loss": 0.0024,
+      "step": 2923
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798017607327324e-06,
+      "loss": 0.0019,
+      "step": 2924
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879759938102507e-06,
+      "loss": 0.0028,
+      "step": 2925
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797181154722818e-06,
+      "loss": 0.1217,
+      "step": 2926
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796762928420566e-06,
+      "loss": 0.0177,
+      "step": 2927
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796344702118316e-06,
+      "loss": 0.4641,
+      "step": 2928
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795926475816064e-06,
+      "loss": 0.1915,
+      "step": 2929
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795508249513811e-06,
+      "loss": 0.0021,
+      "step": 2930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879509002321156e-06,
+      "loss": 0.0033,
+      "step": 2931
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794671796909307e-06,
+      "loss": 0.7674,
+      "step": 2932
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794253570607055e-06,
+      "loss": 2.727,
+      "step": 2933
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793835344304803e-06,
+      "loss": 0.0018,
+      "step": 2934
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879341711800255e-06,
+      "loss": 0.0111,
+      "step": 2935
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792998891700299e-06,
+      "loss": 1.4279,
+      "step": 2936
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792580665398047e-06,
+      "loss": 0.0019,
+      "step": 2937
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792162439095794e-06,
+      "loss": 1.6503,
+      "step": 2938
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879174421279354e-06,
+      "loss": 0.002,
+      "step": 2939
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791325986491288e-06,
+      "loss": 0.0021,
+      "step": 2940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790907760189036e-06,
+      "loss": 0.0025,
+      "step": 2941
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790489533886786e-06,
+      "loss": 0.0037,
+      "step": 2942
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790071307584534e-06,
+      "loss": 0.0023,
+      "step": 2943
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789653081282282e-06,
+      "loss": 0.6193,
+      "step": 2944
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878923485498003e-06,
+      "loss": 0.0047,
+      "step": 2945
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788816628677777e-06,
+      "loss": 0.0022,
+      "step": 2946
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788398402375525e-06,
+      "loss": 0.0045,
+      "step": 2947
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787980176073273e-06,
+      "loss": 0.002,
+      "step": 2948
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878756194977102e-06,
+      "loss": 0.0025,
+      "step": 2949
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787143723468769e-06,
+      "loss": 0.0452,
+      "step": 2950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786725497166517e-06,
+      "loss": 0.0019,
+      "step": 2951
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786307270864262e-06,
+      "loss": 0.0024,
+      "step": 2952
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.878588904456201e-06,
+      "loss": 0.0029,
+      "step": 2953
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8785470818259758e-06,
+      "loss": 0.0358,
+      "step": 2954
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8785052591957506e-06,
+      "loss": 1.8153,
+      "step": 2955
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8784634365655254e-06,
+      "loss": 3.2699,
+      "step": 2956
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8784216139353004e-06,
+      "loss": 0.0028,
+      "step": 2957
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8783797913050752e-06,
+      "loss": 3.1434,
+      "step": 2958
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87833796867485e-06,
+      "loss": 0.0019,
+      "step": 2959
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8782961460446247e-06,
+      "loss": 0.0058,
+      "step": 2960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8782543234143995e-06,
+      "loss": 0.0019,
+      "step": 2961
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8782125007841743e-06,
+      "loss": 0.0034,
+      "step": 2962
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.878170678153949e-06,
+      "loss": 0.002,
+      "step": 2963
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8781288555237239e-06,
+      "loss": 0.0022,
+      "step": 2964
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8780870328934987e-06,
+      "loss": 0.3383,
+      "step": 2965
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8780452102632732e-06,
+      "loss": 0.003,
+      "step": 2966
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.878003387633048e-06,
+      "loss": 0.0019,
+      "step": 2967
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8779615650028228e-06,
+      "loss": 0.002,
+      "step": 2968
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8779197423725976e-06,
+      "loss": 0.0024,
+      "step": 2969
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8778779197423724e-06,
+      "loss": 0.0205,
+      "step": 2970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8778360971121472e-06,
+      "loss": 0.0017,
+      "step": 2971
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8777942744819222e-06,
+      "loss": 0.0043,
+      "step": 2972
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.877752451851697e-06,
+      "loss": 0.3294,
+      "step": 2973
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8777106292214717e-06,
+      "loss": 0.0033,
+      "step": 2974
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8776688065912465e-06,
+      "loss": 0.0119,
+      "step": 2975
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8776269839610213e-06,
+      "loss": 0.0021,
+      "step": 2976
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8775851613307961e-06,
+      "loss": 0.0055,
+      "step": 2977
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8775433387005709e-06,
+      "loss": 0.0019,
+      "step": 2978
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8775015160703455e-06,
+      "loss": 0.3247,
+      "step": 2979
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8774596934401203e-06,
+      "loss": 0.6049,
+      "step": 2980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.877417870809895e-06,
+      "loss": 0.0025,
+      "step": 2981
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8773760481796698e-06,
+      "loss": 0.0022,
+      "step": 2982
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8773342255494446e-06,
+      "loss": 0.0058,
+      "step": 2983
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8772924029192194e-06,
+      "loss": 3.5198,
+      "step": 2984
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8772505802889942e-06,
+      "loss": 0.0022,
+      "step": 2985
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8772087576587692e-06,
+      "loss": 0.0018,
+      "step": 2986
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.877166935028544e-06,
+      "loss": 0.0017,
+      "step": 2987
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8771251123983188e-06,
+      "loss": 0.0018,
+      "step": 2988
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8770832897680935e-06,
+      "loss": 0.0079,
+      "step": 2989
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8770414671378683e-06,
+      "loss": 1.699,
+      "step": 2990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8769996445076431e-06,
+      "loss": 0.0023,
+      "step": 2991
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876957821877418e-06,
+      "loss": 2.1606,
+      "step": 2992
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8769159992471925e-06,
+      "loss": 0.0016,
+      "step": 2993
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8768741766169673e-06,
+      "loss": 0.2426,
+      "step": 2994
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876832353986742e-06,
+      "loss": 0.0021,
+      "step": 2995
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8767905313565168e-06,
+      "loss": 0.003,
+      "step": 2996
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8767487087262916e-06,
+      "loss": 1.0577,
+      "step": 2997
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8767068860960664e-06,
+      "loss": 2.1042,
+      "step": 2998
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8766650634658412e-06,
+      "loss": 0.0051,
+      "step": 2999
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876623240835616e-06,
+      "loss": 0.0033,
+      "step": 3000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876581418205391e-06,
+      "loss": 0.0018,
+      "step": 3001
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8765395955751658e-06,
+      "loss": 0.0097,
+      "step": 3002
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764977729449406e-06,
+      "loss": 0.3475,
+      "step": 3003
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764559503147153e-06,
+      "loss": 0.002,
+      "step": 3004
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764141276844901e-06,
+      "loss": 0.7991,
+      "step": 3005
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763723050542647e-06,
+      "loss": 0.0031,
+      "step": 3006
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763304824240395e-06,
+      "loss": 1.8888,
+      "step": 3007
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762886597938143e-06,
+      "loss": 0.0044,
+      "step": 3008
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876246837163589e-06,
+      "loss": 0.0023,
+      "step": 3009
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762050145333638e-06,
+      "loss": 0.177,
+      "step": 3010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761631919031386e-06,
+      "loss": 0.1947,
+      "step": 3011
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761213692729134e-06,
+      "loss": 0.0686,
+      "step": 3012
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760795466426882e-06,
+      "loss": 0.0029,
+      "step": 3013
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876037724012463e-06,
+      "loss": 0.0029,
+      "step": 3014
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759959013822378e-06,
+      "loss": 0.7111,
+      "step": 3015
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759540787520128e-06,
+      "loss": 1.0266,
+      "step": 3016
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759122561217876e-06,
+      "loss": 0.0037,
+      "step": 3017
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758704334915623e-06,
+      "loss": 1.2005,
+      "step": 3018
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758286108613371e-06,
+      "loss": 0.0027,
+      "step": 3019
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757867882311117e-06,
+      "loss": 0.0049,
+      "step": 3020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757449656008865e-06,
+      "loss": 0.0023,
+      "step": 3021
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757031429706613e-06,
+      "loss": 0.349,
+      "step": 3022
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875661320340436e-06,
+      "loss": 0.2121,
+      "step": 3023
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756194977102109e-06,
+      "loss": 0.0046,
+      "step": 3024
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755776750799856e-06,
+      "loss": 0.0943,
+      "step": 3025
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755358524497604e-06,
+      "loss": 0.003,
+      "step": 3026
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8754940298195352e-06,
+      "loss": 0.4607,
+      "step": 3027
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87545220718931e-06,
+      "loss": 0.2706,
+      "step": 3028
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8754103845590848e-06,
+      "loss": 0.8251,
+      "step": 3029
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753685619288598e-06,
+      "loss": 0.0037,
+      "step": 3030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753267392986346e-06,
+      "loss": 0.0021,
+      "step": 3031
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752849166684094e-06,
+      "loss": 0.0026,
+      "step": 3032
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752430940381841e-06,
+      "loss": 0.0033,
+      "step": 3033
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752012714079587e-06,
+      "loss": 0.9544,
+      "step": 3034
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751594487777335e-06,
+      "loss": 0.7259,
+      "step": 3035
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751176261475083e-06,
+      "loss": 0.0028,
+      "step": 3036
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875075803517283e-06,
+      "loss": 0.005,
+      "step": 3037
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750339808870579e-06,
+      "loss": 0.0028,
+      "step": 3038
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749921582568326e-06,
+      "loss": 0.0131,
+      "step": 3039
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749503356266074e-06,
+      "loss": 0.002,
+      "step": 3040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749085129963822e-06,
+      "loss": 2.7018,
+      "step": 3041
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874866690366157e-06,
+      "loss": 0.0028,
+      "step": 3042
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748248677359318e-06,
+      "loss": 0.5948,
+      "step": 3043
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747830451057066e-06,
+      "loss": 0.0112,
+      "step": 3044
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747412224754816e-06,
+      "loss": 0.0018,
+      "step": 3045
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746993998452564e-06,
+      "loss": 0.8105,
+      "step": 3046
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874657577215031e-06,
+      "loss": 1.6894,
+      "step": 3047
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746157545848057e-06,
+      "loss": 0.8668,
+      "step": 3048
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745739319545805e-06,
+      "loss": 0.009,
+      "step": 3049
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745321093243553e-06,
+      "loss": 0.0023,
+      "step": 3050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87449028669413e-06,
+      "loss": 0.0952,
+      "step": 3051
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744484640639049e-06,
+      "loss": 0.0198,
+      "step": 3052
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744066414336797e-06,
+      "loss": 0.0086,
+      "step": 3053
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743648188034544e-06,
+      "loss": 1.6175,
+      "step": 3054
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743229961732292e-06,
+      "loss": 0.2214,
+      "step": 3055
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874281173543004e-06,
+      "loss": 0.002,
+      "step": 3056
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8742393509127788e-06,
+      "loss": 1.4147,
+      "step": 3057
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741975282825536e-06,
+      "loss": 3.0874,
+      "step": 3058
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741557056523286e-06,
+      "loss": 1.7568,
+      "step": 3059
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741138830221034e-06,
+      "loss": 1.9401,
+      "step": 3060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874072060391878e-06,
+      "loss": 0.0022,
+      "step": 3061
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740302377616527e-06,
+      "loss": 0.9471,
+      "step": 3062
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739884151314275e-06,
+      "loss": 0.0038,
+      "step": 3063
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739465925012023e-06,
+      "loss": 0.0017,
+      "step": 3064
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873904769870977e-06,
+      "loss": 2.5929,
+      "step": 3065
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738629472407519e-06,
+      "loss": 0.0055,
+      "step": 3066
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738211246105267e-06,
+      "loss": 0.0053,
+      "step": 3067
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737793019803015e-06,
+      "loss": 0.9243,
+      "step": 3068
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737374793500762e-06,
+      "loss": 0.0033,
+      "step": 3069
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873695656719851e-06,
+      "loss": 0.0031,
+      "step": 3070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736538340896258e-06,
+      "loss": 0.0031,
+      "step": 3071
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736120114594006e-06,
+      "loss": 0.0039,
+      "step": 3072
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735701888291754e-06,
+      "loss": 0.0068,
+      "step": 3073
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735283661989502e-06,
+      "loss": 0.0179,
+      "step": 3074
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873486543568725e-06,
+      "loss": 0.0032,
+      "step": 3075
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734447209384997e-06,
+      "loss": 1.3383,
+      "step": 3076
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734028983082745e-06,
+      "loss": 0.0059,
+      "step": 3077
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733610756780493e-06,
+      "loss": 0.0068,
+      "step": 3078
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873319253047824e-06,
+      "loss": 0.0028,
+      "step": 3079
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732774304175989e-06,
+      "loss": 0.0022,
+      "step": 3080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732356077873737e-06,
+      "loss": 0.0053,
+      "step": 3081
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731937851571485e-06,
+      "loss": 0.0043,
+      "step": 3082
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731519625269232e-06,
+      "loss": 0.0036,
+      "step": 3083
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873110139896698e-06,
+      "loss": 0.0041,
+      "step": 3084
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730683172664728e-06,
+      "loss": 0.4667,
+      "step": 3085
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730264946362476e-06,
+      "loss": 0.0199,
+      "step": 3086
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729846720060222e-06,
+      "loss": 0.0045,
+      "step": 3087
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872942849375797e-06,
+      "loss": 0.1311,
+      "step": 3088
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872901026745572e-06,
+      "loss": 0.003,
+      "step": 3089
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728592041153468e-06,
+      "loss": 0.0077,
+      "step": 3090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728173814851215e-06,
+      "loss": 0.0352,
+      "step": 3091
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727755588548963e-06,
+      "loss": 0.003,
+      "step": 3092
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727337362246711e-06,
+      "loss": 0.0019,
+      "step": 3093
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872691913594446e-06,
+      "loss": 0.3587,
+      "step": 3094
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726500909642207e-06,
+      "loss": 2.427,
+      "step": 3095
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726082683339955e-06,
+      "loss": 0.3807,
+      "step": 3096
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725664457037703e-06,
+      "loss": 0.0028,
+      "step": 3097
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872524623073545e-06,
+      "loss": 0.0052,
+      "step": 3098
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724828004433198e-06,
+      "loss": 0.003,
+      "step": 3099
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724409778130946e-06,
+      "loss": 0.0025,
+      "step": 3100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723991551828692e-06,
+      "loss": 0.0064,
+      "step": 3101
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872357332552644e-06,
+      "loss": 0.2344,
+      "step": 3102
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872315509922419e-06,
+      "loss": 0.0018,
+      "step": 3103
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722736872921938e-06,
+      "loss": 0.0031,
+      "step": 3104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722318646619685e-06,
+      "loss": 0.3343,
+      "step": 3105
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721900420317433e-06,
+      "loss": 0.0021,
+      "step": 3106
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721482194015181e-06,
+      "loss": 0.2721,
+      "step": 3107
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872106396771293e-06,
+      "loss": 0.0027,
+      "step": 3108
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720645741410677e-06,
+      "loss": 0.0025,
+      "step": 3109
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720227515108425e-06,
+      "loss": 0.9337,
+      "step": 3110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719809288806173e-06,
+      "loss": 0.0022,
+      "step": 3111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.871939106250392e-06,
+      "loss": 1.7672,
+      "step": 3112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8718972836201668e-06,
+      "loss": 0.002,
+      "step": 3113
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8718554609899414e-06,
+      "loss": 1.9409,
+      "step": 3114
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8718136383597162e-06,
+      "loss": 0.1365,
+      "step": 3115
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.871771815729491e-06,
+      "loss": 0.0019,
+      "step": 3116
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8717299930992658e-06,
+      "loss": 0.0019,
+      "step": 3117
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8716881704690408e-06,
+      "loss": 0.0029,
+      "step": 3118
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8716463478388156e-06,
+      "loss": 0.003,
+      "step": 3119
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8716045252085903e-06,
+      "loss": 0.0054,
+      "step": 3120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8715627025783651e-06,
+      "loss": 0.723,
+      "step": 3121
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.87152087994814e-06,
+      "loss": 0.0031,
+      "step": 3122
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8714790573179147e-06,
+      "loss": 0.0028,
+      "step": 3123
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8714372346876895e-06,
+      "loss": 2.1354,
+      "step": 3124
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8713954120574643e-06,
+      "loss": 0.0101,
+      "step": 3125
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.871353589427239e-06,
+      "loss": 0.0186,
+      "step": 3126
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8713117667970138e-06,
+      "loss": 0.1025,
+      "step": 3127
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8712699441667884e-06,
+      "loss": 0.0031,
+      "step": 3128
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8712281215365632e-06,
+      "loss": 0.0526,
+      "step": 3129
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.871186298906338e-06,
+      "loss": 0.0574,
+      "step": 3130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8711444762761128e-06,
+      "loss": 0.0085,
+      "step": 3131
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8711026536458876e-06,
+      "loss": 0.8825,
+      "step": 3132
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8710608310156626e-06,
+      "loss": 0.0014,
+      "step": 3133
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8710190083854373e-06,
+      "loss": 0.0391,
+      "step": 3134
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8709771857552121e-06,
+      "loss": 0.5934,
+      "step": 3135
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.870935363124987e-06,
+      "loss": 0.0018,
+      "step": 3136
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8708935404947617e-06,
+      "loss": 0.0018,
+      "step": 3137
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8708517178645365e-06,
+      "loss": 1.5187,
+      "step": 3138
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8708098952343113e-06,
+      "loss": 0.1093,
+      "step": 3139
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.870768072604086e-06,
+      "loss": 1.0485,
+      "step": 3140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8707262499738609e-06,
+      "loss": 0.0068,
+      "step": 3141
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8706844273436354e-06,
+      "loss": 0.0041,
+      "step": 3142
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8706426047134102e-06,
+      "loss": 0.9392,
+      "step": 3143
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.870600782083185e-06,
+      "loss": 0.0031,
+      "step": 3144
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8705589594529598e-06,
+      "loss": 0.0091,
+      "step": 3145
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8705171368227346e-06,
+      "loss": 0.0778,
+      "step": 3146
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8704753141925096e-06,
+      "loss": 0.002,
+      "step": 3147
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8704334915622844e-06,
+      "loss": 0.0022,
+      "step": 3148
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8703916689320591e-06,
+      "loss": 1.7742,
+      "step": 3149
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.870349846301834e-06,
+      "loss": 0.0023,
+      "step": 3150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8703080236716087e-06,
+      "loss": 0.0062,
+      "step": 3151
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8702662010413835e-06,
+      "loss": 0.0157,
+      "step": 3152
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8702243784111583e-06,
+      "loss": 1.3064,
+      "step": 3153
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.870182555780933e-06,
+      "loss": 0.3808,
+      "step": 3154
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8701407331507077e-06,
+      "loss": 1.6538,
+      "step": 3155
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8700989105204824e-06,
+      "loss": 0.3159,
+      "step": 3156
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8700570878902572e-06,
+      "loss": 0.0117,
+      "step": 3157
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.870015265260032e-06,
+      "loss": 0.0036,
+      "step": 3158
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8699734426298068e-06,
+      "loss": 0.7947,
+      "step": 3159
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8699316199995816e-06,
+      "loss": 0.0095,
+      "step": 3160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8698897973693564e-06,
+      "loss": 0.0028,
+      "step": 3161
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8698479747391314e-06,
+      "loss": 0.0046,
+      "step": 3162
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8698061521089062e-06,
+      "loss": 0.0042,
+      "step": 3163
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869764329478681e-06,
+      "loss": 0.1355,
+      "step": 3164
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8697225068484557e-06,
+      "loss": 0.0027,
+      "step": 3165
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8696806842182305e-06,
+      "loss": 0.0041,
+      "step": 3166
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8696388615880053e-06,
+      "loss": 0.3309,
+      "step": 3167
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86959703895778e-06,
+      "loss": 0.0022,
+      "step": 3168
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8695552163275547e-06,
+      "loss": 3.5326,
+      "step": 3169
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8695133936973294e-06,
+      "loss": 0.003,
+      "step": 3170
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8694715710671042e-06,
+      "loss": 0.0049,
+      "step": 3171
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869429748436879e-06,
+      "loss": 0.2224,
+      "step": 3172
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8693879258066538e-06,
+      "loss": 0.1213,
+      "step": 3173
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8693461031764286e-06,
+      "loss": 0.003,
+      "step": 3174
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8693042805462034e-06,
+      "loss": 0.0024,
+      "step": 3175
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8692624579159782e-06,
+      "loss": 0.0276,
+      "step": 3176
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8692206352857532e-06,
+      "loss": 0.0023,
+      "step": 3177
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869178812655528e-06,
+      "loss": 0.0035,
+      "step": 3178
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8691369900253027e-06,
+      "loss": 0.003,
+      "step": 3179
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8690951673950775e-06,
+      "loss": 0.0035,
+      "step": 3180
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8690533447648523e-06,
+      "loss": 0.0034,
+      "step": 3181
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8690115221346269e-06,
+      "loss": 0.0079,
+      "step": 3182
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8689696995044017e-06,
+      "loss": 0.0839,
+      "step": 3183
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8689278768741765e-06,
+      "loss": 0.0027,
+      "step": 3184
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8688860542439512e-06,
+      "loss": 0.0018,
+      "step": 3185
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868844231613726e-06,
+      "loss": 0.0037,
+      "step": 3186
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8688024089835008e-06,
+      "loss": 0.2479,
+      "step": 3187
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8687605863532756e-06,
+      "loss": 0.0022,
+      "step": 3188
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8687187637230504e-06,
+      "loss": 0.0032,
+      "step": 3189
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8686769410928252e-06,
+      "loss": 0.0023,
+      "step": 3190
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8686351184626002e-06,
+      "loss": 0.1256,
+      "step": 3191
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868593295832375e-06,
+      "loss": 0.0554,
+      "step": 3192
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8685514732021497e-06,
+      "loss": 0.0017,
+      "step": 3193
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8685096505719245e-06,
+      "loss": 0.0043,
+      "step": 3194
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8684678279416993e-06,
+      "loss": 0.0022,
+      "step": 3195
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8684260053114739e-06,
+      "loss": 0.2753,
+      "step": 3196
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8683841826812487e-06,
+      "loss": 0.0025,
+      "step": 3197
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8683423600510235e-06,
+      "loss": 0.004,
+      "step": 3198
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8683005374207982e-06,
+      "loss": 0.0024,
+      "step": 3199
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868258714790573e-06,
+      "loss": 0.0044,
+      "step": 3200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8682168921603478e-06,
+      "loss": 0.0024,
+      "step": 3201
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8681750695301226e-06,
+      "loss": 0.5451,
+      "step": 3202
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8681332468998974e-06,
+      "loss": 0.2772,
+      "step": 3203
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8680914242696722e-06,
+      "loss": 0.0045,
+      "step": 3204
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868049601639447e-06,
+      "loss": 2.2474,
+      "step": 3205
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868007779009222e-06,
+      "loss": 0.0036,
+      "step": 3206
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8679659563789968e-06,
+      "loss": 0.0023,
+      "step": 3207
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8679241337487715e-06,
+      "loss": 0.0022,
+      "step": 3208
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8678823111185461e-06,
+      "loss": 0.0062,
+      "step": 3209
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867840488488321e-06,
+      "loss": 3.0629,
+      "step": 3210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8677986658580957e-06,
+      "loss": 0.0017,
+      "step": 3211
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8677568432278705e-06,
+      "loss": 0.0081,
+      "step": 3212
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8677150205976453e-06,
+      "loss": 3.8692,
+      "step": 3213
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86767319796742e-06,
+      "loss": 0.0446,
+      "step": 3214
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8676313753371948e-06,
+      "loss": 0.0015,
+      "step": 3215
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8675895527069696e-06,
+      "loss": 0.0017,
+      "step": 3216
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8675477300767444e-06,
+      "loss": 0.0016,
+      "step": 3217
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8675059074465192e-06,
+      "loss": 0.0013,
+      "step": 3218
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867464084816294e-06,
+      "loss": 0.002,
+      "step": 3219
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8674222621860688e-06,
+      "loss": 0.0024,
+      "step": 3220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8673804395558438e-06,
+      "loss": 0.0026,
+      "step": 3221
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8673386169256185e-06,
+      "loss": 0.003,
+      "step": 3222
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8672967942953931e-06,
+      "loss": 1.0745,
+      "step": 3223
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867254971665168e-06,
+      "loss": 0.0018,
+      "step": 3224
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8672131490349427e-06,
+      "loss": 0.0015,
+      "step": 3225
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8671713264047175e-06,
+      "loss": 0.0016,
+      "step": 3226
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8671295037744923e-06,
+      "loss": 0.0451,
+      "step": 3227
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867087681144267e-06,
+      "loss": 0.0073,
+      "step": 3228
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8670458585140418e-06,
+      "loss": 0.0016,
+      "step": 3229
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8670040358838166e-06,
+      "loss": 1.7102,
+      "step": 3230
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8669622132535914e-06,
+      "loss": 2.9779,
+      "step": 3231
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8669203906233662e-06,
+      "loss": 0.0819,
+      "step": 3232
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866878567993141e-06,
+      "loss": 0.002,
+      "step": 3233
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8668367453629158e-06,
+      "loss": 0.0887,
+      "step": 3234
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8667949227326908e-06,
+      "loss": 0.004,
+      "step": 3235
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8667531001024656e-06,
+      "loss": 0.0045,
+      "step": 3236
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8667112774722401e-06,
+      "loss": 0.0025,
+      "step": 3237
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866669454842015e-06,
+      "loss": 0.1177,
+      "step": 3238
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8666276322117897e-06,
+      "loss": 0.0022,
+      "step": 3239
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8665858095815645e-06,
+      "loss": 0.6783,
+      "step": 3240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8665439869513393e-06,
+      "loss": 0.0045,
+      "step": 3241
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866502164321114e-06,
+      "loss": 0.0101,
+      "step": 3242
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8664603416908888e-06,
+      "loss": 0.0029,
+      "step": 3243
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8664185190606636e-06,
+      "loss": 2.6739,
+      "step": 3244
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8663766964304384e-06,
+      "loss": 0.0019,
+      "step": 3245
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8663348738002132e-06,
+      "loss": 0.0027,
+      "step": 3246
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866293051169988e-06,
+      "loss": 1.9315,
+      "step": 3247
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8662512285397628e-06,
+      "loss": 0.0144,
+      "step": 3248
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8662094059095376e-06,
+      "loss": 3.612,
+      "step": 3249
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8661675832793124e-06,
+      "loss": 0.0019,
+      "step": 3250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8661257606490871e-06,
+      "loss": 0.0019,
+      "step": 3251
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866083938018862e-06,
+      "loss": 0.002,
+      "step": 3252
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8660421153886367e-06,
+      "loss": 0.576,
+      "step": 3253
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8660002927584115e-06,
+      "loss": 2.2874,
+      "step": 3254
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8659584701281863e-06,
+      "loss": 0.0021,
+      "step": 3255
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865916647497961e-06,
+      "loss": 0.0028,
+      "step": 3256
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8658748248677359e-06,
+      "loss": 1.8802,
+      "step": 3257
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8658330022375106e-06,
+      "loss": 0.0078,
+      "step": 3258
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8657911796072854e-06,
+      "loss": 0.0052,
+      "step": 3259
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8657493569770602e-06,
+      "loss": 0.0025,
+      "step": 3260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865707534346835e-06,
+      "loss": 0.0151,
+      "step": 3261
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8656657117166098e-06,
+      "loss": 0.7493,
+      "step": 3262
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8656238890863844e-06,
+      "loss": 0.0114,
+      "step": 3263
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8655820664561594e-06,
+      "loss": 1.8713,
+      "step": 3264
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8655402438259341e-06,
+      "loss": 0.0034,
+      "step": 3265
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865498421195709e-06,
+      "loss": 0.0079,
+      "step": 3266
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8654565985654837e-06,
+      "loss": 0.0553,
+      "step": 3267
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8654147759352585e-06,
+      "loss": 0.0046,
+      "step": 3268
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8653729533050333e-06,
+      "loss": 0.0031,
+      "step": 3269
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865331130674808e-06,
+      "loss": 0.0035,
+      "step": 3270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8652893080445829e-06,
+      "loss": 0.0081,
+      "step": 3271
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8652474854143577e-06,
+      "loss": 0.0021,
+      "step": 3272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8652056627841324e-06,
+      "loss": 0.0015,
+      "step": 3273
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8651638401539072e-06,
+      "loss": 0.003,
+      "step": 3274
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.865122017523682e-06,
+      "loss": 0.0032,
+      "step": 3275
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8650801948934568e-06,
+      "loss": 0.0192,
+      "step": 3276
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8650383722632314e-06,
+      "loss": 2.5532,
+      "step": 3277
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8649965496330062e-06,
+      "loss": 0.0026,
+      "step": 3278
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8649547270027812e-06,
+      "loss": 0.222,
+      "step": 3279
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.864912904372556e-06,
+      "loss": 0.0027,
+      "step": 3280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8648710817423307e-06,
+      "loss": 0.0089,
+      "step": 3281
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8648292591121055e-06,
+      "loss": 0.0065,
+      "step": 3282
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8647874364818803e-06,
+      "loss": 0.0014,
+      "step": 3283
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.864745613851655e-06,
+      "loss": 0.6449,
+      "step": 3284
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8647037912214299e-06,
+      "loss": 2.3621,
+      "step": 3285
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8646619685912047e-06,
+      "loss": 0.0581,
+      "step": 3286
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8646201459609794e-06,
+      "loss": 0.0018,
+      "step": 3287
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8645783233307542e-06,
+      "loss": 0.0495,
+      "step": 3288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.864536500700529e-06,
+      "loss": 0.0022,
+      "step": 3289
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8644946780703036e-06,
+      "loss": 0.0044,
+      "step": 3290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8644528554400784e-06,
+      "loss": 0.0094,
+      "step": 3291
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8644110328098532e-06,
+      "loss": 2.122,
+      "step": 3292
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.864369210179628e-06,
+      "loss": 0.0022,
+      "step": 3293
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.864327387549403e-06,
+      "loss": 0.2294,
+      "step": 3294
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8642855649191777e-06,
+      "loss": 0.0028,
+      "step": 3295
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8642437422889525e-06,
+      "loss": 0.0032,
+      "step": 3296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8642019196587273e-06,
+      "loss": 0.0186,
+      "step": 3297
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.864160097028502e-06,
+      "loss": 0.0038,
+      "step": 3298
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8641182743982769e-06,
+      "loss": 0.7909,
+      "step": 3299
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8640764517680517e-06,
+      "loss": 2.5275,
+      "step": 3300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8640346291378265e-06,
+      "loss": 0.0045,
+      "step": 3301
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8639928065076012e-06,
+      "loss": 0.0028,
+      "step": 3302
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.863950983877376e-06,
+      "loss": 0.0027,
+      "step": 3303
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8639091612471506e-06,
+      "loss": 0.0027,
+      "step": 3304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8638673386169254e-06,
+      "loss": 0.0039,
+      "step": 3305
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8638255159867002e-06,
+      "loss": 0.0169,
+      "step": 3306
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.863783693356475e-06,
+      "loss": 0.0019,
+      "step": 3307
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.86374187072625e-06,
+      "loss": 0.1534,
+      "step": 3308
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8637000480960247e-06,
+      "loss": 0.0019,
+      "step": 3309
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8636582254657995e-06,
+      "loss": 0.0019,
+      "step": 3310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8636164028355743e-06,
+      "loss": 0.002,
+      "step": 3311
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.863574580205349e-06,
+      "loss": 1.4906,
+      "step": 3312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8635327575751239e-06,
+      "loss": 0.0028,
+      "step": 3313
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8634909349448987e-06,
+      "loss": 0.004,
+      "step": 3314
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8634491123146735e-06,
+      "loss": 0.0023,
+      "step": 3315
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8634072896844482e-06,
+      "loss": 0.0023,
+      "step": 3316
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8633654670542228e-06,
+      "loss": 0.0036,
+      "step": 3317
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8633236444239976e-06,
+      "loss": 0.0015,
+      "step": 3318
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8632818217937724e-06,
+      "loss": 0.0024,
+      "step": 3319
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8632399991635472e-06,
+      "loss": 0.0104,
+      "step": 3320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.863198176533322e-06,
+      "loss": 0.0034,
+      "step": 3321
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8631563539030968e-06,
+      "loss": 0.0015,
+      "step": 3322
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8631145312728718e-06,
+      "loss": 0.0018,
+      "step": 3323
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8630727086426465e-06,
+      "loss": 0.0035,
+      "step": 3324
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8630308860124213e-06,
+      "loss": 0.0018,
+      "step": 3325
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8629890633821961e-06,
+      "loss": 1.6122,
+      "step": 3326
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862947240751971e-06,
+      "loss": 1.289,
+      "step": 3327
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8629054181217457e-06,
+      "loss": 0.0018,
+      "step": 3328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8628635954915205e-06,
+      "loss": 0.0018,
+      "step": 3329
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8628217728612953e-06,
+      "loss": 0.0028,
+      "step": 3330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8627799502310698e-06,
+      "loss": 0.0013,
+      "step": 3331
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8627381276008446e-06,
+      "loss": 0.0017,
+      "step": 3332
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8626963049706194e-06,
+      "loss": 0.3924,
+      "step": 3333
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8626544823403942e-06,
+      "loss": 0.0026,
+      "step": 3334
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862612659710169e-06,
+      "loss": 3.7134,
+      "step": 3335
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8625708370799438e-06,
+      "loss": 2.7922,
+      "step": 3336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8625290144497186e-06,
+      "loss": 0.0022,
+      "step": 3337
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8624871918194935e-06,
+      "loss": 0.0818,
+      "step": 3338
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8624453691892683e-06,
+      "loss": 0.0083,
+      "step": 3339
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8624035465590431e-06,
+      "loss": 0.8309,
+      "step": 3340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862361723928818e-06,
+      "loss": 0.0028,
+      "step": 3341
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8623199012985927e-06,
+      "loss": 0.0025,
+      "step": 3342
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8622780786683675e-06,
+      "loss": 0.0032,
+      "step": 3343
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8622362560381423e-06,
+      "loss": 0.0016,
+      "step": 3344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8621944334079168e-06,
+      "loss": 0.2633,
+      "step": 3345
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8621526107776916e-06,
+      "loss": 0.0045,
+      "step": 3346
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8621107881474664e-06,
+      "loss": 0.0028,
+      "step": 3347
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8620689655172412e-06,
+      "loss": 0.0031,
+      "step": 3348
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862027142887016e-06,
+      "loss": 2.734,
+      "step": 3349
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8619853202567908e-06,
+      "loss": 0.0015,
+      "step": 3350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8619434976265656e-06,
+      "loss": 0.2286,
+      "step": 3351
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8619016749963406e-06,
+      "loss": 0.0035,
+      "step": 3352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8618598523661153e-06,
+      "loss": 0.0031,
+      "step": 3353
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8618180297358901e-06,
+      "loss": 0.0019,
+      "step": 3354
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861776207105665e-06,
+      "loss": 0.0026,
+      "step": 3355
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8617343844754397e-06,
+      "loss": 0.0033,
+      "step": 3356
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8616925618452145e-06,
+      "loss": 0.002,
+      "step": 3357
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861650739214989e-06,
+      "loss": 0.0015,
+      "step": 3358
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8616089165847638e-06,
+      "loss": 0.0019,
+      "step": 3359
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8615670939545386e-06,
+      "loss": 2.722,
+      "step": 3360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8615252713243134e-06,
+      "loss": 0.1198,
+      "step": 3361
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8614834486940882e-06,
+      "loss": 0.0015,
+      "step": 3362
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861441626063863e-06,
+      "loss": 0.0014,
+      "step": 3363
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8613998034336378e-06,
+      "loss": 0.0017,
+      "step": 3364
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8613579808034126e-06,
+      "loss": 0.0016,
+      "step": 3365
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8613161581731874e-06,
+      "loss": 0.0017,
+      "step": 3366
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8612743355429624e-06,
+      "loss": 0.002,
+      "step": 3367
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8612325129127371e-06,
+      "loss": 0.4272,
+      "step": 3368
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861190690282512e-06,
+      "loss": 1.9455,
+      "step": 3369
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8611488676522867e-06,
+      "loss": 0.0021,
+      "step": 3370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8611070450220615e-06,
+      "loss": 1.2533,
+      "step": 3371
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861065222391836e-06,
+      "loss": 0.5678,
+      "step": 3372
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8610233997616109e-06,
+      "loss": 1.9836,
+      "step": 3373
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8609815771313856e-06,
+      "loss": 2.1955,
+      "step": 3374
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8609397545011604e-06,
+      "loss": 0.6324,
+      "step": 3375
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8608979318709352e-06,
+      "loss": 0.0028,
+      "step": 3376
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.86085610924071e-06,
+      "loss": 0.0016,
+      "step": 3377
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8608142866104848e-06,
+      "loss": 0.3031,
+      "step": 3378
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8607724639802596e-06,
+      "loss": 0.0021,
+      "step": 3379
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8607306413500344e-06,
+      "loss": 0.0018,
+      "step": 3380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8606888187198091e-06,
+      "loss": 0.002,
+      "step": 3381
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8606469960895841e-06,
+      "loss": 0.0024,
+      "step": 3382
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860605173459359e-06,
+      "loss": 0.0032,
+      "step": 3383
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8605633508291337e-06,
+      "loss": 0.0037,
+      "step": 3384
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8605215281989083e-06,
+      "loss": 0.0036,
+      "step": 3385
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860479705568683e-06,
+      "loss": 0.0019,
+      "step": 3386
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8604378829384579e-06,
+      "loss": 0.0017,
+      "step": 3387
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8603960603082327e-06,
+      "loss": 0.0107,
+      "step": 3388
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8603542376780074e-06,
+      "loss": 0.0042,
+      "step": 3389
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8603124150477822e-06,
+      "loss": 0.0026,
+      "step": 3390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860270592417557e-06,
+      "loss": 0.0527,
+      "step": 3391
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8602287697873318e-06,
+      "loss": 0.2162,
+      "step": 3392
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8601869471571066e-06,
+      "loss": 0.3105,
+      "step": 3393
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8601451245268814e-06,
+      "loss": 0.0029,
+      "step": 3394
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8601033018966562e-06,
+      "loss": 0.5063,
+      "step": 3395
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8600614792664312e-06,
+      "loss": 0.0027,
+      "step": 3396
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860019656636206e-06,
+      "loss": 0.0031,
+      "step": 3397
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8599778340059807e-06,
+      "loss": 0.0019,
+      "step": 3398
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8599360113757553e-06,
+      "loss": 0.002,
+      "step": 3399
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.85989418874553e-06,
+      "loss": 0.0038,
+      "step": 3400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8598523661153049e-06,
+      "loss": 0.0034,
+      "step": 3401
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8598105434850797e-06,
+      "loss": 0.971,
+      "step": 3402
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8597687208548544e-06,
+      "loss": 0.0022,
+      "step": 3403
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8597268982246292e-06,
+      "loss": 0.0022,
+      "step": 3404
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859685075594404e-06,
+      "loss": 0.0017,
+      "step": 3405
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8596432529641788e-06,
+      "loss": 0.0057,
+      "step": 3406
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8596014303339536e-06,
+      "loss": 0.0072,
+      "step": 3407
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8595596077037284e-06,
+      "loss": 0.0019,
+      "step": 3408
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8595177850735032e-06,
+      "loss": 0.003,
+      "step": 3409
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859475962443278e-06,
+      "loss": 0.0015,
+      "step": 3410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859434139813053e-06,
+      "loss": 0.0015,
+      "step": 3411
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8593923171828275e-06,
+      "loss": 0.0046,
+      "step": 3412
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8593504945526023e-06,
+      "loss": 0.0023,
+      "step": 3413
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859308671922377e-06,
+      "loss": 0.0018,
+      "step": 3414
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8592668492921519e-06,
+      "loss": 0.7897,
+      "step": 3415
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8592250266619267e-06,
+      "loss": 0.8198,
+      "step": 3416
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8591832040317015e-06,
+      "loss": 0.0095,
+      "step": 3417
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8591413814014762e-06,
+      "loss": 0.0022,
+      "step": 3418
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859099558771251e-06,
+      "loss": 0.7503,
+      "step": 3419
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8590577361410258e-06,
+      "loss": 0.0017,
+      "step": 3420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8590159135108006e-06,
+      "loss": 3.4917,
+      "step": 3421
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8589740908805754e-06,
+      "loss": 0.0014,
+      "step": 3422
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8589322682503502e-06,
+      "loss": 0.6693,
+      "step": 3423
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858890445620125e-06,
+      "loss": 0.0024,
+      "step": 3424
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8588486229899e-06,
+      "loss": 0.0041,
+      "step": 3425
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8588068003596745e-06,
+      "loss": 0.0031,
+      "step": 3426
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8587649777294493e-06,
+      "loss": 0.003,
+      "step": 3427
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8587231550992241e-06,
+      "loss": 0.0019,
+      "step": 3428
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8586813324689989e-06,
+      "loss": 0.0018,
+      "step": 3429
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8586395098387737e-06,
+      "loss": 0.0018,
+      "step": 3430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8585976872085485e-06,
+      "loss": 0.0017,
+      "step": 3431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8585558645783233e-06,
+      "loss": 3.2662,
+      "step": 3432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.858514041948098e-06,
+      "loss": 0.0047,
+      "step": 3433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8584722193178728e-06,
+      "loss": 0.0021,
+      "step": 3434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8584303966876476e-06,
+      "loss": 1.715,
+      "step": 3435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8583885740574224e-06,
+      "loss": 0.0027,
+      "step": 3436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8583467514271972e-06,
+      "loss": 0.003,
+      "step": 3437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.858304928796972e-06,
+      "loss": 0.0018,
+      "step": 3438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8582631061667465e-06,
+      "loss": 0.0763,
+      "step": 3439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8582212835365215e-06,
+      "loss": 0.0013,
+      "step": 3440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8581794609062963e-06,
+      "loss": 0.0014,
+      "step": 3441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8581376382760711e-06,
+      "loss": 0.0029,
+      "step": 3442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.858095815645846e-06,
+      "loss": 0.0022,
+      "step": 3443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8580539930156207e-06,
+      "loss": 0.0035,
+      "step": 3444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8580121703853955e-06,
+      "loss": 0.5271,
+      "step": 3445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8579703477551703e-06,
+      "loss": 0.0045,
+      "step": 3446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.857928525124945e-06,
+      "loss": 0.0014,
+      "step": 3447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8578867024947198e-06,
+      "loss": 0.0031,
+      "step": 3448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8578448798644946e-06,
+      "loss": 0.0015,
+      "step": 3449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8578030572342694e-06,
+      "loss": 0.0341,
+      "step": 3450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8577612346040442e-06,
+      "loss": 0.9172,
+      "step": 3451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.857719411973819e-06,
+      "loss": 1.4148,
+      "step": 3452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8576775893435936e-06,
+      "loss": 0.7867,
+      "step": 3453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8576357667133683e-06,
+      "loss": 0.0028,
+      "step": 3454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8575939440831433e-06,
+      "loss": 0.0018,
+      "step": 3455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8575521214529181e-06,
+      "loss": 0.0048,
+      "step": 3456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.857510298822693e-06,
+      "loss": 0.0021,
+      "step": 3457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8574684761924677e-06,
+      "loss": 4.1205,
+      "step": 3458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8574266535622425e-06,
+      "loss": 0.0014,
+      "step": 3459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8573848309320173e-06,
+      "loss": 0.0024,
+      "step": 3460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.857343008301792e-06,
+      "loss": 3.7348,
+      "step": 3461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8573011856715668e-06,
+      "loss": 0.0031,
+      "step": 3462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8572593630413416e-06,
+      "loss": 0.0087,
+      "step": 3463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8572175404111164e-06,
+      "loss": 0.0054,
+      "step": 3464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8571757177808912e-06,
+      "loss": 1.7801,
+      "step": 3465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8571338951506658e-06,
+      "loss": 0.0026,
+      "step": 3466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8570920725204406e-06,
+      "loss": 0.0028,
+      "step": 3467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8570502498902153e-06,
+      "loss": 0.0035,
+      "step": 3468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8570084272599903e-06,
+      "loss": 0.4737,
+      "step": 3469
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8569666046297651e-06,
+      "loss": 0.5366,
+      "step": 3470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85692478199954e-06,
+      "loss": 3.8737,
+      "step": 3471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8568829593693147e-06,
+      "loss": 0.0018,
+      "step": 3472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8568411367390895e-06,
+      "loss": 0.5599,
+      "step": 3473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8567993141088643e-06,
+      "loss": 0.0049,
+      "step": 3474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856757491478639e-06,
+      "loss": 0.0019,
+      "step": 3475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8567156688484139e-06,
+      "loss": 0.0028,
+      "step": 3476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8566738462181886e-06,
+      "loss": 2.1396,
+      "step": 3477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8566320235879634e-06,
+      "loss": 0.0081,
+      "step": 3478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8565902009577382e-06,
+      "loss": 0.0057,
+      "step": 3479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8565483783275128e-06,
+      "loss": 0.0027,
+      "step": 3480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8565065556972876e-06,
+      "loss": 0.0022,
+      "step": 3481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8564647330670624e-06,
+      "loss": 1.16,
+      "step": 3482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8564229104368371e-06,
+      "loss": 0.2277,
+      "step": 3483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8563810878066121e-06,
+      "loss": 1.5056,
+      "step": 3484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856339265176387e-06,
+      "loss": 0.0024,
+      "step": 3485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8562974425461617e-06,
+      "loss": 0.0049,
+      "step": 3486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8562556199159365e-06,
+      "loss": 0.0313,
+      "step": 3487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8562137972857113e-06,
+      "loss": 1.6052,
+      "step": 3488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856171974655486e-06,
+      "loss": 0.0022,
+      "step": 3489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8561301520252609e-06,
+      "loss": 0.0026,
+      "step": 3490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8560883293950356e-06,
+      "loss": 0.0369,
+      "step": 3491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8560465067648104e-06,
+      "loss": 0.0047,
+      "step": 3492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856004684134585e-06,
+      "loss": 0.0045,
+      "step": 3493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8559628615043598e-06,
+      "loss": 2.3913,
+      "step": 3494
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8559210388741346e-06,
+      "loss": 0.0054,
+      "step": 3495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8558792162439094e-06,
+      "loss": 0.0038,
+      "step": 3496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8558373936136842e-06,
+      "loss": 0.0016,
+      "step": 3497
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855795570983459e-06,
+      "loss": 1.3393,
+      "step": 3498
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855753748353234e-06,
+      "loss": 1.5841,
+      "step": 3499
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8557119257230087e-06,
+      "loss": 0.0026,
+      "step": 3500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8556701030927835e-06,
+      "loss": 0.0125,
+      "step": 3501
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8556282804625583e-06,
+      "loss": 0.0195,
+      "step": 3502
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855586457832333e-06,
+      "loss": 0.0026,
+      "step": 3503
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8555446352021079e-06,
+      "loss": 0.0588,
+      "step": 3504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8555028125718827e-06,
+      "loss": 0.005,
+      "step": 3505
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8554609899416574e-06,
+      "loss": 0.0018,
+      "step": 3506
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855419167311432e-06,
+      "loss": 0.0029,
+      "step": 3507
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8553773446812068e-06,
+      "loss": 0.0021,
+      "step": 3508
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8553355220509816e-06,
+      "loss": 0.0762,
+      "step": 3509
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8552936994207564e-06,
+      "loss": 0.0078,
+      "step": 3510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8552518767905312e-06,
+      "loss": 0.0081,
+      "step": 3511
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855210054160306e-06,
+      "loss": 0.0017,
+      "step": 3512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855168231530081e-06,
+      "loss": 0.0021,
+      "step": 3513
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8551264088998557e-06,
+      "loss": 0.0095,
+      "step": 3514
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8550845862696305e-06,
+      "loss": 0.0015,
+      "step": 3515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8550427636394053e-06,
+      "loss": 0.0015,
+      "step": 3516
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85500094100918e-06,
+      "loss": 0.0024,
+      "step": 3517
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8549591183789549e-06,
+      "loss": 0.0018,
+      "step": 3518
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8549172957487297e-06,
+      "loss": 0.0017,
+      "step": 3519
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8548754731185042e-06,
+      "loss": 0.0179,
+      "step": 3520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854833650488279e-06,
+      "loss": 3.4199,
+      "step": 3521
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8547918278580538e-06,
+      "loss": 0.0016,
+      "step": 3522
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8547500052278286e-06,
+      "loss": 0.0025,
+      "step": 3523
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8547081825976034e-06,
+      "loss": 0.0016,
+      "step": 3524
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8546663599673782e-06,
+      "loss": 0.0018,
+      "step": 3525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854624537337153e-06,
+      "loss": 0.355,
+      "step": 3526
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8545827147069277e-06,
+      "loss": 0.0052,
+      "step": 3527
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8545408920767027e-06,
+      "loss": 2.3428,
+      "step": 3528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8544990694464775e-06,
+      "loss": 0.0032,
+      "step": 3529
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8544572468162523e-06,
+      "loss": 0.336,
+      "step": 3530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854415424186027e-06,
+      "loss": 0.0024,
+      "step": 3531
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8543736015558019e-06,
+      "loss": 1.3605,
+      "step": 3532
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8543317789255767e-06,
+      "loss": 0.0026,
+      "step": 3533
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8542899562953512e-06,
+      "loss": 0.0029,
+      "step": 3534
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854248133665126e-06,
+      "loss": 0.5919,
+      "step": 3535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8542063110349008e-06,
+      "loss": 0.0036,
+      "step": 3536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8541644884046756e-06,
+      "loss": 3.3682,
+      "step": 3537
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8541226657744504e-06,
+      "loss": 0.3102,
+      "step": 3538
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8540808431442252e-06,
+      "loss": 0.0026,
+      "step": 3539
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854039020514e-06,
+      "loss": 0.0016,
+      "step": 3540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8539971978837747e-06,
+      "loss": 0.0024,
+      "step": 3541
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8539553752535495e-06,
+      "loss": 0.002,
+      "step": 3542
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8539135526233245e-06,
+      "loss": 1.9209,
+      "step": 3543
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8538717299930993e-06,
+      "loss": 0.0905,
+      "step": 3544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8538299073628741e-06,
+      "loss": 0.0025,
+      "step": 3545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8537880847326489e-06,
+      "loss": 0.004,
+      "step": 3546
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8537462621024237e-06,
+      "loss": 0.003,
+      "step": 3547
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8537044394721983e-06,
+      "loss": 0.0757,
+      "step": 3548
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853662616841973e-06,
+      "loss": 0.0032,
+      "step": 3549
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8536207942117478e-06,
+      "loss": 0.0041,
+      "step": 3550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8535789715815226e-06,
+      "loss": 0.0046,
+      "step": 3551
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8535371489512974e-06,
+      "loss": 3.6032,
+      "step": 3552
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8534953263210722e-06,
+      "loss": 0.6012,
+      "step": 3553
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853453503690847e-06,
+      "loss": 0.0019,
+      "step": 3554
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8534116810606218e-06,
+      "loss": 0.0015,
+      "step": 3555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8533698584303965e-06,
+      "loss": 1.1175,
+      "step": 3556
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8533280358001715e-06,
+      "loss": 0.002,
+      "step": 3557
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8532862131699463e-06,
+      "loss": 2.1984,
+      "step": 3558
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8532443905397211e-06,
+      "loss": 0.0043,
+      "step": 3559
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853202567909496e-06,
+      "loss": 0.0031,
+      "step": 3560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8531607452792705e-06,
+      "loss": 0.0014,
+      "step": 3561
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8531189226490453e-06,
+      "loss": 0.0016,
+      "step": 3562
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85307710001882e-06,
+      "loss": 0.2884,
+      "step": 3563
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8530352773885948e-06,
+      "loss": 0.014,
+      "step": 3564
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8529934547583696e-06,
+      "loss": 2.4458,
+      "step": 3565
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8529516321281444e-06,
+      "loss": 0.5192,
+      "step": 3566
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8529098094979192e-06,
+      "loss": 0.601,
+      "step": 3567
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852867986867694e-06,
+      "loss": 0.0015,
+      "step": 3568
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8528261642374688e-06,
+      "loss": 0.3275,
+      "step": 3569
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8527843416072436e-06,
+      "loss": 0.3321,
+      "step": 3570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8527425189770183e-06,
+      "loss": 0.078,
+      "step": 3571
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8527006963467933e-06,
+      "loss": 2.835,
+      "step": 3572
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8526588737165681e-06,
+      "loss": 0.7875,
+      "step": 3573
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852617051086343e-06,
+      "loss": 0.6596,
+      "step": 3574
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8525752284561175e-06,
+      "loss": 0.0022,
+      "step": 3575
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8525334058258923e-06,
+      "loss": 0.0017,
+      "step": 3576
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852491583195667e-06,
+      "loss": 0.0015,
+      "step": 3577
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8524497605654418e-06,
+      "loss": 0.002,
+      "step": 3578
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8524079379352166e-06,
+      "loss": 0.003,
+      "step": 3579
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8523661153049914e-06,
+      "loss": 0.0922,
+      "step": 3580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8523242926747662e-06,
+      "loss": 0.0032,
+      "step": 3581
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852282470044541e-06,
+      "loss": 0.0012,
+      "step": 3582
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8522406474143158e-06,
+      "loss": 0.0024,
+      "step": 3583
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8521988247840906e-06,
+      "loss": 0.0017,
+      "step": 3584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8521570021538653e-06,
+      "loss": 0.0043,
+      "step": 3585
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8521151795236401e-06,
+      "loss": 0.4338,
+      "step": 3586
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8520733568934151e-06,
+      "loss": 0.0038,
+      "step": 3587
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8520315342631897e-06,
+      "loss": 0.0037,
+      "step": 3588
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8519897116329645e-06,
+      "loss": 0.0083,
+      "step": 3589
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8519478890027393e-06,
+      "loss": 0.0014,
+      "step": 3590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.851906066372514e-06,
+      "loss": 0.0659,
+      "step": 3591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8518642437422889e-06,
+      "loss": 3.6894,
+      "step": 3592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8518224211120636e-06,
+      "loss": 0.0015,
+      "step": 3593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8517805984818384e-06,
+      "loss": 0.7376,
+      "step": 3594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8517387758516132e-06,
+      "loss": 1.3164,
+      "step": 3595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.851696953221388e-06,
+      "loss": 2.2307,
+      "step": 3596
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8516551305911628e-06,
+      "loss": 0.0038,
+      "step": 3597
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8516133079609376e-06,
+      "loss": 0.4111,
+      "step": 3598
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8515714853307124e-06,
+      "loss": 0.0045,
+      "step": 3599
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8515296627004871e-06,
+      "loss": 0.2717,
+      "step": 3600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8514878400702621e-06,
+      "loss": 0.0035,
+      "step": 3601
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8514460174400367e-06,
+      "loss": 0.4588,
+      "step": 3602
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8514041948098115e-06,
+      "loss": 0.0042,
+      "step": 3603
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8513623721795863e-06,
+      "loss": 0.0031,
+      "step": 3604
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.851320549549361e-06,
+      "loss": 0.0022,
+      "step": 3605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8512787269191359e-06,
+      "loss": 0.0034,
+      "step": 3606
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8512369042889106e-06,
+      "loss": 0.2812,
+      "step": 3607
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8511950816586854e-06,
+      "loss": 0.0016,
+      "step": 3608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8511532590284602e-06,
+      "loss": 0.0122,
+      "step": 3609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.851111436398235e-06,
+      "loss": 0.0015,
+      "step": 3610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8510696137680098e-06,
+      "loss": 1.6077,
+      "step": 3611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8510277911377846e-06,
+      "loss": 0.0054,
+      "step": 3612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8509859685075594e-06,
+      "loss": 0.0024,
+      "step": 3613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8509441458773342e-06,
+      "loss": 0.6792,
+      "step": 3614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8509023232471087e-06,
+      "loss": 0.0024,
+      "step": 3615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8508605006168837e-06,
+      "loss": 1.0166,
+      "step": 3616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8508186779866585e-06,
+      "loss": 1.8307,
+      "step": 3617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8507768553564333e-06,
+      "loss": 1.0784,
+      "step": 3618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.850735032726208e-06,
+      "loss": 0.0022,
+      "step": 3619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8506932100959829e-06,
+      "loss": 0.021,
+      "step": 3620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8506513874657577e-06,
+      "loss": 0.0025,
+      "step": 3621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8506095648355324e-06,
+      "loss": 0.0145,
+      "step": 3622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8505677422053072e-06,
+      "loss": 0.0024,
+      "step": 3623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.850525919575082e-06,
+      "loss": 0.0152,
+      "step": 3624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8504840969448568e-06,
+      "loss": 3.0117,
+      "step": 3625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8504422743146316e-06,
+      "loss": 0.0112,
+      "step": 3626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8504004516844064e-06,
+      "loss": 0.0236,
+      "step": 3627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.850358629054181e-06,
+      "loss": 0.286,
+      "step": 3628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8503168064239557e-06,
+      "loss": 0.0033,
+      "step": 3629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8502749837937307e-06,
+      "loss": 0.01,
+      "step": 3630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8502331611635055e-06,
+      "loss": 0.2176,
+      "step": 3631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8501913385332803e-06,
+      "loss": 0.0036,
+      "step": 3632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.850149515903055e-06,
+      "loss": 0.0854,
+      "step": 3633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8501076932728299e-06,
+      "loss": 0.849,
+      "step": 3634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8500658706426047e-06,
+      "loss": 0.1199,
+      "step": 3635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8500240480123795e-06,
+      "loss": 0.0094,
+      "step": 3636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8499822253821542e-06,
+      "loss": 0.5076,
+      "step": 3637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849940402751929e-06,
+      "loss": 0.003,
+      "step": 3638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8498985801217038e-06,
+      "loss": 0.002,
+      "step": 3639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8498567574914786e-06,
+      "loss": 0.0059,
+      "step": 3640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8498149348612534e-06,
+      "loss": 0.0025,
+      "step": 3641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849773112231028e-06,
+      "loss": 0.0551,
+      "step": 3642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8497312896008027e-06,
+      "loss": 1.0321,
+      "step": 3643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8496894669705775e-06,
+      "loss": 0.0182,
+      "step": 3644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8496476443403525e-06,
+      "loss": 0.0047,
+      "step": 3645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8496058217101273e-06,
+      "loss": 0.348,
+      "step": 3646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849563999079902e-06,
+      "loss": 2.5862,
+      "step": 3647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8495221764496769e-06,
+      "loss": 1.6309,
+      "step": 3648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8494803538194517e-06,
+      "loss": 0.0075,
+      "step": 3649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8494385311892265e-06,
+      "loss": 0.0034,
+      "step": 3650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8493967085590012e-06,
+      "loss": 0.0043,
+      "step": 3651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849354885928776e-06,
+      "loss": 0.1657,
+      "step": 3652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8493130632985508e-06,
+      "loss": 1.5377,
+      "step": 3653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8492712406683256e-06,
+      "loss": 0.0019,
+      "step": 3654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8492294180381004e-06,
+      "loss": 0.004,
+      "step": 3655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849187595407875e-06,
+      "loss": 0.0039,
+      "step": 3656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8491457727776498e-06,
+      "loss": 0.0021,
+      "step": 3657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8491039501474245e-06,
+      "loss": 0.0383,
+      "step": 3658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8490621275171993e-06,
+      "loss": 0.0023,
+      "step": 3659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8490203048869743e-06,
+      "loss": 0.1491,
+      "step": 3660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8489784822567491e-06,
+      "loss": 0.0026,
+      "step": 3661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848936659626524e-06,
+      "loss": 0.8926,
+      "step": 3662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8488948369962987e-06,
+      "loss": 0.0023,
+      "step": 3663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8488530143660735e-06,
+      "loss": 0.0021,
+      "step": 3664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8488111917358483e-06,
+      "loss": 0.0018,
+      "step": 3665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848769369105623e-06,
+      "loss": 0.0017,
+      "step": 3666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8487275464753978e-06,
+      "loss": 0.0021,
+      "step": 3667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8486857238451726e-06,
+      "loss": 0.0016,
+      "step": 3668
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8486439012149472e-06,
+      "loss": 0.0016,
+      "step": 3669
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848602078584722e-06,
+      "loss": 0.682,
+      "step": 3670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8485602559544968e-06,
+      "loss": 0.0016,
+      "step": 3671
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8485184333242715e-06,
+      "loss": 0.0019,
+      "step": 3672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8484766106940463e-06,
+      "loss": 0.0087,
+      "step": 3673
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8484347880638213e-06,
+      "loss": 0.0211,
+      "step": 3674
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8483929654335961e-06,
+      "loss": 0.0491,
+      "step": 3675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848351142803371e-06,
+      "loss": 0.002,
+      "step": 3676
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8483093201731457e-06,
+      "loss": 0.0333,
+      "step": 3677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8482674975429205e-06,
+      "loss": 0.306,
+      "step": 3678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8482256749126953e-06,
+      "loss": 0.4974,
+      "step": 3679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84818385228247e-06,
+      "loss": 0.0559,
+      "step": 3680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8481420296522448e-06,
+      "loss": 0.1361,
+      "step": 3681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8481002070220196e-06,
+      "loss": 0.0021,
+      "step": 3682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8480583843917942e-06,
+      "loss": 0.0025,
+      "step": 3683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848016561761569e-06,
+      "loss": 0.0016,
+      "step": 3684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8479747391313438e-06,
+      "loss": 0.4163,
+      "step": 3685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8479329165011186e-06,
+      "loss": 2.9706,
+      "step": 3686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8478910938708933e-06,
+      "loss": 0.0016,
+      "step": 3687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8478492712406681e-06,
+      "loss": 0.3132,
+      "step": 3688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8478074486104431e-06,
+      "loss": 0.0172,
+      "step": 3689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847765625980218e-06,
+      "loss": 0.002,
+      "step": 3690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8477238033499927e-06,
+      "loss": 0.0023,
+      "step": 3691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8476819807197675e-06,
+      "loss": 0.0017,
+      "step": 3692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8476401580895423e-06,
+      "loss": 0.0061,
+      "step": 3693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847598335459317e-06,
+      "loss": 0.1414,
+      "step": 3694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8475565128290918e-06,
+      "loss": 0.0073,
+      "step": 3695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8475146901988664e-06,
+      "loss": 0.002,
+      "step": 3696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8474728675686412e-06,
+      "loss": 0.0021,
+      "step": 3697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847431044938416e-06,
+      "loss": 3.4504,
+      "step": 3698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8473892223081908e-06,
+      "loss": 0.0236,
+      "step": 3699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8473473996779656e-06,
+      "loss": 0.0321,
+      "step": 3700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8473055770477404e-06,
+      "loss": 0.0092,
+      "step": 3701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8472637544175151e-06,
+      "loss": 2.9326,
+      "step": 3702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84722193178729e-06,
+      "loss": 0.0018,
+      "step": 3703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847180109157065e-06,
+      "loss": 0.0068,
+      "step": 3704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8471382865268397e-06,
+      "loss": 0.003,
+      "step": 3705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8470964638966145e-06,
+      "loss": 0.0022,
+      "step": 3706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8470546412663893e-06,
+      "loss": 0.1568,
+      "step": 3707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847012818636164e-06,
+      "loss": 0.0344,
+      "step": 3708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8469709960059389e-06,
+      "loss": 0.0046,
+      "step": 3709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8469291733757134e-06,
+      "loss": 0.0026,
+      "step": 3710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8468873507454882e-06,
+      "loss": 0.0038,
+      "step": 3711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846845528115263e-06,
+      "loss": 0.0017,
+      "step": 3712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8468037054850378e-06,
+      "loss": 0.0015,
+      "step": 3713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8467618828548126e-06,
+      "loss": 0.0054,
+      "step": 3714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8467200602245874e-06,
+      "loss": 0.0015,
+      "step": 3715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8466782375943621e-06,
+      "loss": 0.0013,
+      "step": 3716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846636414964137e-06,
+      "loss": 0.0315,
+      "step": 3717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846594592333912e-06,
+      "loss": 0.0023,
+      "step": 3718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8465527697036867e-06,
+      "loss": 0.0104,
+      "step": 3719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8465109470734615e-06,
+      "loss": 2.4892,
+      "step": 3720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8464691244432363e-06,
+      "loss": 0.0014,
+      "step": 3721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846427301813011e-06,
+      "loss": 0.0022,
+      "step": 3722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8463854791827857e-06,
+      "loss": 0.2383,
+      "step": 3723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8463436565525604e-06,
+      "loss": 0.0025,
+      "step": 3724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8463018339223352e-06,
+      "loss": 0.0014,
+      "step": 3725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84626001129211e-06,
+      "loss": 0.0014,
+      "step": 3726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8462181886618848e-06,
+      "loss": 0.002,
+      "step": 3727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8461763660316596e-06,
+      "loss": 0.0017,
+      "step": 3728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8461345434014344e-06,
+      "loss": 0.0041,
+      "step": 3729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8460927207712092e-06,
+      "loss": 0.0018,
+      "step": 3730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846050898140984e-06,
+      "loss": 1.2098,
+      "step": 3731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8460090755107587e-06,
+      "loss": 0.013,
+      "step": 3732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8459672528805337e-06,
+      "loss": 0.0066,
+      "step": 3733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8459254302503085e-06,
+      "loss": 1.3828,
+      "step": 3734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8458836076200833e-06,
+      "loss": 2.0182,
+      "step": 3735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845841784989858e-06,
+      "loss": 0.0019,
+      "step": 3736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8457999623596327e-06,
+      "loss": 0.1068,
+      "step": 3737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8457581397294074e-06,
+      "loss": 0.0015,
+      "step": 3738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8457163170991822e-06,
+      "loss": 0.0018,
+      "step": 3739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845674494468957e-06,
+      "loss": 2.0729,
+      "step": 3740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8456326718387318e-06,
+      "loss": 0.0021,
+      "step": 3741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8455908492085066e-06,
+      "loss": 0.6512,
+      "step": 3742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8455490265782814e-06,
+      "loss": 0.0016,
+      "step": 3743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8455072039480562e-06,
+      "loss": 0.0017,
+      "step": 3744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845465381317831e-06,
+      "loss": 0.3418,
+      "step": 3745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8454235586876057e-06,
+      "loss": 0.0127,
+      "step": 3746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8453817360573805e-06,
+      "loss": 0.0327,
+      "step": 3747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8453399134271555e-06,
+      "loss": 0.0139,
+      "step": 3748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8452980907969303e-06,
+      "loss": 1.5911,
+      "step": 3749
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.845256268166705e-06,
+      "loss": 0.0057,
+      "step": 3750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8452144455364797e-06,
+      "loss": 0.0012,
+      "step": 3751
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8451726229062545e-06,
+      "loss": 0.0033,
+      "step": 3752
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8451308002760292e-06,
+      "loss": 0.0036,
+      "step": 3753
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.845088977645804e-06,
+      "loss": 1.7043,
+      "step": 3754
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8450471550155788e-06,
+      "loss": 0.0018,
+      "step": 3755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8450053323853536e-06,
+      "loss": 0.0026,
+      "step": 3756
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8449635097551284e-06,
+      "loss": 2.9204,
+      "step": 3757
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8449216871249032e-06,
+      "loss": 2.4751,
+      "step": 3758
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.844879864494678e-06,
+      "loss": 0.0028,
+      "step": 3759
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8448380418644527e-06,
+      "loss": 0.7173,
+      "step": 3760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8447962192342275e-06,
+      "loss": 0.3778,
+      "step": 3761
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8447543966040025e-06,
+      "loss": 0.5393,
+      "step": 3762
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8447125739737773e-06,
+      "loss": 0.0012,
+      "step": 3763
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8446707513435519e-06,
+      "loss": 0.0016,
+      "step": 3764
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8446289287133267e-06,
+      "loss": 2.1839,
+      "step": 3765
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8445871060831015e-06,
+      "loss": 0.0023,
+      "step": 3766
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8445452834528762e-06,
+      "loss": 0.0038,
+      "step": 3767
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.844503460822651e-06,
+      "loss": 0.0019,
+      "step": 3768
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8444616381924258e-06,
+      "loss": 0.003,
+      "step": 3769
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8444198155622006e-06,
+      "loss": 0.0016,
+      "step": 3770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8443779929319754e-06,
+      "loss": 0.002,
+      "step": 3771
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8443361703017502e-06,
+      "loss": 0.2628,
+      "step": 3772
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.844294347671525e-06,
+      "loss": 0.0019,
+      "step": 3773
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8442525250412998e-06,
+      "loss": 0.1615,
+      "step": 3774
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8442107024110745e-06,
+      "loss": 0.0104,
+      "step": 3775
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8441688797808493e-06,
+      "loss": 0.1566,
+      "step": 3776
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8441270571506243e-06,
+      "loss": 0.0016,
+      "step": 3777
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.844085234520399e-06,
+      "loss": 0.2209,
+      "step": 3778
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8440434118901737e-06,
+      "loss": 1.6008,
+      "step": 3779
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8440015892599485e-06,
+      "loss": 0.0056,
+      "step": 3780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8439597666297233e-06,
+      "loss": 0.0135,
+      "step": 3781
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843917943999498e-06,
+      "loss": 3.2239,
+      "step": 3782
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8438761213692728e-06,
+      "loss": 0.0676,
+      "step": 3783
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8438342987390476e-06,
+      "loss": 0.0049,
+      "step": 3784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8437924761088224e-06,
+      "loss": 0.0038,
+      "step": 3785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8437506534785972e-06,
+      "loss": 0.0017,
+      "step": 3786
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843708830848372e-06,
+      "loss": 0.0021,
+      "step": 3787
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8436670082181468e-06,
+      "loss": 0.2902,
+      "step": 3788
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8436251855879215e-06,
+      "loss": 0.0019,
+      "step": 3789
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8435833629576963e-06,
+      "loss": 0.004,
+      "step": 3790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843541540327471e-06,
+      "loss": 0.0035,
+      "step": 3791
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843499717697246e-06,
+      "loss": 0.0018,
+      "step": 3792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8434578950670207e-06,
+      "loss": 0.0076,
+      "step": 3793
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8434160724367955e-06,
+      "loss": 0.002,
+      "step": 3794
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8433742498065703e-06,
+      "loss": 0.0037,
+      "step": 3795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843332427176345e-06,
+      "loss": 0.0016,
+      "step": 3796
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8432906045461198e-06,
+      "loss": 0.0014,
+      "step": 3797
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8432487819158946e-06,
+      "loss": 0.0027,
+      "step": 3798
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8432069592856694e-06,
+      "loss": 0.0016,
+      "step": 3799
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8431651366554442e-06,
+      "loss": 0.0018,
+      "step": 3800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843123314025219e-06,
+      "loss": 1.9787,
+      "step": 3801
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8430814913949938e-06,
+      "loss": 0.4043,
+      "step": 3802
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8430396687647686e-06,
+      "loss": 0.014,
+      "step": 3803
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8429978461345431e-06,
+      "loss": 0.002,
+      "step": 3804
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842956023504318e-06,
+      "loss": 0.0019,
+      "step": 3805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842914200874093e-06,
+      "loss": 0.0096,
+      "step": 3806
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8428723782438677e-06,
+      "loss": 0.0087,
+      "step": 3807
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8428305556136425e-06,
+      "loss": 0.0046,
+      "step": 3808
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8427887329834173e-06,
+      "loss": 0.0016,
+      "step": 3809
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842746910353192e-06,
+      "loss": 0.0067,
+      "step": 3810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8427050877229668e-06,
+      "loss": 0.0569,
+      "step": 3811
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8426632650927416e-06,
+      "loss": 0.0025,
+      "step": 3812
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8426214424625164e-06,
+      "loss": 0.0016,
+      "step": 3813
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8425796198322912e-06,
+      "loss": 0.0017,
+      "step": 3814
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842537797202066e-06,
+      "loss": 0.7093,
+      "step": 3815
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8424959745718408e-06,
+      "loss": 0.0028,
+      "step": 3816
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8424541519416156e-06,
+      "loss": 0.0018,
+      "step": 3817
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8424123293113901e-06,
+      "loss": 1.2693,
+      "step": 3818
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842370506681165e-06,
+      "loss": 2.7319,
+      "step": 3819
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8423286840509397e-06,
+      "loss": 0.0026,
+      "step": 3820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8422868614207147e-06,
+      "loss": 1.324,
+      "step": 3821
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8422450387904895e-06,
+      "loss": 0.0016,
+      "step": 3822
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8422032161602643e-06,
+      "loss": 0.0046,
+      "step": 3823
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842161393530039e-06,
+      "loss": 0.0019,
+      "step": 3824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8421195708998139e-06,
+      "loss": 3.7366,
+      "step": 3825
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8420777482695886e-06,
+      "loss": 0.0014,
+      "step": 3826
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8420359256393634e-06,
+      "loss": 0.0046,
+      "step": 3827
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8419941030091382e-06,
+      "loss": 0.0034,
+      "step": 3828
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841952280378913e-06,
+      "loss": 2.1737,
+      "step": 3829
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8419104577486878e-06,
+      "loss": 0.0033,
+      "step": 3830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8418686351184624e-06,
+      "loss": 0.0019,
+      "step": 3831
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8418268124882371e-06,
+      "loss": 0.0013,
+      "step": 3832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841784989858012e-06,
+      "loss": 0.6522,
+      "step": 3833
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8417431672277867e-06,
+      "loss": 0.0013,
+      "step": 3834
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8417013445975617e-06,
+      "loss": 0.0019,
+      "step": 3835
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8416595219673365e-06,
+      "loss": 0.0018,
+      "step": 3836
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8416176993371113e-06,
+      "loss": 0.0013,
+      "step": 3837
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841575876706886e-06,
+      "loss": 0.0052,
+      "step": 3838
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8415340540766609e-06,
+      "loss": 0.0015,
+      "step": 3839
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8414922314464357e-06,
+      "loss": 0.0034,
+      "step": 3840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8414504088162104e-06,
+      "loss": 0.0017,
+      "step": 3841
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8414085861859852e-06,
+      "loss": 2.219,
+      "step": 3842
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.84136676355576e-06,
+      "loss": 0.0018,
+      "step": 3843
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8413249409255348e-06,
+      "loss": 0.0013,
+      "step": 3844
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8412831182953094e-06,
+      "loss": 0.0017,
+      "step": 3845
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8412412956650842e-06,
+      "loss": 0.643,
+      "step": 3846
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841199473034859e-06,
+      "loss": 0.0023,
+      "step": 3847
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8411576504046337e-06,
+      "loss": 3.6424,
+      "step": 3848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8411158277744085e-06,
+      "loss": 0.766,
+      "step": 3849
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8410740051441835e-06,
+      "loss": 0.0015,
+      "step": 3850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8410321825139583e-06,
+      "loss": 2.2314,
+      "step": 3851
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840990359883733e-06,
+      "loss": 0.0041,
+      "step": 3852
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8409485372535079e-06,
+      "loss": 0.0019,
+      "step": 3853
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8409067146232827e-06,
+      "loss": 1.4347,
+      "step": 3854
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8408648919930574e-06,
+      "loss": 0.0031,
+      "step": 3855
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8408230693628322e-06,
+      "loss": 0.0024,
+      "step": 3856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840781246732607e-06,
+      "loss": 0.0033,
+      "step": 3857
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8407394241023818e-06,
+      "loss": 1.3037,
+      "step": 3858
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8406976014721564e-06,
+      "loss": 0.0031,
+      "step": 3859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8406557788419312e-06,
+      "loss": 0.0019,
+      "step": 3860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840613956211706e-06,
+      "loss": 0.4699,
+      "step": 3861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8405721335814807e-06,
+      "loss": 0.0138,
+      "step": 3862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8405303109512555e-06,
+      "loss": 0.0143,
+      "step": 3863
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8404884883210303e-06,
+      "loss": 0.0042,
+      "step": 3864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8404466656908053e-06,
+      "loss": 0.8164,
+      "step": 3865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.84040484306058e-06,
+      "loss": 0.0016,
+      "step": 3866
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8403630204303549e-06,
+      "loss": 0.2957,
+      "step": 3867
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8403211978001297e-06,
+      "loss": 1.2114,
+      "step": 3868
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8402793751699045e-06,
+      "loss": 2.0009,
+      "step": 3869
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8402375525396792e-06,
+      "loss": 1.8285,
+      "step": 3870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840195729909454e-06,
+      "loss": 0.004,
+      "step": 3871
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8401539072792286e-06,
+      "loss": 0.6175,
+      "step": 3872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8401120846490034e-06,
+      "loss": 0.2679,
+      "step": 3873
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8400702620187782e-06,
+      "loss": 1.4846,
+      "step": 3874
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840028439388553e-06,
+      "loss": 0.0832,
+      "step": 3875
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8399866167583277e-06,
+      "loss": 0.0039,
+      "step": 3876
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8399447941281025e-06,
+      "loss": 0.1037,
+      "step": 3877
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8399029714978773e-06,
+      "loss": 0.0258,
+      "step": 3878
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8398611488676523e-06,
+      "loss": 0.0079,
+      "step": 3879
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839819326237427e-06,
+      "loss": 0.0295,
+      "step": 3880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8397775036072019e-06,
+      "loss": 0.1054,
+      "step": 3881
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8397356809769767e-06,
+      "loss": 0.1226,
+      "step": 3882
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8396938583467515e-06,
+      "loss": 0.2748,
+      "step": 3883
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8396520357165262e-06,
+      "loss": 0.0017,
+      "step": 3884
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839610213086301e-06,
+      "loss": 0.0058,
+      "step": 3885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8395683904560756e-06,
+      "loss": 0.3823,
+      "step": 3886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8395265678258504e-06,
+      "loss": 0.005,
+      "step": 3887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8394847451956252e-06,
+      "loss": 0.2811,
+      "step": 3888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8394429225654e-06,
+      "loss": 0.0597,
+      "step": 3889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8394010999351748e-06,
+      "loss": 0.0038,
+      "step": 3890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8393592773049495e-06,
+      "loss": 0.002,
+      "step": 3891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8393174546747243e-06,
+      "loss": 1.4638,
+      "step": 3892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8392756320444991e-06,
+      "loss": 0.0036,
+      "step": 3893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8392338094142741e-06,
+      "loss": 1.0812,
+      "step": 3894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839191986784049e-06,
+      "loss": 0.2007,
+      "step": 3895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8391501641538237e-06,
+      "loss": 0.0018,
+      "step": 3896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8391083415235985e-06,
+      "loss": 0.0018,
+      "step": 3897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8390665188933733e-06,
+      "loss": 0.2532,
+      "step": 3898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8390246962631478e-06,
+      "loss": 0.0035,
+      "step": 3899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8389828736329226e-06,
+      "loss": 0.1111,
+      "step": 3900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8389410510026974e-06,
+      "loss": 1.6275,
+      "step": 3901
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8388992283724722e-06,
+      "loss": 0.0032,
+      "step": 3902
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838857405742247e-06,
+      "loss": 0.0021,
+      "step": 3903
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8388155831120218e-06,
+      "loss": 0.039,
+      "step": 3904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8387737604817966e-06,
+      "loss": 0.6757,
+      "step": 3905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8387319378515713e-06,
+      "loss": 0.0027,
+      "step": 3906
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8386901152213461e-06,
+      "loss": 0.0015,
+      "step": 3907
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838648292591121e-06,
+      "loss": 0.0018,
+      "step": 3908
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838606469960896e-06,
+      "loss": 0.1666,
+      "step": 3909
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8385646473306707e-06,
+      "loss": 0.0019,
+      "step": 3910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8385228247004455e-06,
+      "loss": 3.5303,
+      "step": 3911
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8384810020702203e-06,
+      "loss": 0.0017,
+      "step": 3912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8384391794399948e-06,
+      "loss": 0.0032,
+      "step": 3913
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8383973568097696e-06,
+      "loss": 0.0016,
+      "step": 3914
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8383555341795444e-06,
+      "loss": 0.0019,
+      "step": 3915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8383137115493192e-06,
+      "loss": 0.007,
+      "step": 3916
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.838271888919094e-06,
+      "loss": 0.0014,
+      "step": 3917
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8382300662888688e-06,
+      "loss": 0.0018,
+      "step": 3918
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8381882436586436e-06,
+      "loss": 0.0019,
+      "step": 3919
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8381464210284183e-06,
+      "loss": 0.0014,
+      "step": 3920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8381045983981931e-06,
+      "loss": 0.0238,
+      "step": 3921
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.838062775767968e-06,
+      "loss": 0.0556,
+      "step": 3922
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.838020953137743e-06,
+      "loss": 0.0013,
+      "step": 3923
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8379791305075177e-06,
+      "loss": 0.0021,
+      "step": 3924
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8379373078772925e-06,
+      "loss": 3.4951,
+      "step": 3925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.837895485247067e-06,
+      "loss": 0.0014,
+      "step": 3926
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8378536626168418e-06,
+      "loss": 0.015,
+      "step": 3927
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8378118399866166e-06,
+      "loss": 0.0012,
+      "step": 3928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8377700173563914e-06,
+      "loss": 0.0014,
+      "step": 3929
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8377281947261662e-06,
+      "loss": 0.0034,
+      "step": 3930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.837686372095941e-06,
+      "loss": 1.957,
+      "step": 3931
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8376445494657158e-06,
+      "loss": 0.2427,
+      "step": 3932
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8376027268354906e-06,
+      "loss": 0.0018,
+      "step": 3933
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8375609042052654e-06,
+      "loss": 0.0031,
+      "step": 3934
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8375190815750401e-06,
+      "loss": 0.0019,
+      "step": 3935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.837477258944815e-06,
+      "loss": 0.7586,
+      "step": 3936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8374354363145897e-06,
+      "loss": 0.0032,
+      "step": 3937
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8373936136843647e-06,
+      "loss": 0.0017,
+      "step": 3938
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8373517910541395e-06,
+      "loss": 0.0011,
+      "step": 3939
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.837309968423914e-06,
+      "loss": 0.003,
+      "step": 3940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8372681457936889e-06,
+      "loss": 0.0013,
+      "step": 3941
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8372263231634636e-06,
+      "loss": 0.0025,
+      "step": 3942
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8371845005332384e-06,
+      "loss": 0.0023,
+      "step": 3943
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8371426779030132e-06,
+      "loss": 2.2309,
+      "step": 3944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.837100855272788e-06,
+      "loss": 0.7813,
+      "step": 3945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8370590326425628e-06,
+      "loss": 0.0051,
+      "step": 3946
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8370172100123376e-06,
+      "loss": 0.0014,
+      "step": 3947
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8369753873821124e-06,
+      "loss": 0.0195,
+      "step": 3948
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8369335647518871e-06,
+      "loss": 0.0021,
+      "step": 3949
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836891742121662e-06,
+      "loss": 0.0013,
+      "step": 3950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8368499194914367e-06,
+      "loss": 0.0155,
+      "step": 3951
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8368080968612115e-06,
+      "loss": 0.0019,
+      "step": 3952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8367662742309865e-06,
+      "loss": 0.0024,
+      "step": 3953
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836724451600761e-06,
+      "loss": 0.0018,
+      "step": 3954
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8366826289705359e-06,
+      "loss": 2.1416,
+      "step": 3955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8366408063403107e-06,
+      "loss": 0.0193,
+      "step": 3956
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8365989837100854e-06,
+      "loss": 0.002,
+      "step": 3957
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8365571610798602e-06,
+      "loss": 0.0025,
+      "step": 3958
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836515338449635e-06,
+      "loss": 0.0151,
+      "step": 3959
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8364735158194098e-06,
+      "loss": 0.0018,
+      "step": 3960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8364316931891846e-06,
+      "loss": 0.0012,
+      "step": 3961
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8363898705589594e-06,
+      "loss": 0.0531,
+      "step": 3962
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8363480479287342e-06,
+      "loss": 0.0016,
+      "step": 3963
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836306225298509e-06,
+      "loss": 0.0024,
+      "step": 3964
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8362644026682837e-06,
+      "loss": 0.0029,
+      "step": 3965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8362225800380585e-06,
+      "loss": 0.0014,
+      "step": 3966
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8361807574078333e-06,
+      "loss": 0.0422,
+      "step": 3967
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836138934777608e-06,
+      "loss": 0.0014,
+      "step": 3968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8360971121473829e-06,
+      "loss": 0.8577,
+      "step": 3969
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8360552895171577e-06,
+      "loss": 0.0021,
+      "step": 3970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8360134668869324e-06,
+      "loss": 0.0012,
+      "step": 3971
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8359716442567072e-06,
+      "loss": 0.0011,
+      "step": 3972
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835929821626482e-06,
+      "loss": 0.0088,
+      "step": 3973
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8358879989962568e-06,
+      "loss": 0.0014,
+      "step": 3974
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8358461763660316e-06,
+      "loss": 0.0011,
+      "step": 3975
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8358043537358064e-06,
+      "loss": 0.0015,
+      "step": 3976
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8357625311055812e-06,
+      "loss": 0.0012,
+      "step": 3977
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835720708475356e-06,
+      "loss": 1.7664,
+      "step": 3978
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8356788858451307e-06,
+      "loss": 0.0034,
+      "step": 3979
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8356370632149053e-06,
+      "loss": 0.0043,
+      "step": 3980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83559524058468e-06,
+      "loss": 0.0331,
+      "step": 3981
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835553417954455e-06,
+      "loss": 0.001,
+      "step": 3982
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8355115953242299e-06,
+      "loss": 0.0013,
+      "step": 3983
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8354697726940047e-06,
+      "loss": 0.0053,
+      "step": 3984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8354279500637795e-06,
+      "loss": 0.0043,
+      "step": 3985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8353861274335542e-06,
+      "loss": 0.0015,
+      "step": 3986
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835344304803329e-06,
+      "loss": 0.1999,
+      "step": 3987
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8353024821731038e-06,
+      "loss": 0.0838,
+      "step": 3988
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8352606595428786e-06,
+      "loss": 1.6413,
+      "step": 3989
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8352188369126534e-06,
+      "loss": 0.009,
+      "step": 3990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8351770142824282e-06,
+      "loss": 2.7432,
+      "step": 3991
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835135191652203e-06,
+      "loss": 0.001,
+      "step": 3992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8350933690219777e-06,
+      "loss": 0.0516,
+      "step": 3993
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8350515463917523e-06,
+      "loss": 0.0035,
+      "step": 3994
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8350097237615271e-06,
+      "loss": 0.0013,
+      "step": 3995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834967901131302e-06,
+      "loss": 0.0014,
+      "step": 3996
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8349260785010769e-06,
+      "loss": 0.0036,
+      "step": 3997
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8348842558708517e-06,
+      "loss": 0.001,
+      "step": 3998
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8348424332406265e-06,
+      "loss": 0.6982,
+      "step": 3999
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8348006106104013e-06,
+      "loss": 0.0014,
+      "step": 4000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834758787980176e-06,
+      "loss": 0.0033,
+      "step": 4001
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8347169653499508e-06,
+      "loss": 0.067,
+      "step": 4002
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8346751427197256e-06,
+      "loss": 0.557,
+      "step": 4003
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8346333200895004e-06,
+      "loss": 1.3324,
+      "step": 4004
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8345914974592752e-06,
+      "loss": 0.0011,
+      "step": 4005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83454967482905e-06,
+      "loss": 0.0072,
+      "step": 4006
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8345078521988245e-06,
+      "loss": 0.0011,
+      "step": 4007
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8344660295685993e-06,
+      "loss": 0.001,
+      "step": 4008
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8344242069383741e-06,
+      "loss": 0.001,
+      "step": 4009
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834382384308149e-06,
+      "loss": 0.0019,
+      "step": 4010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834340561677924e-06,
+      "loss": 0.5994,
+      "step": 4011
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8342987390476987e-06,
+      "loss": 0.6809,
+      "step": 4012
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8342569164174735e-06,
+      "loss": 0.0027,
+      "step": 4013
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8342150937872483e-06,
+      "loss": 0.0017,
+      "step": 4014
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834173271157023e-06,
+      "loss": 0.0295,
+      "step": 4015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8341314485267978e-06,
+      "loss": 0.0028,
+      "step": 4016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8340896258965726e-06,
+      "loss": 0.0015,
+      "step": 4017
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8340478032663474e-06,
+      "loss": 0.0211,
+      "step": 4018
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8340059806361222e-06,
+      "loss": 1.3541,
+      "step": 4019
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833964158005897e-06,
+      "loss": 1.3534,
+      "step": 4020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8339223353756716e-06,
+      "loss": 0.0031,
+      "step": 4021
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8338805127454463e-06,
+      "loss": 0.0032,
+      "step": 4022
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8338386901152211e-06,
+      "loss": 0.0018,
+      "step": 4023
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833796867484996e-06,
+      "loss": 0.4614,
+      "step": 4024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8337550448547707e-06,
+      "loss": 0.0027,
+      "step": 4025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8337132222245457e-06,
+      "loss": 0.2312,
+      "step": 4026
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8336713995943205e-06,
+      "loss": 3.6568,
+      "step": 4027
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8336295769640953e-06,
+      "loss": 0.0024,
+      "step": 4028
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83358775433387e-06,
+      "loss": 0.0016,
+      "step": 4029
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8335459317036448e-06,
+      "loss": 0.0013,
+      "step": 4030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8335041090734196e-06,
+      "loss": 0.4948,
+      "step": 4031
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8334622864431944e-06,
+      "loss": 0.0013,
+      "step": 4032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8334204638129692e-06,
+      "loss": 0.0021,
+      "step": 4033
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8333786411827438e-06,
+      "loss": 0.2383,
+      "step": 4034
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8333368185525186e-06,
+      "loss": 0.0018,
+      "step": 4035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8332949959222933e-06,
+      "loss": 0.1923,
+      "step": 4036
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8332531732920681e-06,
+      "loss": 0.0018,
+      "step": 4037
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833211350661843e-06,
+      "loss": 0.0061,
+      "step": 4038
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8331695280316177e-06,
+      "loss": 0.0011,
+      "step": 4039
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8331277054013927e-06,
+      "loss": 0.0355,
+      "step": 4040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8330858827711675e-06,
+      "loss": 0.0256,
+      "step": 4041
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8330440601409423e-06,
+      "loss": 0.0011,
+      "step": 4042
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833002237510717e-06,
+      "loss": 0.001,
+      "step": 4043
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8329604148804919e-06,
+      "loss": 0.0016,
+      "step": 4044
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8329185922502666e-06,
+      "loss": 0.0022,
+      "step": 4045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8328767696200414e-06,
+      "loss": 0.0018,
+      "step": 4046
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8328349469898162e-06,
+      "loss": 0.0022,
+      "step": 4047
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8327931243595908e-06,
+      "loss": 0.0047,
+      "step": 4048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8327513017293656e-06,
+      "loss": 0.6503,
+      "step": 4049
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8327094790991404e-06,
+      "loss": 0.2426,
+      "step": 4050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8326676564689151e-06,
+      "loss": 0.0024,
+      "step": 4051
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83262583383869e-06,
+      "loss": 0.0012,
+      "step": 4052
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8325840112084647e-06,
+      "loss": 0.0014,
+      "step": 4053
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8325421885782395e-06,
+      "loss": 0.701,
+      "step": 4054
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8325003659480145e-06,
+      "loss": 0.0013,
+      "step": 4055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8324585433177893e-06,
+      "loss": 0.1924,
+      "step": 4056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832416720687564e-06,
+      "loss": 0.0017,
+      "step": 4057
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8323748980573389e-06,
+      "loss": 0.0999,
+      "step": 4058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8323330754271136e-06,
+      "loss": 0.0023,
+      "step": 4059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8322912527968884e-06,
+      "loss": 0.0012,
+      "step": 4060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8322494301666632e-06,
+      "loss": 0.0016,
+      "step": 4061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8322076075364378e-06,
+      "loss": 0.0009,
+      "step": 4062
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8321657849062126e-06,
+      "loss": 0.0021,
+      "step": 4063
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8321239622759874e-06,
+      "loss": 0.0014,
+      "step": 4064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8320821396457622e-06,
+      "loss": 0.0133,
+      "step": 4065
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832040317015537e-06,
+      "loss": 1.5014,
+      "step": 4066
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8319984943853117e-06,
+      "loss": 0.7416,
+      "step": 4067
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8319566717550865e-06,
+      "loss": 3.9272,
+      "step": 4068
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8319148491248613e-06,
+      "loss": 3.6253,
+      "step": 4069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8318730264946363e-06,
+      "loss": 1.8415,
+      "step": 4070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.831831203864411e-06,
+      "loss": 0.0035,
+      "step": 4071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8317893812341859e-06,
+      "loss": 0.0011,
+      "step": 4072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8317475586039607e-06,
+      "loss": 0.0016,
+      "step": 4073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8317057359737354e-06,
+      "loss": 0.001,
+      "step": 4074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.83166391334351e-06,
+      "loss": 2.6869,
+      "step": 4075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8316220907132848e-06,
+      "loss": 0.0031,
+      "step": 4076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8315802680830596e-06,
+      "loss": 0.0196,
+      "step": 4077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8315384454528344e-06,
+      "loss": 0.0045,
+      "step": 4078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8314966228226092e-06,
+      "loss": 0.8103,
+      "step": 4079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.831454800192384e-06,
+      "loss": 0.0022,
+      "step": 4080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8314129775621587e-06,
+      "loss": 0.0016,
+      "step": 4081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8313711549319335e-06,
+      "loss": 0.0017,
+      "step": 4082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8313293323017083e-06,
+      "loss": 0.0011,
+      "step": 4083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8312875096714833e-06,
+      "loss": 0.7088,
+      "step": 4084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.831245687041258e-06,
+      "loss": 0.0025,
+      "step": 4085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8312038644110329e-06,
+      "loss": 0.0865,
+      "step": 4086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8311620417808077e-06,
+      "loss": 0.0464,
+      "step": 4087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8311202191505824e-06,
+      "loss": 0.0029,
+      "step": 4088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.831078396520357e-06,
+      "loss": 0.1468,
+      "step": 4089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8310365738901318e-06,
+      "loss": 3.3282,
+      "step": 4090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8309947512599066e-06,
+      "loss": 0.0019,
+      "step": 4091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8309529286296814e-06,
+      "loss": 0.0025,
+      "step": 4092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8309111059994562e-06,
+      "loss": 0.146,
+      "step": 4093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.830869283369231e-06,
+      "loss": 0.001,
+      "step": 4094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8308274607390057e-06,
+      "loss": 2.2031,
+      "step": 4095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8307856381087805e-06,
+      "loss": 0.0014,
+      "step": 4096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8307438154785553e-06,
+      "loss": 0.0019,
+      "step": 4097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.83070199284833e-06,
+      "loss": 2.6609,
+      "step": 4098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.830660170218105e-06,
+      "loss": 0.0012,
+      "step": 4099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8306183475878799e-06,
+      "loss": 0.0137,
+      "step": 4100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8305765249576547e-06,
+      "loss": 0.4287,
+      "step": 4101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8305347023274292e-06,
+      "loss": 1.613,
+      "step": 4102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.830492879697204e-06,
+      "loss": 1.8952,
+      "step": 4103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8304510570669788e-06,
+      "loss": 0.4364,
+      "step": 4104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8304092344367536e-06,
+      "loss": 0.0023,
+      "step": 4105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8303674118065284e-06,
+      "loss": 0.1479,
+      "step": 4106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8303255891763032e-06,
+      "loss": 0.0014,
+      "step": 4107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.830283766546078e-06,
+      "loss": 0.0081,
+      "step": 4108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8302419439158527e-06,
+      "loss": 0.2634,
+      "step": 4109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8302001212856275e-06,
+      "loss": 0.0013,
+      "step": 4110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8301582986554023e-06,
+      "loss": 0.0013,
+      "step": 4111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8301164760251771e-06,
+      "loss": 0.8594,
+      "step": 4112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.830074653394952e-06,
+      "loss": 0.1892,
+      "step": 4113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8300328307647269e-06,
+      "loss": 0.0132,
+      "step": 4114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8299910081345017e-06,
+      "loss": 0.0087,
+      "step": 4115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8299491855042763e-06,
+      "loss": 1.1837,
+      "step": 4116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829907362874051e-06,
+      "loss": 0.0014,
+      "step": 4117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8298655402438258e-06,
+      "loss": 0.0016,
+      "step": 4118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8298237176136006e-06,
+      "loss": 0.002,
+      "step": 4119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8297818949833754e-06,
+      "loss": 0.0033,
+      "step": 4120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8297400723531502e-06,
+      "loss": 1.518,
+      "step": 4121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829698249722925e-06,
+      "loss": 0.1601,
+      "step": 4122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8296564270926998e-06,
+      "loss": 0.0032,
+      "step": 4123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8296146044624745e-06,
+      "loss": 0.0026,
+      "step": 4124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8295727818322493e-06,
+      "loss": 0.0014,
+      "step": 4125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8295309592020241e-06,
+      "loss": 1.3122,
+      "step": 4126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829489136571799e-06,
+      "loss": 0.0018,
+      "step": 4127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829447313941574e-06,
+      "loss": 0.0024,
+      "step": 4128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8294054913113485e-06,
+      "loss": 0.0057,
+      "step": 4129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8293636686811233e-06,
+      "loss": 0.0014,
+      "step": 4130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829321846050898e-06,
+      "loss": 1.7801,
+      "step": 4131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8292800234206728e-06,
+      "loss": 0.4408,
+      "step": 4132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8292382007904476e-06,
+      "loss": 0.0058,
+      "step": 4133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8291963781602224e-06,
+      "loss": 0.7042,
+      "step": 4134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8291545555299972e-06,
+      "loss": 0.0013,
+      "step": 4135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829112732899772e-06,
+      "loss": 0.4264,
+      "step": 4136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8290709102695468e-06,
+      "loss": 0.2071,
+      "step": 4137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8290290876393216e-06,
+      "loss": 0.0034,
+      "step": 4138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8289872650090963e-06,
+      "loss": 0.7344,
+      "step": 4139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8289454423788711e-06,
+      "loss": 1.4234,
+      "step": 4140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828903619748646e-06,
+      "loss": 0.0014,
+      "step": 4141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8288617971184205e-06,
+      "loss": 0.0014,
+      "step": 4142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8288199744881955e-06,
+      "loss": 0.0068,
+      "step": 4143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8287781518579703e-06,
+      "loss": 0.0249,
+      "step": 4144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828736329227745e-06,
+      "loss": 0.002,
+      "step": 4145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8286945065975198e-06,
+      "loss": 0.0024,
+      "step": 4146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8286526839672946e-06,
+      "loss": 0.3143,
+      "step": 4147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8286108613370694e-06,
+      "loss": 0.0019,
+      "step": 4148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8285690387068442e-06,
+      "loss": 0.0053,
+      "step": 4149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828527216076619e-06,
+      "loss": 0.0026,
+      "step": 4150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8284853934463938e-06,
+      "loss": 0.0066,
+      "step": 4151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8284435708161686e-06,
+      "loss": 0.0339,
+      "step": 4152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8284017481859433e-06,
+      "loss": 0.0031,
+      "step": 4153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8283599255557181e-06,
+      "loss": 0.0031,
+      "step": 4154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828318102925493e-06,
+      "loss": 0.0018,
+      "step": 4155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8282762802952675e-06,
+      "loss": 0.005,
+      "step": 4156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8282344576650423e-06,
+      "loss": 0.0065,
+      "step": 4157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8281926350348173e-06,
+      "loss": 0.0017,
+      "step": 4158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828150812404592e-06,
+      "loss": 0.0011,
+      "step": 4159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8281089897743669e-06,
+      "loss": 0.0022,
+      "step": 4160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8280671671441416e-06,
+      "loss": 0.0013,
+      "step": 4161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8280253445139164e-06,
+      "loss": 0.0023,
+      "step": 4162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8279835218836912e-06,
+      "loss": 0.8334,
+      "step": 4163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827941699253466e-06,
+      "loss": 0.0055,
+      "step": 4164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8278998766232408e-06,
+      "loss": 0.0071,
+      "step": 4165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8278580539930156e-06,
+      "loss": 0.131,
+      "step": 4166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8278162313627904e-06,
+      "loss": 0.0012,
+      "step": 4167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8277744087325651e-06,
+      "loss": 0.1502,
+      "step": 4168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.82773258610234e-06,
+      "loss": 0.0018,
+      "step": 4169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8276907634721145e-06,
+      "loss": 0.1224,
+      "step": 4170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8276489408418893e-06,
+      "loss": 1.2875,
+      "step": 4171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8276071182116643e-06,
+      "loss": 0.0025,
+      "step": 4172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827565295581439e-06,
+      "loss": 0.0039,
+      "step": 4173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8275234729512139e-06,
+      "loss": 0.0013,
+      "step": 4174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8274816503209886e-06,
+      "loss": 0.0014,
+      "step": 4175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8274398276907634e-06,
+      "loss": 0.0015,
+      "step": 4176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8273980050605382e-06,
+      "loss": 0.0017,
+      "step": 4177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827356182430313e-06,
+      "loss": 0.0027,
+      "step": 4178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8273143598000878e-06,
+      "loss": 0.0012,
+      "step": 4179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8272725371698626e-06,
+      "loss": 0.267,
+      "step": 4180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8272307145396374e-06,
+      "loss": 0.0054,
+      "step": 4181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8271888919094122e-06,
+      "loss": 0.0046,
+      "step": 4182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8271470692791867e-06,
+      "loss": 0.003,
+      "step": 4183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8271052466489615e-06,
+      "loss": 2.8976,
+      "step": 4184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8270634240187363e-06,
+      "loss": 0.0013,
+      "step": 4185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827021601388511e-06,
+      "loss": 1.7615,
+      "step": 4186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826979778758286e-06,
+      "loss": 0.002,
+      "step": 4187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8269379561280609e-06,
+      "loss": 0.0011,
+      "step": 4188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8268961334978357e-06,
+      "loss": 0.0049,
+      "step": 4189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8268543108676104e-06,
+      "loss": 3.6238,
+      "step": 4190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8268124882373852e-06,
+      "loss": 0.0018,
+      "step": 4191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.82677066560716e-06,
+      "loss": 0.0033,
+      "step": 4192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8267288429769348e-06,
+      "loss": 0.0014,
+      "step": 4193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8266870203467096e-06,
+      "loss": 0.0013,
+      "step": 4194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8266451977164844e-06,
+      "loss": 0.0016,
+      "step": 4195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8266033750862592e-06,
+      "loss": 0.0012,
+      "step": 4196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8265615524560337e-06,
+      "loss": 2.1157,
+      "step": 4197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8265197298258085e-06,
+      "loss": 0.7744,
+      "step": 4198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8264779071955833e-06,
+      "loss": 0.003,
+      "step": 4199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826436084565358e-06,
+      "loss": 0.1319,
+      "step": 4200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8263942619351329e-06,
+      "loss": 0.0031,
+      "step": 4201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8263524393049079e-06,
+      "loss": 0.0017,
+      "step": 4202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8263106166746827e-06,
+      "loss": 0.0036,
+      "step": 4203
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8262687940444575e-06,
+      "loss": 0.0016,
+      "step": 4204
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8262269714142322e-06,
+      "loss": 0.005,
+      "step": 4205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826185148784007e-06,
+      "loss": 0.0013,
+      "step": 4206
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8261433261537818e-06,
+      "loss": 0.3467,
+      "step": 4207
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8261015035235566e-06,
+      "loss": 0.0014,
+      "step": 4208
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8260596808933314e-06,
+      "loss": 0.0071,
+      "step": 4209
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826017858263106e-06,
+      "loss": 0.0019,
+      "step": 4210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8259760356328807e-06,
+      "loss": 2.8573,
+      "step": 4211
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8259342130026555e-06,
+      "loss": 0.0085,
+      "step": 4212
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8258923903724303e-06,
+      "loss": 0.0053,
+      "step": 4213
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825850567742205e-06,
+      "loss": 3.3423,
+      "step": 4214
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8258087451119799e-06,
+      "loss": 0.0091,
+      "step": 4215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8257669224817549e-06,
+      "loss": 3.8166,
+      "step": 4216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8257250998515297e-06,
+      "loss": 0.0015,
+      "step": 4217
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8256832772213045e-06,
+      "loss": 0.0015,
+      "step": 4218
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8256414545910792e-06,
+      "loss": 0.004,
+      "step": 4219
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825599631960854e-06,
+      "loss": 0.0016,
+      "step": 4220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8255578093306288e-06,
+      "loss": 0.0026,
+      "step": 4221
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8255159867004036e-06,
+      "loss": 0.0011,
+      "step": 4222
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8254741640701784e-06,
+      "loss": 0.3727,
+      "step": 4223
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825432341439953e-06,
+      "loss": 0.7953,
+      "step": 4224
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8253905188097278e-06,
+      "loss": 0.0014,
+      "step": 4225
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8253486961795025e-06,
+      "loss": 0.0027,
+      "step": 4226
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8253068735492773e-06,
+      "loss": 0.0121,
+      "step": 4227
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8252650509190521e-06,
+      "loss": 0.0026,
+      "step": 4228
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.825223228288827e-06,
+      "loss": 0.0519,
+      "step": 4229
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8251814056586017e-06,
+      "loss": 0.0052,
+      "step": 4230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8251395830283767e-06,
+      "loss": 0.0011,
+      "step": 4231
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8250977603981515e-06,
+      "loss": 0.0145,
+      "step": 4232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8250559377679263e-06,
+      "loss": 0.0016,
+      "step": 4233
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.825014115137701e-06,
+      "loss": 3.1025,
+      "step": 4234
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8249722925074758e-06,
+      "loss": 0.0012,
+      "step": 4235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8249304698772506e-06,
+      "loss": 0.0041,
+      "step": 4236
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8248886472470252e-06,
+      "loss": 0.0014,
+      "step": 4237
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8248468246168e-06,
+      "loss": 0.2783,
+      "step": 4238
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8248050019865748e-06,
+      "loss": 0.0027,
+      "step": 4239
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8247631793563495e-06,
+      "loss": 0.078,
+      "step": 4240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8247213567261243e-06,
+      "loss": 0.0034,
+      "step": 4241
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8246795340958991e-06,
+      "loss": 0.0011,
+      "step": 4242
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.824637711465674e-06,
+      "loss": 0.0155,
+      "step": 4243
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8245958888354487e-06,
+      "loss": 0.002,
+      "step": 4244
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8245540662052237e-06,
+      "loss": 0.01,
+      "step": 4245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8245122435749985e-06,
+      "loss": 0.1741,
+      "step": 4246
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8244704209447733e-06,
+      "loss": 2.2084,
+      "step": 4247
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.824428598314548e-06,
+      "loss": 0.0513,
+      "step": 4248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8243867756843228e-06,
+      "loss": 0.0011,
+      "step": 4249
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8243449530540976e-06,
+      "loss": 0.1607,
+      "step": 4250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8243031304238722e-06,
+      "loss": 1.2841,
+      "step": 4251
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.824261307793647e-06,
+      "loss": 0.0017,
+      "step": 4252
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8242194851634218e-06,
+      "loss": 0.0054,
+      "step": 4253
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8241776625331966e-06,
+      "loss": 0.0015,
+      "step": 4254
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8241358399029713e-06,
+      "loss": 0.0012,
+      "step": 4255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8240940172727461e-06,
+      "loss": 0.0016,
+      "step": 4256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.824052194642521e-06,
+      "loss": 0.0027,
+      "step": 4257
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8240103720122957e-06,
+      "loss": 0.0015,
+      "step": 4258
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8239685493820705e-06,
+      "loss": 0.7597,
+      "step": 4259
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8239267267518455e-06,
+      "loss": 0.3634,
+      "step": 4260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8238849041216203e-06,
+      "loss": 0.018,
+      "step": 4261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823843081491395e-06,
+      "loss": 0.0014,
+      "step": 4262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8238012588611698e-06,
+      "loss": 0.0016,
+      "step": 4263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8237594362309446e-06,
+      "loss": 0.1928,
+      "step": 4264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8237176136007192e-06,
+      "loss": 1.1128,
+      "step": 4265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823675790970494e-06,
+      "loss": 0.0143,
+      "step": 4266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8236339683402688e-06,
+      "loss": 0.026,
+      "step": 4267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8235921457100436e-06,
+      "loss": 0.0018,
+      "step": 4268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8235503230798184e-06,
+      "loss": 0.0017,
+      "step": 4269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8235085004495931e-06,
+      "loss": 2.0287,
+      "step": 4270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823466677819368e-06,
+      "loss": 0.0051,
+      "step": 4271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8234248551891427e-06,
+      "loss": 0.0018,
+      "step": 4272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8233830325589175e-06,
+      "loss": 0.3841,
+      "step": 4273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8233412099286923e-06,
+      "loss": 1.9045,
+      "step": 4274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8232993872984673e-06,
+      "loss": 0.0015,
+      "step": 4275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823257564668242e-06,
+      "loss": 0.0401,
+      "step": 4276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8232157420380169e-06,
+      "loss": 0.0019,
+      "step": 4277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8231739194077914e-06,
+      "loss": 0.003,
+      "step": 4278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8231320967775662e-06,
+      "loss": 0.0013,
+      "step": 4279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823090274147341e-06,
+      "loss": 0.0015,
+      "step": 4280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8230484515171158e-06,
+      "loss": 0.0014,
+      "step": 4281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8230066288868906e-06,
+      "loss": 0.0029,
+      "step": 4282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8229648062566654e-06,
+      "loss": 0.0016,
+      "step": 4283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8229229836264401e-06,
+      "loss": 0.0015,
+      "step": 4284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822881160996215e-06,
+      "loss": 0.0016,
+      "step": 4285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8228393383659897e-06,
+      "loss": 0.0071,
+      "step": 4286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8227975157357645e-06,
+      "loss": 0.0277,
+      "step": 4287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8227556931055393e-06,
+      "loss": 0.0091,
+      "step": 4288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8227138704753143e-06,
+      "loss": 1.6599,
+      "step": 4289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822672047845089e-06,
+      "loss": 1.3371,
+      "step": 4290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8226302252148639e-06,
+      "loss": 1.5308,
+      "step": 4291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8225884025846384e-06,
+      "loss": 1.8465,
+      "step": 4292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8225465799544132e-06,
+      "loss": 0.0017,
+      "step": 4293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822504757324188e-06,
+      "loss": 0.0019,
+      "step": 4294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8224629346939628e-06,
+      "loss": 2.3386,
+      "step": 4295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8224211120637376e-06,
+      "loss": 0.2363,
+      "step": 4296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8223792894335124e-06,
+      "loss": 0.0073,
+      "step": 4297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8223374668032872e-06,
+      "loss": 0.0022,
+      "step": 4298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822295644173062e-06,
+      "loss": 0.1418,
+      "step": 4299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8222538215428367e-06,
+      "loss": 0.0022,
+      "step": 4300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8222119989126115e-06,
+      "loss": 0.001,
+      "step": 4301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8221701762823863e-06,
+      "loss": 0.01,
+      "step": 4302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822128353652161e-06,
+      "loss": 0.0058,
+      "step": 4303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822086531021936e-06,
+      "loss": 0.79,
+      "step": 4304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8220447083917107e-06,
+      "loss": 0.001,
+      "step": 4305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8220028857614854e-06,
+      "loss": 2.8993,
+      "step": 4306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8219610631312602e-06,
+      "loss": 0.001,
+      "step": 4307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821919240501035e-06,
+      "loss": 0.0077,
+      "step": 4308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8218774178708098e-06,
+      "loss": 0.0018,
+      "step": 4309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8218355952405846e-06,
+      "loss": 3.4808,
+      "step": 4310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8217937726103594e-06,
+      "loss": 0.0017,
+      "step": 4311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8217519499801342e-06,
+      "loss": 0.1753,
+      "step": 4312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821710127349909e-06,
+      "loss": 0.0011,
+      "step": 4313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8216683047196837e-06,
+      "loss": 0.0049,
+      "step": 4314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8216264820894585e-06,
+      "loss": 0.5704,
+      "step": 4315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8215846594592333e-06,
+      "loss": 0.0016,
+      "step": 4316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821542836829008e-06,
+      "loss": 0.0029,
+      "step": 4317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8215010141987827e-06,
+      "loss": 0.0054,
+      "step": 4318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8214591915685577e-06,
+      "loss": 0.0017,
+      "step": 4319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8214173689383325e-06,
+      "loss": 0.0361,
+      "step": 4320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8213755463081072e-06,
+      "loss": 0.0022,
+      "step": 4321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821333723677882e-06,
+      "loss": 0.0011,
+      "step": 4322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8212919010476568e-06,
+      "loss": 0.0104,
+      "step": 4323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8212500784174316e-06,
+      "loss": 0.429,
+      "step": 4324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8212082557872064e-06,
+      "loss": 0.0017,
+      "step": 4325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8211664331569812e-06,
+      "loss": 0.0125,
+      "step": 4326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821124610526756e-06,
+      "loss": 0.0039,
+      "step": 4327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8210827878965307e-06,
+      "loss": 0.0012,
+      "step": 4328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8210409652663055e-06,
+      "loss": 0.0012,
+      "step": 4329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8209991426360803e-06,
+      "loss": 0.0014,
+      "step": 4330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820957320005855e-06,
+      "loss": 0.0012,
+      "step": 4331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8209154973756297e-06,
+      "loss": 0.0017,
+      "step": 4332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8208736747454047e-06,
+      "loss": 2.6673,
+      "step": 4333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8208318521151795e-06,
+      "loss": 0.0015,
+      "step": 4334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8207900294849542e-06,
+      "loss": 2.6462,
+      "step": 4335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820748206854729e-06,
+      "loss": 0.0037,
+      "step": 4336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8207063842245038e-06,
+      "loss": 0.5912,
+      "step": 4337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8206645615942786e-06,
+      "loss": 0.0016,
+      "step": 4338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8206227389640534e-06,
+      "loss": 0.0017,
+      "step": 4339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8205809163338282e-06,
+      "loss": 0.1198,
+      "step": 4340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820539093703603e-06,
+      "loss": 0.0017,
+      "step": 4341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8204972710733778e-06,
+      "loss": 0.0652,
+      "step": 4342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8204554484431525e-06,
+      "loss": 0.0021,
+      "step": 4343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8204136258129273e-06,
+      "loss": 0.0013,
+      "step": 4344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820371803182702e-06,
+      "loss": 0.0014,
+      "step": 4345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8203299805524767e-06,
+      "loss": 0.0215,
+      "step": 4346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8202881579222515e-06,
+      "loss": 0.0018,
+      "step": 4347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8202463352920265e-06,
+      "loss": 1.8677,
+      "step": 4348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8202045126618013e-06,
+      "loss": 0.0049,
+      "step": 4349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820162690031576e-06,
+      "loss": 0.0013,
+      "step": 4350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8201208674013508e-06,
+      "loss": 0.002,
+      "step": 4351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8200790447711256e-06,
+      "loss": 0.0016,
+      "step": 4352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8200372221409004e-06,
+      "loss": 1.9009,
+      "step": 4353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8199953995106752e-06,
+      "loss": 3.6407,
+      "step": 4354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.81995357688045e-06,
+      "loss": 0.48,
+      "step": 4355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8199117542502248e-06,
+      "loss": 0.0013,
+      "step": 4356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8198699316199995e-06,
+      "loss": 0.0019,
+      "step": 4357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8198281089897743e-06,
+      "loss": 0.0032,
+      "step": 4358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819786286359549e-06,
+      "loss": 0.0013,
+      "step": 4359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8197444637293237e-06,
+      "loss": 0.0024,
+      "step": 4360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8197026410990985e-06,
+      "loss": 0.0077,
+      "step": 4361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8196608184688733e-06,
+      "loss": 0.0054,
+      "step": 4362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8196189958386483e-06,
+      "loss": 1.0367,
+      "step": 4363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819577173208423e-06,
+      "loss": 0.4907,
+      "step": 4364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8195353505781978e-06,
+      "loss": 0.0014,
+      "step": 4365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8194935279479726e-06,
+      "loss": 0.0014,
+      "step": 4366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8194517053177474e-06,
+      "loss": 0.0027,
+      "step": 4367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8194098826875222e-06,
+      "loss": 0.0013,
+      "step": 4368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819368060057297e-06,
+      "loss": 0.0018,
+      "step": 4369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8193262374270718e-06,
+      "loss": 0.0024,
+      "step": 4370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8192844147968466e-06,
+      "loss": 0.0074,
+      "step": 4371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8192425921666213e-06,
+      "loss": 0.9416,
+      "step": 4372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819200769536396e-06,
+      "loss": 0.002,
+      "step": 4373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8191589469061707e-06,
+      "loss": 0.0044,
+      "step": 4374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8191171242759455e-06,
+      "loss": 0.0278,
+      "step": 4375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8190753016457203e-06,
+      "loss": 0.0016,
+      "step": 4376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8190334790154953e-06,
+      "loss": 0.0023,
+      "step": 4377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.81899165638527e-06,
+      "loss": 0.0825,
+      "step": 4378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8189498337550448e-06,
+      "loss": 0.0022,
+      "step": 4379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8189080111248196e-06,
+      "loss": 0.0026,
+      "step": 4380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8188661884945944e-06,
+      "loss": 0.002,
+      "step": 4381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8188243658643692e-06,
+      "loss": 0.0239,
+      "step": 4382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818782543234144e-06,
+      "loss": 1.516,
+      "step": 4383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8187407206039188e-06,
+      "loss": 0.0619,
+      "step": 4384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8186988979736936e-06,
+      "loss": 1.5943,
+      "step": 4385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8186570753434681e-06,
+      "loss": 1.9261,
+      "step": 4386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818615252713243e-06,
+      "loss": 2.4472,
+      "step": 4387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8185734300830177e-06,
+      "loss": 0.0011,
+      "step": 4388
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8185316074527925e-06,
+      "loss": 0.002,
+      "step": 4389
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8184897848225673e-06,
+      "loss": 1.9768,
+      "step": 4390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.818447962192342e-06,
+      "loss": 0.0012,
+      "step": 4391
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.818406139562117e-06,
+      "loss": 0.0025,
+      "step": 4392
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8183643169318919e-06,
+      "loss": 0.0021,
+      "step": 4393
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8183224943016666e-06,
+      "loss": 0.0233,
+      "step": 4394
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8182806716714414e-06,
+      "loss": 0.0033,
+      "step": 4395
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8182388490412162e-06,
+      "loss": 0.0023,
+      "step": 4396
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.818197026410991e-06,
+      "loss": 0.5221,
+      "step": 4397
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8181552037807658e-06,
+      "loss": 0.0553,
+      "step": 4398
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8181133811505406e-06,
+      "loss": 0.3714,
+      "step": 4399
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8180715585203151e-06,
+      "loss": 0.0033,
+      "step": 4400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81802973589009e-06,
+      "loss": 0.0059,
+      "step": 4401
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8179879132598647e-06,
+      "loss": 0.0033,
+      "step": 4402
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8179460906296395e-06,
+      "loss": 0.0031,
+      "step": 4403
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8179042679994143e-06,
+      "loss": 2.4902,
+      "step": 4404
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.817862445369189e-06,
+      "loss": 0.0745,
+      "step": 4405
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8178206227389639e-06,
+      "loss": 0.0145,
+      "step": 4406
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8177788001087389e-06,
+      "loss": 0.6511,
+      "step": 4407
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8177369774785137e-06,
+      "loss": 0.002,
+      "step": 4408
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8176951548482884e-06,
+      "loss": 0.0468,
+      "step": 4409
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8176533322180632e-06,
+      "loss": 0.7636,
+      "step": 4410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.817611509587838e-06,
+      "loss": 0.2039,
+      "step": 4411
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8175696869576128e-06,
+      "loss": 0.0037,
+      "step": 4412
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8175278643273874e-06,
+      "loss": 0.1544,
+      "step": 4413
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8174860416971622e-06,
+      "loss": 1.9299,
+      "step": 4414
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.817444219066937e-06,
+      "loss": 0.0192,
+      "step": 4415
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8174023964367117e-06,
+      "loss": 0.4049,
+      "step": 4416
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8173605738064865e-06,
+      "loss": 0.5472,
+      "step": 4417
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8173187511762613e-06,
+      "loss": 0.0075,
+      "step": 4418
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.817276928546036e-06,
+      "loss": 0.0104,
+      "step": 4419
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8172351059158109e-06,
+      "loss": 0.323,
+      "step": 4420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8171932832855859e-06,
+      "loss": 0.0118,
+      "step": 4421
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8171514606553607e-06,
+      "loss": 0.0015,
+      "step": 4422
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8171096380251354e-06,
+      "loss": 0.0015,
+      "step": 4423
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8170678153949102e-06,
+      "loss": 1.8872,
+      "step": 4424
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.817025992764685e-06,
+      "loss": 0.0062,
+      "step": 4425
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8169841701344598e-06,
+      "loss": 3.4291,
+      "step": 4426
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8169423475042344e-06,
+      "loss": 0.0019,
+      "step": 4427
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8169005248740092e-06,
+      "loss": 0.0022,
+      "step": 4428
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.816858702243784e-06,
+      "loss": 0.0904,
+      "step": 4429
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8168168796135587e-06,
+      "loss": 0.002,
+      "step": 4430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8167750569833335e-06,
+      "loss": 0.0019,
+      "step": 4431
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8167332343531083e-06,
+      "loss": 0.0078,
+      "step": 4432
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.816691411722883e-06,
+      "loss": 2.3767,
+      "step": 4433
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8166495890926579e-06,
+      "loss": 0.0024,
+      "step": 4434
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8166077664624327e-06,
+      "loss": 0.4394,
+      "step": 4435
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8165659438322077e-06,
+      "loss": 0.0015,
+      "step": 4436
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8165241212019825e-06,
+      "loss": 0.1092,
+      "step": 4437
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8164822985717572e-06,
+      "loss": 0.18,
+      "step": 4438
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.816440475941532e-06,
+      "loss": 0.0025,
+      "step": 4439
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8163986533113066e-06,
+      "loss": 0.0315,
+      "step": 4440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8163568306810814e-06,
+      "loss": 0.0038,
+      "step": 4441
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8163150080508562e-06,
+      "loss": 0.0035,
+      "step": 4442
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.816273185420631e-06,
+      "loss": 0.0949,
+      "step": 4443
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8162313627904057e-06,
+      "loss": 1.5137,
+      "step": 4444
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8161895401601805e-06,
+      "loss": 0.0017,
+      "step": 4445
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8161477175299553e-06,
+      "loss": 0.0211,
+      "step": 4446
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81610589489973e-06,
+      "loss": 0.0038,
+      "step": 4447
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8160640722695049e-06,
+      "loss": 0.0016,
+      "step": 4448
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8160222496392797e-06,
+      "loss": 1.1264,
+      "step": 4449
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8159804270090547e-06,
+      "loss": 1.2447,
+      "step": 4450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8159386043788295e-06,
+      "loss": 1.0795,
+      "step": 4451
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8158967817486042e-06,
+      "loss": 0.0026,
+      "step": 4452
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815854959118379e-06,
+      "loss": 0.0051,
+      "step": 4453
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8158131364881536e-06,
+      "loss": 0.0909,
+      "step": 4454
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8157713138579284e-06,
+      "loss": 0.0167,
+      "step": 4455
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8157294912277032e-06,
+      "loss": 0.0037,
+      "step": 4456
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815687668597478e-06,
+      "loss": 0.0011,
+      "step": 4457
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8156458459672528e-06,
+      "loss": 0.0101,
+      "step": 4458
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8156040233370275e-06,
+      "loss": 1.3447,
+      "step": 4459
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8155622007068023e-06,
+      "loss": 1.6106,
+      "step": 4460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8155203780765771e-06,
+      "loss": 0.0092,
+      "step": 4461
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815478555446352e-06,
+      "loss": 0.0456,
+      "step": 4462
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8154367328161267e-06,
+      "loss": 0.0011,
+      "step": 4463
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8153949101859015e-06,
+      "loss": 0.0018,
+      "step": 4464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8153530875556765e-06,
+      "loss": 0.0099,
+      "step": 4465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8153112649254513e-06,
+      "loss": 0.0024,
+      "step": 4466
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815269442295226e-06,
+      "loss": 1.5966,
+      "step": 4467
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8152276196650006e-06,
+      "loss": 0.0178,
+      "step": 4468
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8151857970347754e-06,
+      "loss": 0.1214,
+      "step": 4469
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8151439744045502e-06,
+      "loss": 0.0051,
+      "step": 4470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815102151774325e-06,
+      "loss": 0.0019,
+      "step": 4471
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8150603291440998e-06,
+      "loss": 0.0018,
+      "step": 4472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8150185065138746e-06,
+      "loss": 0.0019,
+      "step": 4473
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8149766838836493e-06,
+      "loss": 0.0014,
+      "step": 4474
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8149348612534241e-06,
+      "loss": 0.0022,
+      "step": 4475
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814893038623199e-06,
+      "loss": 0.0013,
+      "step": 4476
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8148512159929737e-06,
+      "loss": 0.0081,
+      "step": 4477
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8148093933627485e-06,
+      "loss": 0.0018,
+      "step": 4478
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8147675707325233e-06,
+      "loss": 0.0015,
+      "step": 4479
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8147257481022983e-06,
+      "loss": 0.0013,
+      "step": 4480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8146839254720728e-06,
+      "loss": 1.692,
+      "step": 4481
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8146421028418476e-06,
+      "loss": 0.0091,
+      "step": 4482
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8146002802116224e-06,
+      "loss": 0.0017,
+      "step": 4483
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8145584575813972e-06,
+      "loss": 0.0029,
+      "step": 4484
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814516634951172e-06,
+      "loss": 2.6151,
+      "step": 4485
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8144748123209468e-06,
+      "loss": 0.0012,
+      "step": 4486
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8144329896907216e-06,
+      "loss": 0.5011,
+      "step": 4487
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8143911670604963e-06,
+      "loss": 0.0108,
+      "step": 4488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8143493444302711e-06,
+      "loss": 0.0064,
+      "step": 4489
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814307521800046e-06,
+      "loss": 0.0038,
+      "step": 4490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8142656991698207e-06,
+      "loss": 0.0031,
+      "step": 4491
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8142238765395955e-06,
+      "loss": 0.152,
+      "step": 4492
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8141820539093703e-06,
+      "loss": 0.8271,
+      "step": 4493
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8141402312791453e-06,
+      "loss": 0.7125,
+      "step": 4494
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8140984086489198e-06,
+      "loss": 0.0021,
+      "step": 4495
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8140565860186946e-06,
+      "loss": 0.6356,
+      "step": 4496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8140147633884694e-06,
+      "loss": 0.1798,
+      "step": 4497
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8139729407582442e-06,
+      "loss": 0.0034,
+      "step": 4498
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813931118128019e-06,
+      "loss": 0.7039,
+      "step": 4499
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8138892954977938e-06,
+      "loss": 0.0011,
+      "step": 4500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8138474728675686e-06,
+      "loss": 0.0018,
+      "step": 4501
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8138056502373434e-06,
+      "loss": 0.0011,
+      "step": 4502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8137638276071181e-06,
+      "loss": 0.004,
+      "step": 4503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813722004976893e-06,
+      "loss": 0.0015,
+      "step": 4504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8136801823466677e-06,
+      "loss": 0.0023,
+      "step": 4505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8136383597164425e-06,
+      "loss": 0.0016,
+      "step": 4506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8135965370862173e-06,
+      "loss": 0.0029,
+      "step": 4507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8135547144559919e-06,
+      "loss": 0.002,
+      "step": 4508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8135128918257669e-06,
+      "loss": 2.7613,
+      "step": 4509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8134710691955416e-06,
+      "loss": 0.001,
+      "step": 4510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8134292465653164e-06,
+      "loss": 0.2575,
+      "step": 4511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8133874239350912e-06,
+      "loss": 0.0173,
+      "step": 4512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813345601304866e-06,
+      "loss": 2.4179,
+      "step": 4513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8133037786746408e-06,
+      "loss": 1.9116,
+      "step": 4514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8132619560444156e-06,
+      "loss": 0.0011,
+      "step": 4515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8132201334141904e-06,
+      "loss": 3.3294,
+      "step": 4516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8131783107839651e-06,
+      "loss": 0.0017,
+      "step": 4517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81313648815374e-06,
+      "loss": 0.0011,
+      "step": 4518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8130946655235147e-06,
+      "loss": 0.0015,
+      "step": 4519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8130528428932895e-06,
+      "loss": 0.1375,
+      "step": 4520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813011020263064e-06,
+      "loss": 0.002,
+      "step": 4521
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8129691976328389e-06,
+      "loss": 0.0012,
+      "step": 4522
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8129273750026137e-06,
+      "loss": 0.0032,
+      "step": 4523
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8128855523723887e-06,
+      "loss": 0.0052,
+      "step": 4524
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8128437297421634e-06,
+      "loss": 0.0013,
+      "step": 4525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8128019071119382e-06,
+      "loss": 0.0014,
+      "step": 4526
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812760084481713e-06,
+      "loss": 1.2342,
+      "step": 4527
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8127182618514878e-06,
+      "loss": 0.0048,
+      "step": 4528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8126764392212626e-06,
+      "loss": 0.0037,
+      "step": 4529
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8126346165910374e-06,
+      "loss": 0.0019,
+      "step": 4530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8125927939608122e-06,
+      "loss": 2.7534,
+      "step": 4531
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812550971330587e-06,
+      "loss": 0.0015,
+      "step": 4532
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8125091487003617e-06,
+      "loss": 0.0013,
+      "step": 4533
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8124673260701365e-06,
+      "loss": 0.0014,
+      "step": 4534
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812425503439911e-06,
+      "loss": 0.0021,
+      "step": 4535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8123836808096859e-06,
+      "loss": 2.0775,
+      "step": 4536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8123418581794607e-06,
+      "loss": 0.0015,
+      "step": 4537
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8123000355492357e-06,
+      "loss": 0.0013,
+      "step": 4538
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8122582129190104e-06,
+      "loss": 0.1802,
+      "step": 4539
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8122163902887852e-06,
+      "loss": 2.0959,
+      "step": 4540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81217456765856e-06,
+      "loss": 0.0028,
+      "step": 4541
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8121327450283348e-06,
+      "loss": 0.0017,
+      "step": 4542
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8120909223981096e-06,
+      "loss": 0.0062,
+      "step": 4543
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8120490997678844e-06,
+      "loss": 0.0018,
+      "step": 4544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8120072771376592e-06,
+      "loss": 0.0016,
+      "step": 4545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811965454507434e-06,
+      "loss": 0.0026,
+      "step": 4546
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8119236318772087e-06,
+      "loss": 0.3652,
+      "step": 4547
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8118818092469833e-06,
+      "loss": 0.002,
+      "step": 4548
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.811839986616758e-06,
+      "loss": 3.8531,
+      "step": 4549
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8117981639865329e-06,
+      "loss": 0.0017,
+      "step": 4550
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8117563413563077e-06,
+      "loss": 0.001,
+      "step": 4551
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8117145187260825e-06,
+      "loss": 1.9614,
+      "step": 4552
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8116726960958575e-06,
+      "loss": 0.0012,
+      "step": 4553
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8116308734656322e-06,
+      "loss": 0.0015,
+      "step": 4554
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.811589050835407e-06,
+      "loss": 0.0015,
+      "step": 4555
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8115472282051818e-06,
+      "loss": 0.0034,
+      "step": 4556
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8115054055749566e-06,
+      "loss": 0.005,
+      "step": 4557
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8114635829447314e-06,
+      "loss": 0.0017,
+      "step": 4558
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8114217603145062e-06,
+      "loss": 0.002,
+      "step": 4559
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.811379937684281e-06,
+      "loss": 1.6571,
+      "step": 4560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8113381150540557e-06,
+      "loss": 0.0016,
+      "step": 4561
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8112962924238303e-06,
+      "loss": 0.0037,
+      "step": 4562
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8112544697936051e-06,
+      "loss": 0.0013,
+      "step": 4563
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.81121264716338e-06,
+      "loss": 0.0011,
+      "step": 4564
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8111708245331547e-06,
+      "loss": 0.3674,
+      "step": 4565
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8111290019029295e-06,
+      "loss": 0.0012,
+      "step": 4566
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8110871792727043e-06,
+      "loss": 0.0018,
+      "step": 4567
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8110453566424793e-06,
+      "loss": 3.6036,
+      "step": 4568
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.811003534012254e-06,
+      "loss": 0.0014,
+      "step": 4569
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8109617113820288e-06,
+      "loss": 0.0009,
+      "step": 4570
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8109198887518036e-06,
+      "loss": 1.7496,
+      "step": 4571
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8108780661215784e-06,
+      "loss": 0.0073,
+      "step": 4572
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8108362434913532e-06,
+      "loss": 0.27,
+      "step": 4573
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.810794420861128e-06,
+      "loss": 0.0015,
+      "step": 4574
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8107525982309028e-06,
+      "loss": 0.0012,
+      "step": 4575
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8107107756006773e-06,
+      "loss": 0.0087,
+      "step": 4576
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8106689529704521e-06,
+      "loss": 0.363,
+      "step": 4577
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.810627130340227e-06,
+      "loss": 0.0013,
+      "step": 4578
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8105853077100017e-06,
+      "loss": 1.5226,
+      "step": 4579
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8105434850797765e-06,
+      "loss": 0.0038,
+      "step": 4580
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8105016624495513e-06,
+      "loss": 0.0021,
+      "step": 4581
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8104598398193263e-06,
+      "loss": 0.0066,
+      "step": 4582
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.810418017189101e-06,
+      "loss": 1.5544,
+      "step": 4583
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8103761945588758e-06,
+      "loss": 0.0011,
+      "step": 4584
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8103343719286506e-06,
+      "loss": 1.1388,
+      "step": 4585
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8102925492984254e-06,
+      "loss": 0.0035,
+      "step": 4586
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8102507266682002e-06,
+      "loss": 0.5992,
+      "step": 4587
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.810208904037975e-06,
+      "loss": 0.0012,
+      "step": 4588
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8101670814077496e-06,
+      "loss": 0.0015,
+      "step": 4589
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8101252587775243e-06,
+      "loss": 1.2314,
+      "step": 4590
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8100834361472991e-06,
+      "loss": 0.0014,
+      "step": 4591
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.810041613517074e-06,
+      "loss": 0.0035,
+      "step": 4592
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8099997908868487e-06,
+      "loss": 0.008,
+      "step": 4593
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8099579682566235e-06,
+      "loss": 1.6468,
+      "step": 4594
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8099161456263983e-06,
+      "loss": 0.0012,
+      "step": 4595
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809874322996173e-06,
+      "loss": 0.3423,
+      "step": 4596
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809832500365948e-06,
+      "loss": 0.0017,
+      "step": 4597
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8097906777357228e-06,
+      "loss": 0.0014,
+      "step": 4598
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8097488551054976e-06,
+      "loss": 0.0705,
+      "step": 4599
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8097070324752724e-06,
+      "loss": 0.001,
+      "step": 4600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8096652098450472e-06,
+      "loss": 0.0018,
+      "step": 4601
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809623387214822e-06,
+      "loss": 0.1004,
+      "step": 4602
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8095815645845966e-06,
+      "loss": 0.0037,
+      "step": 4603
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8095397419543713e-06,
+      "loss": 0.0015,
+      "step": 4604
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8094979193241461e-06,
+      "loss": 0.0018,
+      "step": 4605
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809456096693921e-06,
+      "loss": 0.0012,
+      "step": 4606
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8094142740636957e-06,
+      "loss": 0.0182,
+      "step": 4607
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8093724514334705e-06,
+      "loss": 0.2407,
+      "step": 4608
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8093306288032453e-06,
+      "loss": 0.0025,
+      "step": 4609
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80928880617302e-06,
+      "loss": 0.0051,
+      "step": 4610
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8092469835427949e-06,
+      "loss": 2.4773,
+      "step": 4611
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8092051609125698e-06,
+      "loss": 0.0331,
+      "step": 4612
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8091633382823446e-06,
+      "loss": 0.0021,
+      "step": 4613
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8091215156521194e-06,
+      "loss": 0.4007,
+      "step": 4614
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8090796930218942e-06,
+      "loss": 0.0013,
+      "step": 4615
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8090378703916688e-06,
+      "loss": 0.4154,
+      "step": 4616
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8089960477614436e-06,
+      "loss": 0.0012,
+      "step": 4617
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8089542251312184e-06,
+      "loss": 0.401,
+      "step": 4618
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8089124025009931e-06,
+      "loss": 0.0056,
+      "step": 4619
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808870579870768e-06,
+      "loss": 0.0312,
+      "step": 4620
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8088287572405427e-06,
+      "loss": 0.0026,
+      "step": 4621
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8087869346103175e-06,
+      "loss": 0.0042,
+      "step": 4622
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8087451119800923e-06,
+      "loss": 0.0022,
+      "step": 4623
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808703289349867e-06,
+      "loss": 0.0099,
+      "step": 4624
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8086614667196419e-06,
+      "loss": 1.4984,
+      "step": 4625
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8086196440894169e-06,
+      "loss": 0.0014,
+      "step": 4626
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8085778214591916e-06,
+      "loss": 0.0018,
+      "step": 4627
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8085359988289664e-06,
+      "loss": 0.6825,
+      "step": 4628
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8084941761987412e-06,
+      "loss": 0.0025,
+      "step": 4629
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8084523535685158e-06,
+      "loss": 2.025,
+      "step": 4630
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8084105309382906e-06,
+      "loss": 0.0291,
+      "step": 4631
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8083687083080654e-06,
+      "loss": 0.0578,
+      "step": 4632
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8083268856778402e-06,
+      "loss": 0.0013,
+      "step": 4633
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808285063047615e-06,
+      "loss": 0.0343,
+      "step": 4634
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8082432404173897e-06,
+      "loss": 0.0013,
+      "step": 4635
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8082014177871645e-06,
+      "loss": 0.0047,
+      "step": 4636
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8081595951569393e-06,
+      "loss": 0.0036,
+      "step": 4637
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808117772526714e-06,
+      "loss": 0.0021,
+      "step": 4638
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8080759498964889e-06,
+      "loss": 0.3183,
+      "step": 4639
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8080341272662637e-06,
+      "loss": 0.0016,
+      "step": 4640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8079923046360387e-06,
+      "loss": 0.006,
+      "step": 4641
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8079504820058134e-06,
+      "loss": 0.0021,
+      "step": 4642
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807908659375588e-06,
+      "loss": 0.4468,
+      "step": 4643
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8078668367453628e-06,
+      "loss": 0.0013,
+      "step": 4644
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8078250141151376e-06,
+      "loss": 0.0025,
+      "step": 4645
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8077831914849124e-06,
+      "loss": 0.4989,
+      "step": 4646
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8077413688546872e-06,
+      "loss": 0.6038,
+      "step": 4647
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807699546224462e-06,
+      "loss": 0.0018,
+      "step": 4648
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8076577235942367e-06,
+      "loss": 0.4728,
+      "step": 4649
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8076159009640115e-06,
+      "loss": 2.0496,
+      "step": 4650
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8075740783337863e-06,
+      "loss": 0.0031,
+      "step": 4651
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807532255703561e-06,
+      "loss": 0.0195,
+      "step": 4652
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8074904330733359e-06,
+      "loss": 0.1344,
+      "step": 4653
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8074486104431107e-06,
+      "loss": 1.0362,
+      "step": 4654
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8074067878128857e-06,
+      "loss": 0.0019,
+      "step": 4655
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8073649651826604e-06,
+      "loss": 0.0029,
+      "step": 4656
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807323142552435e-06,
+      "loss": 0.0195,
+      "step": 4657
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8072813199222098e-06,
+      "loss": 0.0053,
+      "step": 4658
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8072394972919846e-06,
+      "loss": 0.8515,
+      "step": 4659
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8071976746617594e-06,
+      "loss": 0.0021,
+      "step": 4660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8071558520315342e-06,
+      "loss": 0.0014,
+      "step": 4661
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807114029401309e-06,
+      "loss": 0.0011,
+      "step": 4662
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8070722067710837e-06,
+      "loss": 0.0017,
+      "step": 4663
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8070303841408585e-06,
+      "loss": 1.3791,
+      "step": 4664
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8069885615106333e-06,
+      "loss": 0.008,
+      "step": 4665
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806946738880408e-06,
+      "loss": 0.0096,
+      "step": 4666
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8069049162501829e-06,
+      "loss": 0.0071,
+      "step": 4667
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8068630936199577e-06,
+      "loss": 0.0012,
+      "step": 4668
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8068212709897325e-06,
+      "loss": 0.0565,
+      "step": 4669
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8067794483595075e-06,
+      "loss": 0.0012,
+      "step": 4670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806737625729282e-06,
+      "loss": 0.002,
+      "step": 4671
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8066958030990568e-06,
+      "loss": 0.0202,
+      "step": 4672
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8066539804688316e-06,
+      "loss": 2.6179,
+      "step": 4673
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8066121578386064e-06,
+      "loss": 0.0795,
+      "step": 4674
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8065703352083812e-06,
+      "loss": 0.0066,
+      "step": 4675
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806528512578156e-06,
+      "loss": 0.001,
+      "step": 4676
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8064866899479307e-06,
+      "loss": 0.0936,
+      "step": 4677
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8064448673177055e-06,
+      "loss": 0.0012,
+      "step": 4678
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8064030446874803e-06,
+      "loss": 0.003,
+      "step": 4679
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8063612220572551e-06,
+      "loss": 3.4785,
+      "step": 4680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80631939942703e-06,
+      "loss": 0.0596,
+      "step": 4681
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8062775767968047e-06,
+      "loss": 0.0304,
+      "step": 4682
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8062357541665795e-06,
+      "loss": 0.0024,
+      "step": 4683
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806193931536354e-06,
+      "loss": 0.0012,
+      "step": 4684
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806152108906129e-06,
+      "loss": 0.0012,
+      "step": 4685
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8061102862759038e-06,
+      "loss": 0.0019,
+      "step": 4686
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8060684636456786e-06,
+      "loss": 0.0083,
+      "step": 4687
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8060266410154534e-06,
+      "loss": 0.0011,
+      "step": 4688
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8059848183852282e-06,
+      "loss": 0.4155,
+      "step": 4689
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805942995755003e-06,
+      "loss": 0.0035,
+      "step": 4690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8059011731247778e-06,
+      "loss": 0.0008,
+      "step": 4691
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8058593504945525e-06,
+      "loss": 0.1205,
+      "step": 4692
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8058175278643273e-06,
+      "loss": 0.0026,
+      "step": 4693
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8057757052341021e-06,
+      "loss": 0.0011,
+      "step": 4694
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805733882603877e-06,
+      "loss": 0.0147,
+      "step": 4695
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8056920599736517e-06,
+      "loss": 0.0014,
+      "step": 4696
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8056502373434263e-06,
+      "loss": 0.0027,
+      "step": 4697
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805608414713201e-06,
+      "loss": 0.1547,
+      "step": 4698
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805566592082976e-06,
+      "loss": 0.0471,
+      "step": 4699
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8055247694527508e-06,
+      "loss": 0.0014,
+      "step": 4700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8054829468225256e-06,
+      "loss": 0.0067,
+      "step": 4701
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8054411241923004e-06,
+      "loss": 2.3455,
+      "step": 4702
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8053993015620752e-06,
+      "loss": 2.5776,
+      "step": 4703
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80535747893185e-06,
+      "loss": 0.0267,
+      "step": 4704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8053156563016248e-06,
+      "loss": 0.6537,
+      "step": 4705
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8052738336713996e-06,
+      "loss": 0.4797,
+      "step": 4706
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8052320110411743e-06,
+      "loss": 0.0013,
+      "step": 4707
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8051901884109491e-06,
+      "loss": 0.3241,
+      "step": 4708
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.805148365780724e-06,
+      "loss": 2.9905,
+      "step": 4709
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8051065431504987e-06,
+      "loss": 0.0039,
+      "step": 4710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8050647205202733e-06,
+      "loss": 0.0012,
+      "step": 4711
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.805022897890048e-06,
+      "loss": 0.0013,
+      "step": 4712
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8049810752598228e-06,
+      "loss": 0.0023,
+      "step": 4713
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8049392526295978e-06,
+      "loss": 0.0023,
+      "step": 4714
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8048974299993726e-06,
+      "loss": 1.8294,
+      "step": 4715
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8048556073691474e-06,
+      "loss": 0.003,
+      "step": 4716
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8048137847389222e-06,
+      "loss": 0.0012,
+      "step": 4717
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.804771962108697e-06,
+      "loss": 0.001,
+      "step": 4718
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8047301394784718e-06,
+      "loss": 0.0023,
+      "step": 4719
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8046883168482466e-06,
+      "loss": 2.2314,
+      "step": 4720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8046464942180213e-06,
+      "loss": 0.5074,
+      "step": 4721
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8046046715877961e-06,
+      "loss": 0.0019,
+      "step": 4722
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.804562848957571e-06,
+      "loss": 0.0012,
+      "step": 4723
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8045210263273455e-06,
+      "loss": 1.6183,
+      "step": 4724
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8044792036971203e-06,
+      "loss": 0.0037,
+      "step": 4725
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.804437381066895e-06,
+      "loss": 0.007,
+      "step": 4726
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8043955584366699e-06,
+      "loss": 3.365,
+      "step": 4727
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8043537358064446e-06,
+      "loss": 0.4396,
+      "step": 4728
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8043119131762196e-06,
+      "loss": 0.0703,
+      "step": 4729
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8042700905459944e-06,
+      "loss": 2.688,
+      "step": 4730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8042282679157692e-06,
+      "loss": 0.1929,
+      "step": 4731
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.804186445285544e-06,
+      "loss": 0.0463,
+      "step": 4732
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8041446226553188e-06,
+      "loss": 0.0037,
+      "step": 4733
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8041028000250936e-06,
+      "loss": 0.0031,
+      "step": 4734
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8040609773948684e-06,
+      "loss": 0.0027,
+      "step": 4735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8040191547646431e-06,
+      "loss": 0.0069,
+      "step": 4736
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803977332134418e-06,
+      "loss": 0.0215,
+      "step": 4737
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8039355095041925e-06,
+      "loss": 0.0155,
+      "step": 4738
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8038936868739673e-06,
+      "loss": 3.3969,
+      "step": 4739
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803851864243742e-06,
+      "loss": 0.0027,
+      "step": 4740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8038100416135169e-06,
+      "loss": 1.5398,
+      "step": 4741
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8037682189832916e-06,
+      "loss": 0.001,
+      "step": 4742
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8037263963530666e-06,
+      "loss": 0.0887,
+      "step": 4743
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8036845737228414e-06,
+      "loss": 2.5375,
+      "step": 4744
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8036427510926162e-06,
+      "loss": 2.1035,
+      "step": 4745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803600928462391e-06,
+      "loss": 3.4089,
+      "step": 4746
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8035591058321658e-06,
+      "loss": 0.0278,
+      "step": 4747
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8035172832019406e-06,
+      "loss": 0.0016,
+      "step": 4748
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8034754605717154e-06,
+      "loss": 0.014,
+      "step": 4749
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8034336379414902e-06,
+      "loss": 0.0015,
+      "step": 4750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8033918153112647e-06,
+      "loss": 0.0053,
+      "step": 4751
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8033499926810395e-06,
+      "loss": 0.3202,
+      "step": 4752
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8033081700508143e-06,
+      "loss": 0.0062,
+      "step": 4753
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803266347420589e-06,
+      "loss": 0.0031,
+      "step": 4754
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8032245247903639e-06,
+      "loss": 0.0016,
+      "step": 4755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8031827021601387e-06,
+      "loss": 0.0897,
+      "step": 4756
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8031408795299134e-06,
+      "loss": 0.2224,
+      "step": 4757
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8030990568996884e-06,
+      "loss": 0.0094,
+      "step": 4758
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8030572342694632e-06,
+      "loss": 0.014,
+      "step": 4759
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803015411639238e-06,
+      "loss": 0.0199,
+      "step": 4760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8029735890090128e-06,
+      "loss": 0.0055,
+      "step": 4761
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8029317663787876e-06,
+      "loss": 0.0034,
+      "step": 4762
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8028899437485624e-06,
+      "loss": 0.0024,
+      "step": 4763
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8028481211183372e-06,
+      "loss": 0.6171,
+      "step": 4764
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8028062984881117e-06,
+      "loss": 0.0026,
+      "step": 4765
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8027644758578865e-06,
+      "loss": 0.0479,
+      "step": 4766
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8027226532276613e-06,
+      "loss": 0.0224,
+      "step": 4767
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802680830597436e-06,
+      "loss": 0.0015,
+      "step": 4768
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8026390079672109e-06,
+      "loss": 0.0413,
+      "step": 4769
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8025971853369857e-06,
+      "loss": 2.5888,
+      "step": 4770
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8025553627067605e-06,
+      "loss": 0.0193,
+      "step": 4771
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8025135400765352e-06,
+      "loss": 0.006,
+      "step": 4772
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8024717174463102e-06,
+      "loss": 0.0028,
+      "step": 4773
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802429894816085e-06,
+      "loss": 0.0875,
+      "step": 4774
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8023880721858598e-06,
+      "loss": 0.001,
+      "step": 4775
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8023462495556346e-06,
+      "loss": 0.0035,
+      "step": 4776
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8023044269254094e-06,
+      "loss": 3.2903,
+      "step": 4777
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8022626042951842e-06,
+      "loss": 2.6209,
+      "step": 4778
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8022207816649587e-06,
+      "loss": 0.0015,
+      "step": 4779
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8021789590347335e-06,
+      "loss": 0.0019,
+      "step": 4780
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8021371364045083e-06,
+      "loss": 0.0096,
+      "step": 4781
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802095313774283e-06,
+      "loss": 0.2372,
+      "step": 4782
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8020534911440579e-06,
+      "loss": 0.0025,
+      "step": 4783
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8020116685138327e-06,
+      "loss": 0.0017,
+      "step": 4784
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8019698458836075e-06,
+      "loss": 0.0022,
+      "step": 4785
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8019280232533822e-06,
+      "loss": 0.0213,
+      "step": 4786
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8018862006231572e-06,
+      "loss": 0.0019,
+      "step": 4787
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801844377992932e-06,
+      "loss": 1.5943,
+      "step": 4788
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8018025553627068e-06,
+      "loss": 2.6028,
+      "step": 4789
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8017607327324816e-06,
+      "loss": 0.0026,
+      "step": 4790
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8017189101022564e-06,
+      "loss": 0.0053,
+      "step": 4791
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801677087472031e-06,
+      "loss": 2.3614,
+      "step": 4792
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8016352648418058e-06,
+      "loss": 0.0022,
+      "step": 4793
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8015934422115805e-06,
+      "loss": 2.1619,
+      "step": 4794
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8015516195813553e-06,
+      "loss": 2.8051,
+      "step": 4795
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8015097969511301e-06,
+      "loss": 0.0024,
+      "step": 4796
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801467974320905e-06,
+      "loss": 0.0018,
+      "step": 4797
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8014261516906797e-06,
+      "loss": 0.5355,
+      "step": 4798
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8013843290604545e-06,
+      "loss": 0.2453,
+      "step": 4799
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8013425064302293e-06,
+      "loss": 1.3679,
+      "step": 4800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801300683800004e-06,
+      "loss": 0.002,
+      "step": 4801
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801258861169779e-06,
+      "loss": 0.0034,
+      "step": 4802
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8012170385395538e-06,
+      "loss": 0.0251,
+      "step": 4803
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8011752159093286e-06,
+      "loss": 0.3101,
+      "step": 4804
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8011333932791034e-06,
+      "loss": 0.6785,
+      "step": 4805
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801091570648878e-06,
+      "loss": 0.0044,
+      "step": 4806
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8010497480186528e-06,
+      "loss": 0.0078,
+      "step": 4807
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8010079253884275e-06,
+      "loss": 0.0743,
+      "step": 4808
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8009661027582023e-06,
+      "loss": 0.003,
+      "step": 4809
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8009242801279771e-06,
+      "loss": 0.0015,
+      "step": 4810
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800882457497752e-06,
+      "loss": 0.3478,
+      "step": 4811
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8008406348675267e-06,
+      "loss": 2.9302,
+      "step": 4812
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8007988122373015e-06,
+      "loss": 0.3272,
+      "step": 4813
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8007569896070763e-06,
+      "loss": 0.0701,
+      "step": 4814
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800715166976851e-06,
+      "loss": 2.1755,
+      "step": 4815
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800673344346626e-06,
+      "loss": 0.0417,
+      "step": 4816
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8006315217164008e-06,
+      "loss": 0.0107,
+      "step": 4817
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8005896990861756e-06,
+      "loss": 1.7206,
+      "step": 4818
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8005478764559502e-06,
+      "loss": 0.0158,
+      "step": 4819
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800506053825725e-06,
+      "loss": 1.9518,
+      "step": 4820
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8004642311954998e-06,
+      "loss": 0.0034,
+      "step": 4821
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8004224085652746e-06,
+      "loss": 0.6187,
+      "step": 4822
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8003805859350493e-06,
+      "loss": 0.0028,
+      "step": 4823
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8003387633048241e-06,
+      "loss": 0.0032,
+      "step": 4824
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800296940674599e-06,
+      "loss": 0.0079,
+      "step": 4825
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8002551180443737e-06,
+      "loss": 0.0349,
+      "step": 4826
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8002132954141485e-06,
+      "loss": 0.0053,
+      "step": 4827
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8001714727839233e-06,
+      "loss": 0.0271,
+      "step": 4828
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800129650153698e-06,
+      "loss": 0.0024,
+      "step": 4829
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8000878275234728e-06,
+      "loss": 0.6317,
+      "step": 4830
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8000460048932478e-06,
+      "loss": 0.0062,
+      "step": 4831
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8000041822630226e-06,
+      "loss": 0.0025,
+      "step": 4832
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7999623596327972e-06,
+      "loss": 0.0036,
+      "step": 4833
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799920537002572e-06,
+      "loss": 0.0022,
+      "step": 4834
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7998787143723468e-06,
+      "loss": 0.0044,
+      "step": 4835
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7998368917421216e-06,
+      "loss": 0.0022,
+      "step": 4836
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7997950691118964e-06,
+      "loss": 0.0466,
+      "step": 4837
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7997532464816711e-06,
+      "loss": 0.0619,
+      "step": 4838
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799711423851446e-06,
+      "loss": 0.0059,
+      "step": 4839
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7996696012212207e-06,
+      "loss": 0.003,
+      "step": 4840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7996277785909955e-06,
+      "loss": 2.9009,
+      "step": 4841
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7995859559607703e-06,
+      "loss": 2.0789,
+      "step": 4842
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799544133330545e-06,
+      "loss": 0.0015,
+      "step": 4843
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7995023107003199e-06,
+      "loss": 0.0031,
+      "step": 4844
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7994604880700946e-06,
+      "loss": 2.0365,
+      "step": 4845
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7994186654398694e-06,
+      "loss": 0.0022,
+      "step": 4846
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7993768428096442e-06,
+      "loss": 1.3512,
+      "step": 4847
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799335020179419e-06,
+      "loss": 0.0035,
+      "step": 4848
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7992931975491938e-06,
+      "loss": 0.0035,
+      "step": 4849
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7992513749189686e-06,
+      "loss": 0.0013,
+      "step": 4850
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7992095522887434e-06,
+      "loss": 0.0027,
+      "step": 4851
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7991677296585181e-06,
+      "loss": 2.0991,
+      "step": 4852
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799125907028293e-06,
+      "loss": 2.2818,
+      "step": 4853
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7990840843980677e-06,
+      "loss": 0.0311,
+      "step": 4854
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7990422617678425e-06,
+      "loss": 0.0074,
+      "step": 4855
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7990004391376173e-06,
+      "loss": 0.0017,
+      "step": 4856
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798958616507392e-06,
+      "loss": 0.0312,
+      "step": 4857
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7989167938771669e-06,
+      "loss": 1.6713,
+      "step": 4858
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7988749712469414e-06,
+      "loss": 0.0021,
+      "step": 4859
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7988331486167164e-06,
+      "loss": 0.0043,
+      "step": 4860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7987913259864912e-06,
+      "loss": 0.0667,
+      "step": 4861
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798749503356266e-06,
+      "loss": 0.0021,
+      "step": 4862
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7987076807260408e-06,
+      "loss": 0.004,
+      "step": 4863
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7986658580958156e-06,
+      "loss": 0.0032,
+      "step": 4864
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7986240354655904e-06,
+      "loss": 1.6457,
+      "step": 4865
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7985822128353652e-06,
+      "loss": 0.0051,
+      "step": 4866
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79854039020514e-06,
+      "loss": 0.0042,
+      "step": 4867
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7984985675749147e-06,
+      "loss": 0.0026,
+      "step": 4868
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7984567449446895e-06,
+      "loss": 0.0032,
+      "step": 4869
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7984149223144643e-06,
+      "loss": 0.058,
+      "step": 4870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.798373099684239e-06,
+      "loss": 0.0036,
+      "step": 4871
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7983312770540139e-06,
+      "loss": 0.0026,
+      "step": 4872
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7982894544237884e-06,
+      "loss": 0.0013,
+      "step": 4873
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7982476317935632e-06,
+      "loss": 1.5115,
+      "step": 4874
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7982058091633382e-06,
+      "loss": 0.3176,
+      "step": 4875
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.798163986533113e-06,
+      "loss": 0.0019,
+      "step": 4876
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7981221639028878e-06,
+      "loss": 0.0019,
+      "step": 4877
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7980803412726626e-06,
+      "loss": 0.0022,
+      "step": 4878
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7980385186424374e-06,
+      "loss": 0.0022,
+      "step": 4879
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7979966960122122e-06,
+      "loss": 0.0026,
+      "step": 4880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.797954873381987e-06,
+      "loss": 0.002,
+      "step": 4881
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7979130507517617e-06,
+      "loss": 0.0035,
+      "step": 4882
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7978712281215365e-06,
+      "loss": 0.0443,
+      "step": 4883
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7978294054913113e-06,
+      "loss": 0.2977,
+      "step": 4884
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.797787582861086e-06,
+      "loss": 0.0666,
+      "step": 4885
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7977457602308609e-06,
+      "loss": 0.0369,
+      "step": 4886
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7977039376006355e-06,
+      "loss": 0.008,
+      "step": 4887
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7976621149704102e-06,
+      "loss": 0.0015,
+      "step": 4888
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.797620292340185e-06,
+      "loss": 0.0024,
+      "step": 4889
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79757846970996e-06,
+      "loss": 0.0053,
+      "step": 4890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7975366470797348e-06,
+      "loss": 0.0023,
+      "step": 4891
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7974948244495096e-06,
+      "loss": 0.0027,
+      "step": 4892
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7974530018192844e-06,
+      "loss": 0.199,
+      "step": 4893
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7974111791890592e-06,
+      "loss": 0.0021,
+      "step": 4894
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.797369356558834e-06,
+      "loss": 0.0014,
+      "step": 4895
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7973275339286087e-06,
+      "loss": 0.0026,
+      "step": 4896
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7972857112983835e-06,
+      "loss": 0.0093,
+      "step": 4897
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7972438886681583e-06,
+      "loss": 0.0016,
+      "step": 4898
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.797202066037933e-06,
+      "loss": 0.0021,
+      "step": 4899
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7971602434077077e-06,
+      "loss": 0.0015,
+      "step": 4900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7971184207774825e-06,
+      "loss": 2.7975,
+      "step": 4901
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7970765981472573e-06,
+      "loss": 0.0083,
+      "step": 4902
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.797034775517032e-06,
+      "loss": 0.0017,
+      "step": 4903
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796992952886807e-06,
+      "loss": 0.0011,
+      "step": 4904
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7969511302565818e-06,
+      "loss": 2.447,
+      "step": 4905
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7969093076263566e-06,
+      "loss": 0.0032,
+      "step": 4906
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7968674849961314e-06,
+      "loss": 0.0011,
+      "step": 4907
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7968256623659062e-06,
+      "loss": 0.0029,
+      "step": 4908
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796783839735681e-06,
+      "loss": 2.538,
+      "step": 4909
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7967420171054558e-06,
+      "loss": 1.5312,
+      "step": 4910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7967001944752305e-06,
+      "loss": 1.7333,
+      "step": 4911
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7966583718450053e-06,
+      "loss": 1.146,
+      "step": 4912
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7966165492147801e-06,
+      "loss": 0.1367,
+      "step": 4913
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7965747265845547e-06,
+      "loss": 0.0019,
+      "step": 4914
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7965329039543295e-06,
+      "loss": 0.0021,
+      "step": 4915
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7964910813241043e-06,
+      "loss": 0.0056,
+      "step": 4916
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796449258693879e-06,
+      "loss": 0.0022,
+      "step": 4917
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7964074360636538e-06,
+      "loss": 1.8552,
+      "step": 4918
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7963656134334288e-06,
+      "loss": 1.5811,
+      "step": 4919
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7963237908032036e-06,
+      "loss": 0.0029,
+      "step": 4920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7962819681729784e-06,
+      "loss": 2.5533,
+      "step": 4921
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7962401455427532e-06,
+      "loss": 0.884,
+      "step": 4922
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796198322912528e-06,
+      "loss": 0.0043,
+      "step": 4923
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7961565002823028e-06,
+      "loss": 0.1485,
+      "step": 4924
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7961146776520775e-06,
+      "loss": 0.0015,
+      "step": 4925
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7960728550218523e-06,
+      "loss": 2.4162,
+      "step": 4926
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796031032391627e-06,
+      "loss": 0.0206,
+      "step": 4927
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7959892097614017e-06,
+      "loss": 1.57,
+      "step": 4928
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7959473871311765e-06,
+      "loss": 0.0018,
+      "step": 4929
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7959055645009513e-06,
+      "loss": 0.0022,
+      "step": 4930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795863741870726e-06,
+      "loss": 1.8694,
+      "step": 4931
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7958219192405008e-06,
+      "loss": 1.2672,
+      "step": 4932
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7957800966102756e-06,
+      "loss": 0.3391,
+      "step": 4933
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7957382739800506e-06,
+      "loss": 0.0019,
+      "step": 4934
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7956964513498254e-06,
+      "loss": 0.0012,
+      "step": 4935
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7956546287196002e-06,
+      "loss": 1.9997,
+      "step": 4936
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795612806089375e-06,
+      "loss": 0.316,
+      "step": 4937
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7955709834591498e-06,
+      "loss": 1.8182,
+      "step": 4938
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7955291608289246e-06,
+      "loss": 0.0087,
+      "step": 4939
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7954873381986993e-06,
+      "loss": 1.9818,
+      "step": 4940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795445515568474e-06,
+      "loss": 0.0019,
+      "step": 4941
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7954036929382487e-06,
+      "loss": 1.2795,
+      "step": 4942
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7953618703080235e-06,
+      "loss": 0.0036,
+      "step": 4943
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7953200476777983e-06,
+      "loss": 0.0131,
+      "step": 4944
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795278225047573e-06,
+      "loss": 0.0019,
+      "step": 4945
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7952364024173478e-06,
+      "loss": 0.0041,
+      "step": 4946
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7951945797871226e-06,
+      "loss": 0.0031,
+      "step": 4947
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7951527571568976e-06,
+      "loss": 0.0021,
+      "step": 4948
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7951109345266724e-06,
+      "loss": 0.0175,
+      "step": 4949
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7950691118964472e-06,
+      "loss": 0.0661,
+      "step": 4950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795027289266222e-06,
+      "loss": 0.0923,
+      "step": 4951
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7949854666359968e-06,
+      "loss": 0.0016,
+      "step": 4952
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7949436440057716e-06,
+      "loss": 0.0024,
+      "step": 4953
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7949018213755461e-06,
+      "loss": 0.0216,
+      "step": 4954
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794859998745321e-06,
+      "loss": 0.0017,
+      "step": 4955
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7948181761150957e-06,
+      "loss": 0.0111,
+      "step": 4956
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7947763534848705e-06,
+      "loss": 1.3201,
+      "step": 4957
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7947345308546453e-06,
+      "loss": 0.0044,
+      "step": 4958
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79469270822442e-06,
+      "loss": 0.0107,
+      "step": 4959
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7946508855941949e-06,
+      "loss": 0.0058,
+      "step": 4960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7946090629639696e-06,
+      "loss": 0.0072,
+      "step": 4961
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7945672403337444e-06,
+      "loss": 0.0026,
+      "step": 4962
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7945254177035194e-06,
+      "loss": 0.0041,
+      "step": 4963
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7944835950732942e-06,
+      "loss": 0.0747,
+      "step": 4964
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794441772443069e-06,
+      "loss": 0.0459,
+      "step": 4965
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7943999498128438e-06,
+      "loss": 0.0021,
+      "step": 4966
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7943581271826186e-06,
+      "loss": 1.676,
+      "step": 4967
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7943163045523931e-06,
+      "loss": 0.0197,
+      "step": 4968
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794274481922168e-06,
+      "loss": 1.4367,
+      "step": 4969
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7942326592919427e-06,
+      "loss": 1.8225,
+      "step": 4970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7941908366617175e-06,
+      "loss": 0.7266,
+      "step": 4971
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7941490140314923e-06,
+      "loss": 0.003,
+      "step": 4972
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794107191401267e-06,
+      "loss": 0.0066,
+      "step": 4973
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7940653687710419e-06,
+      "loss": 0.0036,
+      "step": 4974
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7940235461408167e-06,
+      "loss": 0.0081,
+      "step": 4975
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7939817235105914e-06,
+      "loss": 0.0022,
+      "step": 4976
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7939399008803662e-06,
+      "loss": 0.002,
+      "step": 4977
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7938980782501412e-06,
+      "loss": 0.0013,
+      "step": 4978
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793856255619916e-06,
+      "loss": 0.0207,
+      "step": 4979
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7938144329896908e-06,
+      "loss": 0.1281,
+      "step": 4980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7937726103594656e-06,
+      "loss": 0.98,
+      "step": 4981
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7937307877292402e-06,
+      "loss": 0.176,
+      "step": 4982
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793688965099015e-06,
+      "loss": 0.0012,
+      "step": 4983
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7936471424687897e-06,
+      "loss": 0.0025,
+      "step": 4984
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7936053198385645e-06,
+      "loss": 0.0255,
+      "step": 4985
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7935634972083393e-06,
+      "loss": 0.0068,
+      "step": 4986
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793521674578114e-06,
+      "loss": 2.9236,
+      "step": 4987
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7934798519478889e-06,
+      "loss": 0.0034,
+      "step": 4988
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7934380293176637e-06,
+      "loss": 0.0307,
+      "step": 4989
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7933962066874384e-06,
+      "loss": 0.0014,
+      "step": 4990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7933543840572132e-06,
+      "loss": 0.0112,
+      "step": 4991
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7933125614269882e-06,
+      "loss": 0.0023,
+      "step": 4992
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793270738796763e-06,
+      "loss": 0.3346,
+      "step": 4993
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7932289161665378e-06,
+      "loss": 0.1078,
+      "step": 4994
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7931870935363124e-06,
+      "loss": 0.0031,
+      "step": 4995
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7931452709060872e-06,
+      "loss": 0.0056,
+      "step": 4996
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793103448275862e-06,
+      "loss": 0.0019,
+      "step": 4997
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7930616256456367e-06,
+      "loss": 0.0017,
+      "step": 4998
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7930198030154115e-06,
+      "loss": 0.8835,
+      "step": 4999
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7929779803851863e-06,
+      "loss": 0.0012,
+      "step": 5000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792936157754961e-06,
+      "loss": 1.2766,
+      "step": 5001
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7928943351247359e-06,
+      "loss": 0.0172,
+      "step": 5002
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7928525124945107e-06,
+      "loss": 0.0013,
+      "step": 5003
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7928106898642855e-06,
+      "loss": 0.0017,
+      "step": 5004
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7927688672340602e-06,
+      "loss": 0.0021,
+      "step": 5005
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792727044603835e-06,
+      "loss": 1.7973,
+      "step": 5006
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79268522197361e-06,
+      "loss": 0.002,
+      "step": 5007
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7926433993433848e-06,
+      "loss": 0.002,
+      "step": 5008
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7926015767131594e-06,
+      "loss": 0.0021,
+      "step": 5009
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7925597540829342e-06,
+      "loss": 0.0019,
+      "step": 5010
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792517931452709e-06,
+      "loss": 0.0689,
+      "step": 5011
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7924761088224837e-06,
+      "loss": 0.0015,
+      "step": 5012
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7924342861922585e-06,
+      "loss": 0.0011,
+      "step": 5013
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7923924635620333e-06,
+      "loss": 0.0269,
+      "step": 5014
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792350640931808e-06,
+      "loss": 0.0044,
+      "step": 5015
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7923088183015829e-06,
+      "loss": 0.0052,
+      "step": 5016
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7922669956713577e-06,
+      "loss": 0.0023,
+      "step": 5017
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7922251730411325e-06,
+      "loss": 1.1703,
+      "step": 5018
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7921833504109073e-06,
+      "loss": 0.0061,
+      "step": 5019
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792141527780682e-06,
+      "loss": 0.0047,
+      "step": 5020
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792099705150457e-06,
+      "loss": 0.0121,
+      "step": 5021
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7920578825202316e-06,
+      "loss": 0.0032,
+      "step": 5022
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7920160598900064e-06,
+      "loss": 0.0013,
+      "step": 5023
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7919742372597812e-06,
+      "loss": 0.01,
+      "step": 5024
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791932414629556e-06,
+      "loss": 0.0015,
+      "step": 5025
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7918905919993308e-06,
+      "loss": 0.0035,
+      "step": 5026
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7918487693691055e-06,
+      "loss": 0.0043,
+      "step": 5027
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7918069467388803e-06,
+      "loss": 0.0032,
+      "step": 5028
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7917651241086551e-06,
+      "loss": 2.4054,
+      "step": 5029
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.79172330147843e-06,
+      "loss": 0.4131,
+      "step": 5030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7916814788482047e-06,
+      "loss": 0.0043,
+      "step": 5031
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7916396562179795e-06,
+      "loss": 3.9265,
+      "step": 5032
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7915978335877543e-06,
+      "loss": 2.642,
+      "step": 5033
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.791556010957529e-06,
+      "loss": 0.0024,
+      "step": 5034
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7915141883273036e-06,
+      "loss": 0.0021,
+      "step": 5035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7914723656970786e-06,
+      "loss": 0.0022,
+      "step": 5036
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7914305430668534e-06,
+      "loss": 0.001,
+      "step": 5037
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7913887204366282e-06,
+      "loss": 0.0096,
+      "step": 5038
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.791346897806403e-06,
+      "loss": 0.0042,
+      "step": 5039
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7913050751761778e-06,
+      "loss": 0.0033,
+      "step": 5040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7912632525459525e-06,
+      "loss": 0.0019,
+      "step": 5041
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7912214299157273e-06,
+      "loss": 0.0019,
+      "step": 5042
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7911796072855021e-06,
+      "loss": 0.4431,
+      "step": 5043
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.791137784655277e-06,
+      "loss": 0.011,
+      "step": 5044
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7910959620250517e-06,
+      "loss": 0.0047,
+      "step": 5045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7910541393948265e-06,
+      "loss": 1.6669,
+      "step": 5046
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7910123167646013e-06,
+      "loss": 0.0012,
+      "step": 5047
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.790970494134376e-06,
+      "loss": 0.0015,
+      "step": 5048
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7909286715041506e-06,
+      "loss": 0.001,
+      "step": 5049
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7908868488739254e-06,
+      "loss": 0.0012,
+      "step": 5050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7908450262437004e-06,
+      "loss": 1.494,
+      "step": 5051
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7908032036134752e-06,
+      "loss": 0.001,
+      "step": 5052
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.79076138098325e-06,
+      "loss": 0.0016,
+      "step": 5053
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7907195583530248e-06,
+      "loss": 0.002,
+      "step": 5054
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7906777357227996e-06,
+      "loss": 0.0659,
+      "step": 5055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7906359130925743e-06,
+      "loss": 0.0026,
+      "step": 5056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7905940904623491e-06,
+      "loss": 0.002,
+      "step": 5057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.790552267832124e-06,
+      "loss": 0.0178,
+      "step": 5058
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7905104452018987e-06,
+      "loss": 0.3897,
+      "step": 5059
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7904686225716735e-06,
+      "loss": 0.0779,
+      "step": 5060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7904267999414483e-06,
+      "loss": 0.78,
+      "step": 5061
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7903849773112229e-06,
+      "loss": 0.0269,
+      "step": 5062
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7903431546809976e-06,
+      "loss": 0.0026,
+      "step": 5063
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7903013320507724e-06,
+      "loss": 0.0036,
+      "step": 5064
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7902595094205474e-06,
+      "loss": 0.1483,
+      "step": 5065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7902176867903222e-06,
+      "loss": 1.839,
+      "step": 5066
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.790175864160097e-06,
+      "loss": 0.0406,
+      "step": 5067
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7901340415298718e-06,
+      "loss": 0.0017,
+      "step": 5068
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7900922188996466e-06,
+      "loss": 2.4265,
+      "step": 5069
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7900503962694214e-06,
+      "loss": 0.001,
+      "step": 5070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7900085736391961e-06,
+      "loss": 0.0031,
+      "step": 5071
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789966751008971e-06,
+      "loss": 0.0046,
+      "step": 5072
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7899249283787457e-06,
+      "loss": 1.9843,
+      "step": 5073
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7898831057485205e-06,
+      "loss": 0.0017,
+      "step": 5074
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7898412831182953e-06,
+      "loss": 0.0033,
+      "step": 5075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7897994604880699e-06,
+      "loss": 0.0015,
+      "step": 5076
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7897576378578446e-06,
+      "loss": 0.0008,
+      "step": 5077
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7897158152276194e-06,
+      "loss": 0.0009,
+      "step": 5078
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7896739925973942e-06,
+      "loss": 0.0043,
+      "step": 5079
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7896321699671692e-06,
+      "loss": 0.0162,
+      "step": 5080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789590347336944e-06,
+      "loss": 0.0012,
+      "step": 5081
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7895485247067188e-06,
+      "loss": 0.2848,
+      "step": 5082
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7895067020764936e-06,
+      "loss": 0.0025,
+      "step": 5083
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7894648794462684e-06,
+      "loss": 1.0621,
+      "step": 5084
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7894230568160431e-06,
+      "loss": 0.0008,
+      "step": 5085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789381234185818e-06,
+      "loss": 0.0062,
+      "step": 5086
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7893394115555927e-06,
+      "loss": 0.8377,
+      "step": 5087
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7892975889253675e-06,
+      "loss": 1.9849,
+      "step": 5088
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7892557662951423e-06,
+      "loss": 0.0022,
+      "step": 5089
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7892139436649169e-06,
+      "loss": 0.0084,
+      "step": 5090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7891721210346917e-06,
+      "loss": 2.4173,
+      "step": 5091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7891302984044664e-06,
+      "loss": 0.0023,
+      "step": 5092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7890884757742412e-06,
+      "loss": 1.3467,
+      "step": 5093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789046653144016e-06,
+      "loss": 0.0107,
+      "step": 5094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789004830513791e-06,
+      "loss": 0.0012,
+      "step": 5095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7889630078835658e-06,
+      "loss": 1.5685,
+      "step": 5096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7889211852533406e-06,
+      "loss": 0.0142,
+      "step": 5097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7888793626231154e-06,
+      "loss": 0.0047,
+      "step": 5098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7888375399928902e-06,
+      "loss": 0.0045,
+      "step": 5099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788795717362665e-06,
+      "loss": 0.001,
+      "step": 5100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7887538947324397e-06,
+      "loss": 0.0049,
+      "step": 5101
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7887120721022145e-06,
+      "loss": 0.0009,
+      "step": 5102
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788670249471989e-06,
+      "loss": 0.383,
+      "step": 5103
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7886284268417639e-06,
+      "loss": 0.442,
+      "step": 5104
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7885866042115387e-06,
+      "loss": 0.0011,
+      "step": 5105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7885447815813134e-06,
+      "loss": 0.8958,
+      "step": 5106
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7885029589510882e-06,
+      "loss": 0.5264,
+      "step": 5107
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788461136320863e-06,
+      "loss": 1.7591,
+      "step": 5108
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788419313690638e-06,
+      "loss": 0.1435,
+      "step": 5109
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7883774910604128e-06,
+      "loss": 0.0787,
+      "step": 5110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7883356684301876e-06,
+      "loss": 0.0009,
+      "step": 5111
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7882938457999624e-06,
+      "loss": 0.0014,
+      "step": 5112
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7882520231697372e-06,
+      "loss": 0.0058,
+      "step": 5113
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788210200539512e-06,
+      "loss": 0.002,
+      "step": 5114
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7881683779092867e-06,
+      "loss": 0.0019,
+      "step": 5115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7881265552790615e-06,
+      "loss": 0.0321,
+      "step": 5116
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788084732648836e-06,
+      "loss": 0.1851,
+      "step": 5117
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7880429100186109e-06,
+      "loss": 0.1454,
+      "step": 5118
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7880010873883857e-06,
+      "loss": 0.0029,
+      "step": 5119
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7879592647581605e-06,
+      "loss": 0.0026,
+      "step": 5120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7879174421279352e-06,
+      "loss": 0.0018,
+      "step": 5121
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.78787561949771e-06,
+      "loss": 0.3191,
+      "step": 5122
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7878337968674848e-06,
+      "loss": 0.0011,
+      "step": 5123
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7877919742372598e-06,
+      "loss": 0.0036,
+      "step": 5124
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7877501516070346e-06,
+      "loss": 0.0019,
+      "step": 5125
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7877083289768094e-06,
+      "loss": 0.0018,
+      "step": 5126
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7876665063465842e-06,
+      "loss": 0.7257,
+      "step": 5127
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787624683716359e-06,
+      "loss": 0.0007,
+      "step": 5128
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7875828610861337e-06,
+      "loss": 0.0023,
+      "step": 5129
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7875410384559083e-06,
+      "loss": 0.0015,
+      "step": 5130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7874992158256831e-06,
+      "loss": 0.001,
+      "step": 5131
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7874573931954579e-06,
+      "loss": 0.001,
+      "step": 5132
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7874155705652327e-06,
+      "loss": 0.0009,
+      "step": 5133
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7873737479350075e-06,
+      "loss": 0.0013,
+      "step": 5134
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7873319253047823e-06,
+      "loss": 2.7391,
+      "step": 5135
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787290102674557e-06,
+      "loss": 0.4523,
+      "step": 5136
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7872482800443318e-06,
+      "loss": 0.0879,
+      "step": 5137
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7872064574141066e-06,
+      "loss": 0.0026,
+      "step": 5138
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7871646347838816e-06,
+      "loss": 0.4481,
+      "step": 5139
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7871228121536564e-06,
+      "loss": 0.2998,
+      "step": 5140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7870809895234312e-06,
+      "loss": 0.1247,
+      "step": 5141
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787039166893206e-06,
+      "loss": 0.0013,
+      "step": 5142
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7869973442629808e-06,
+      "loss": 0.0018,
+      "step": 5143
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7869555216327553e-06,
+      "loss": 0.0009,
+      "step": 5144
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7869136990025301e-06,
+      "loss": 0.0034,
+      "step": 5145
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786871876372305e-06,
+      "loss": 0.998,
+      "step": 5146
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7868300537420797e-06,
+      "loss": 0.0167,
+      "step": 5147
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7867882311118545e-06,
+      "loss": 0.0041,
+      "step": 5148
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7867464084816293e-06,
+      "loss": 0.0019,
+      "step": 5149
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786704585851404e-06,
+      "loss": 1.7792,
+      "step": 5150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7866627632211788e-06,
+      "loss": 0.0024,
+      "step": 5151
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7866209405909536e-06,
+      "loss": 0.0009,
+      "step": 5152
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7865791179607286e-06,
+      "loss": 0.0008,
+      "step": 5153
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7865372953305034e-06,
+      "loss": 0.0008,
+      "step": 5154
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7864954727002782e-06,
+      "loss": 0.0009,
+      "step": 5155
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786453650070053e-06,
+      "loss": 1.3482,
+      "step": 5156
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7864118274398276e-06,
+      "loss": 0.0047,
+      "step": 5157
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7863700048096023e-06,
+      "loss": 0.0187,
+      "step": 5158
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7863281821793771e-06,
+      "loss": 3.6521,
+      "step": 5159
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786286359549152e-06,
+      "loss": 0.9233,
+      "step": 5160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7862445369189267e-06,
+      "loss": 0.0013,
+      "step": 5161
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7862027142887015e-06,
+      "loss": 0.0273,
+      "step": 5162
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7861608916584763e-06,
+      "loss": 0.0027,
+      "step": 5163
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786119069028251e-06,
+      "loss": 0.0012,
+      "step": 5164
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7860772463980258e-06,
+      "loss": 0.0013,
+      "step": 5165
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7860354237678006e-06,
+      "loss": 0.0031,
+      "step": 5166
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7859936011375754e-06,
+      "loss": 0.0035,
+      "step": 5167
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7859517785073504e-06,
+      "loss": 2.2161,
+      "step": 5168
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7859099558771252e-06,
+      "loss": 0.3511,
+      "step": 5169
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7858681332469e-06,
+      "loss": 0.0135,
+      "step": 5170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7858263106166746e-06,
+      "loss": 0.0526,
+      "step": 5171
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7857844879864493e-06,
+      "loss": 0.0015,
+      "step": 5172
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7857426653562241e-06,
+      "loss": 0.0024,
+      "step": 5173
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785700842725999e-06,
+      "loss": 0.0009,
+      "step": 5174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7856590200957737e-06,
+      "loss": 0.0017,
+      "step": 5175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7856171974655485e-06,
+      "loss": 0.0054,
+      "step": 5176
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7855753748353233e-06,
+      "loss": 0.0012,
+      "step": 5177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785533552205098e-06,
+      "loss": 1.6802,
+      "step": 5178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7854917295748729e-06,
+      "loss": 0.1697,
+      "step": 5179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7854499069446476e-06,
+      "loss": 0.0011,
+      "step": 5180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7854080843144224e-06,
+      "loss": 0.0012,
+      "step": 5181
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7853662616841972e-06,
+      "loss": 0.0014,
+      "step": 5182
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7853244390539722e-06,
+      "loss": 0.0019,
+      "step": 5183
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785282616423747e-06,
+      "loss": 0.0009,
+      "step": 5184
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7852407937935216e-06,
+      "loss": 2.0156,
+      "step": 5185
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7851989711632964e-06,
+      "loss": 0.0014,
+      "step": 5186
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7851571485330711e-06,
+      "loss": 0.0221,
+      "step": 5187
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.785115325902846e-06,
+      "loss": 0.009,
+      "step": 5188
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7850735032726207e-06,
+      "loss": 0.0008,
+      "step": 5189
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7850316806423955e-06,
+      "loss": 0.0032,
+      "step": 5190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7849898580121703e-06,
+      "loss": 0.0019,
+      "step": 5191
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.784948035381945e-06,
+      "loss": 3.5971,
+      "step": 5192
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7849062127517199e-06,
+      "loss": 0.001,
+      "step": 5193
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7848643901214946e-06,
+      "loss": 0.092,
+      "step": 5194
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7848225674912694e-06,
+      "loss": 0.353,
+      "step": 5195
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7847807448610442e-06,
+      "loss": 0.001,
+      "step": 5196
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7847389222308192e-06,
+      "loss": 0.0017,
+      "step": 5197
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7846970996005938e-06,
+      "loss": 0.5594,
+      "step": 5198
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7846552769703686e-06,
+      "loss": 0.0018,
+      "step": 5199
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7846134543401434e-06,
+      "loss": 0.0029,
+      "step": 5200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7845716317099182e-06,
+      "loss": 1.7061,
+      "step": 5201
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.784529809079693e-06,
+      "loss": 0.001,
+      "step": 5202
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7844879864494677e-06,
+      "loss": 0.0019,
+      "step": 5203
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7844461638192425e-06,
+      "loss": 0.0049,
+      "step": 5204
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7844043411890173e-06,
+      "loss": 0.0022,
+      "step": 5205
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.784362518558792e-06,
+      "loss": 0.2377,
+      "step": 5206
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7843206959285669e-06,
+      "loss": 0.0008,
+      "step": 5207
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7842788732983417e-06,
+      "loss": 0.0345,
+      "step": 5208
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7842370506681164e-06,
+      "loss": 0.048,
+      "step": 5209
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7841952280378912e-06,
+      "loss": 1.0287,
+      "step": 5210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7841534054076658e-06,
+      "loss": 0.4052,
+      "step": 5211
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7841115827774408e-06,
+      "loss": 0.0012,
+      "step": 5212
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7840697601472156e-06,
+      "loss": 0.2305,
+      "step": 5213
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7840279375169904e-06,
+      "loss": 0.0011,
+      "step": 5214
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7839861148867652e-06,
+      "loss": 0.01,
+      "step": 5215
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.78394429225654e-06,
+      "loss": 0.0091,
+      "step": 5216
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7839024696263147e-06,
+      "loss": 0.0011,
+      "step": 5217
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7838606469960895e-06,
+      "loss": 0.3933,
+      "step": 5218
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7838188243658643e-06,
+      "loss": 0.001,
+      "step": 5219
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.783777001735639e-06,
+      "loss": 0.0024,
+      "step": 5220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7837351791054139e-06,
+      "loss": 1.3784,
+      "step": 5221
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7836933564751887e-06,
+      "loss": 0.0019,
+      "step": 5222
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7836515338449634e-06,
+      "loss": 1.4351,
+      "step": 5223
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7836097112147382e-06,
+      "loss": 2.285,
+      "step": 5224
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7835678885845128e-06,
+      "loss": 0.0026,
+      "step": 5225
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7835260659542878e-06,
+      "loss": 0.0069,
+      "step": 5226
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7834842433240626e-06,
+      "loss": 0.0015,
+      "step": 5227
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7834424206938374e-06,
+      "loss": 1.0173,
+      "step": 5228
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7834005980636122e-06,
+      "loss": 0.0016,
+      "step": 5229
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.783358775433387e-06,
+      "loss": 1.488,
+      "step": 5230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7833169528031617e-06,
+      "loss": 0.001,
+      "step": 5231
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7832751301729365e-06,
+      "loss": 3.5006,
+      "step": 5232
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7832333075427113e-06,
+      "loss": 0.4137,
+      "step": 5233
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.783191484912486e-06,
+      "loss": 0.0213,
+      "step": 5234
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7831496622822609e-06,
+      "loss": 1.2393,
+      "step": 5235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7831078396520357e-06,
+      "loss": 0.0011,
+      "step": 5236
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7830660170218105e-06,
+      "loss": 0.1145,
+      "step": 5237
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.783024194391585e-06,
+      "loss": 0.0992,
+      "step": 5238
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7829823717613598e-06,
+      "loss": 0.003,
+      "step": 5239
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7829405491311346e-06,
+      "loss": 0.0015,
+      "step": 5240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7828987265009096e-06,
+      "loss": 0.0015,
+      "step": 5241
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7828569038706844e-06,
+      "loss": 0.0165,
+      "step": 5242
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7828150812404592e-06,
+      "loss": 1.4737,
+      "step": 5243
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782773258610234e-06,
+      "loss": 0.0012,
+      "step": 5244
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7827314359800087e-06,
+      "loss": 0.0113,
+      "step": 5245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7826896133497835e-06,
+      "loss": 0.0097,
+      "step": 5246
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7826477907195583e-06,
+      "loss": 0.0047,
+      "step": 5247
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7826059680893331e-06,
+      "loss": 0.0037,
+      "step": 5248
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782564145459108e-06,
+      "loss": 1.8126,
+      "step": 5249
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7825223228288827e-06,
+      "loss": 2.5416,
+      "step": 5250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7824805001986575e-06,
+      "loss": 0.723,
+      "step": 5251
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782438677568432e-06,
+      "loss": 0.0017,
+      "step": 5252
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7823968549382068e-06,
+      "loss": 0.0225,
+      "step": 5253
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7823550323079816e-06,
+      "loss": 0.0011,
+      "step": 5254
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7823132096777564e-06,
+      "loss": 0.598,
+      "step": 5255
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7822713870475314e-06,
+      "loss": 0.007,
+      "step": 5256
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7822295644173062e-06,
+      "loss": 0.0025,
+      "step": 5257
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782187741787081e-06,
+      "loss": 0.0056,
+      "step": 5258
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7821459191568558e-06,
+      "loss": 0.0022,
+      "step": 5259
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7821040965266305e-06,
+      "loss": 0.0012,
+      "step": 5260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7820622738964053e-06,
+      "loss": 0.4572,
+      "step": 5261
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7820204512661801e-06,
+      "loss": 0.002,
+      "step": 5262
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781978628635955e-06,
+      "loss": 0.0154,
+      "step": 5263
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7819368060057297e-06,
+      "loss": 0.0027,
+      "step": 5264
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7818949833755043e-06,
+      "loss": 0.0009,
+      "step": 5265
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781853160745279e-06,
+      "loss": 0.0123,
+      "step": 5266
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7818113381150538e-06,
+      "loss": 1.7543,
+      "step": 5267
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7817695154848286e-06,
+      "loss": 0.0048,
+      "step": 5268
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7817276928546034e-06,
+      "loss": 0.0014,
+      "step": 5269
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7816858702243784e-06,
+      "loss": 3.142,
+      "step": 5270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7816440475941532e-06,
+      "loss": 0.0206,
+      "step": 5271
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781602224963928e-06,
+      "loss": 3.2275,
+      "step": 5272
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7815604023337028e-06,
+      "loss": 0.0017,
+      "step": 5273
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7815185797034776e-06,
+      "loss": 0.0019,
+      "step": 5274
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7814767570732523e-06,
+      "loss": 0.0065,
+      "step": 5275
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7814349344430271e-06,
+      "loss": 0.0017,
+      "step": 5276
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781393111812802e-06,
+      "loss": 0.6963,
+      "step": 5277
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7813512891825767e-06,
+      "loss": 0.0019,
+      "step": 5278
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7813094665523513e-06,
+      "loss": 0.8272,
+      "step": 5279
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781267643922126e-06,
+      "loss": 0.0017,
+      "step": 5280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7812258212919008e-06,
+      "loss": 1.1904,
+      "step": 5281
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7811839986616756e-06,
+      "loss": 0.0014,
+      "step": 5282
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7811421760314504e-06,
+      "loss": 0.0027,
+      "step": 5283
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7811003534012252e-06,
+      "loss": 0.0028,
+      "step": 5284
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7810585307710002e-06,
+      "loss": 0.0011,
+      "step": 5285
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781016708140775e-06,
+      "loss": 0.0053,
+      "step": 5286
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7809748855105498e-06,
+      "loss": 0.0015,
+      "step": 5287
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7809330628803246e-06,
+      "loss": 0.0125,
+      "step": 5288
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7808912402500993e-06,
+      "loss": 0.6085,
+      "step": 5289
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7808494176198741e-06,
+      "loss": 0.0015,
+      "step": 5290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780807594989649e-06,
+      "loss": 1.7069,
+      "step": 5291
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7807657723594237e-06,
+      "loss": 0.0012,
+      "step": 5292
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7807239497291983e-06,
+      "loss": 0.0043,
+      "step": 5293
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780682127098973e-06,
+      "loss": 0.0018,
+      "step": 5294
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7806403044687479e-06,
+      "loss": 0.0062,
+      "step": 5295
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7805984818385226e-06,
+      "loss": 0.0712,
+      "step": 5296
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7805566592082974e-06,
+      "loss": 0.0955,
+      "step": 5297
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7805148365780722e-06,
+      "loss": 0.1895,
+      "step": 5298
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780473013947847e-06,
+      "loss": 0.0053,
+      "step": 5299
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780431191317622e-06,
+      "loss": 0.0014,
+      "step": 5300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7803893686873968e-06,
+      "loss": 0.0035,
+      "step": 5301
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7803475460571716e-06,
+      "loss": 0.0021,
+      "step": 5302
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7803057234269464e-06,
+      "loss": 0.0143,
+      "step": 5303
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7802639007967211e-06,
+      "loss": 0.003,
+      "step": 5304
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780222078166496e-06,
+      "loss": 0.4608,
+      "step": 5305
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7801802555362705e-06,
+      "loss": 0.0008,
+      "step": 5306
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7801384329060453e-06,
+      "loss": 0.3537,
+      "step": 5307
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.78009661027582e-06,
+      "loss": 0.0013,
+      "step": 5308
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7800547876455949e-06,
+      "loss": 0.0046,
+      "step": 5309
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7800129650153696e-06,
+      "loss": 0.001,
+      "step": 5310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7799711423851444e-06,
+      "loss": 0.0009,
+      "step": 5311
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7799293197549192e-06,
+      "loss": 0.0322,
+      "step": 5312
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779887497124694e-06,
+      "loss": 0.004,
+      "step": 5313
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779845674494469e-06,
+      "loss": 0.0034,
+      "step": 5314
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7798038518642438e-06,
+      "loss": 0.0242,
+      "step": 5315
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7797620292340186e-06,
+      "loss": 0.0008,
+      "step": 5316
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7797202066037934e-06,
+      "loss": 0.0007,
+      "step": 5317
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7796783839735682e-06,
+      "loss": 0.0007,
+      "step": 5318
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779636561343343e-06,
+      "loss": 1.8926,
+      "step": 5319
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7795947387131175e-06,
+      "loss": 0.0034,
+      "step": 5320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7795529160828923e-06,
+      "loss": 0.0015,
+      "step": 5321
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779511093452667e-06,
+      "loss": 0.0012,
+      "step": 5322
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7794692708224419e-06,
+      "loss": 0.0011,
+      "step": 5323
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7794274481922167e-06,
+      "loss": 2.166,
+      "step": 5324
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7793856255619914e-06,
+      "loss": 0.0013,
+      "step": 5325
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7793438029317662e-06,
+      "loss": 0.0028,
+      "step": 5326
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779301980301541e-06,
+      "loss": 0.0015,
+      "step": 5327
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7792601576713158e-06,
+      "loss": 0.001,
+      "step": 5328
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7792183350410908e-06,
+      "loss": 0.0032,
+      "step": 5329
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7791765124108656e-06,
+      "loss": 0.0012,
+      "step": 5330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7791346897806404e-06,
+      "loss": 0.0009,
+      "step": 5331
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7790928671504152e-06,
+      "loss": 0.0011,
+      "step": 5332
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7790510445201897e-06,
+      "loss": 0.0013,
+      "step": 5333
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7790092218899645e-06,
+      "loss": 0.0009,
+      "step": 5334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7789673992597393e-06,
+      "loss": 0.5335,
+      "step": 5335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778925576629514e-06,
+      "loss": 0.0008,
+      "step": 5336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7788837539992889e-06,
+      "loss": 0.0015,
+      "step": 5337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7788419313690637e-06,
+      "loss": 0.0009,
+      "step": 5338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7788001087388385e-06,
+      "loss": 0.0012,
+      "step": 5339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7787582861086132e-06,
+      "loss": 0.0063,
+      "step": 5340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778716463478388e-06,
+      "loss": 0.2645,
+      "step": 5341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7786746408481628e-06,
+      "loss": 0.0011,
+      "step": 5342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7786328182179376e-06,
+      "loss": 2.1671,
+      "step": 5343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7785909955877126e-06,
+      "loss": 0.003,
+      "step": 5344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7785491729574874e-06,
+      "loss": 0.0021,
+      "step": 5345
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7785073503272622e-06,
+      "loss": 0.0009,
+      "step": 5346
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7784655276970367e-06,
+      "loss": 0.0044,
+      "step": 5347
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7784237050668115e-06,
+      "loss": 0.0026,
+      "step": 5348
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7783818824365863e-06,
+      "loss": 0.0306,
+      "step": 5349
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.778340059806361e-06,
+      "loss": 0.0016,
+      "step": 5350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7782982371761359e-06,
+      "loss": 0.6002,
+      "step": 5351
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7782564145459107e-06,
+      "loss": 0.0011,
+      "step": 5352
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7782145919156855e-06,
+      "loss": 0.0104,
+      "step": 5353
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7781727692854602e-06,
+      "loss": 0.0006,
+      "step": 5354
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.778130946655235e-06,
+      "loss": 0.0028,
+      "step": 5355
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7780891240250098e-06,
+      "loss": 0.0022,
+      "step": 5356
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7780473013947846e-06,
+      "loss": 0.008,
+      "step": 5357
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7780054787645596e-06,
+      "loss": 0.0013,
+      "step": 5358
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7779636561343344e-06,
+      "loss": 0.2597,
+      "step": 5359
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.777921833504109e-06,
+      "loss": 0.0033,
+      "step": 5360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7778800108738838e-06,
+      "loss": 0.0067,
+      "step": 5361
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7778381882436585e-06,
+      "loss": 0.0021,
+      "step": 5362
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7777963656134333e-06,
+      "loss": 0.0065,
+      "step": 5363
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7777545429832081e-06,
+      "loss": 0.0008,
+      "step": 5364
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.777712720352983e-06,
+      "loss": 0.0007,
+      "step": 5365
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7776708977227577e-06,
+      "loss": 0.0206,
+      "step": 5366
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7776290750925325e-06,
+      "loss": 0.0013,
+      "step": 5367
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7775872524623073e-06,
+      "loss": 0.0008,
+      "step": 5368
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.777545429832082e-06,
+      "loss": 0.8123,
+      "step": 5369
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7775036072018568e-06,
+      "loss": 0.0079,
+      "step": 5370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7774617845716316e-06,
+      "loss": 0.1233,
+      "step": 5371
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7774199619414064e-06,
+      "loss": 0.0008,
+      "step": 5372
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7773781393111814e-06,
+      "loss": 0.0824,
+      "step": 5373
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.777336316680956e-06,
+      "loss": 0.024,
+      "step": 5374
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7772944940507308e-06,
+      "loss": 0.0038,
+      "step": 5375
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7772526714205055e-06,
+      "loss": 0.0012,
+      "step": 5376
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7772108487902803e-06,
+      "loss": 0.0138,
+      "step": 5377
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7771690261600551e-06,
+      "loss": 0.0007,
+      "step": 5378
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77712720352983e-06,
+      "loss": 0.0033,
+      "step": 5379
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7770853808996047e-06,
+      "loss": 1.266,
+      "step": 5380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7770435582693795e-06,
+      "loss": 0.0018,
+      "step": 5381
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7770017356391543e-06,
+      "loss": 0.0011,
+      "step": 5382
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776959913008929e-06,
+      "loss": 0.0013,
+      "step": 5383
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7769180903787038e-06,
+      "loss": 0.0008,
+      "step": 5384
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7768762677484786e-06,
+      "loss": 0.0038,
+      "step": 5385
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7768344451182534e-06,
+      "loss": 0.0023,
+      "step": 5386
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776792622488028e-06,
+      "loss": 0.0009,
+      "step": 5387
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776750799857803e-06,
+      "loss": 0.0008,
+      "step": 5388
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7767089772275778e-06,
+      "loss": 1.1841,
+      "step": 5389
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7766671545973526e-06,
+      "loss": 0.0007,
+      "step": 5390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7766253319671273e-06,
+      "loss": 0.0023,
+      "step": 5391
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7765835093369021e-06,
+      "loss": 0.0009,
+      "step": 5392
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776541686706677e-06,
+      "loss": 0.0032,
+      "step": 5393
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7764998640764517e-06,
+      "loss": 1.6881,
+      "step": 5394
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7764580414462265e-06,
+      "loss": 0.0015,
+      "step": 5395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7764162188160013e-06,
+      "loss": 0.0007,
+      "step": 5396
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776374396185776e-06,
+      "loss": 0.2987,
+      "step": 5397
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7763325735555508e-06,
+      "loss": 0.1734,
+      "step": 5398
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7762907509253256e-06,
+      "loss": 1.6226,
+      "step": 5399
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7762489282951004e-06,
+      "loss": 0.0017,
+      "step": 5400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776207105664875e-06,
+      "loss": 0.0222,
+      "step": 5401
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77616528303465e-06,
+      "loss": 0.1538,
+      "step": 5402
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7761234604044248e-06,
+      "loss": 0.0009,
+      "step": 5403
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7760816377741996e-06,
+      "loss": 0.0301,
+      "step": 5404
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7760398151439744e-06,
+      "loss": 0.0014,
+      "step": 5405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7759979925137491e-06,
+      "loss": 0.5759,
+      "step": 5406
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775956169883524e-06,
+      "loss": 0.001,
+      "step": 5407
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7759143472532987e-06,
+      "loss": 0.3364,
+      "step": 5408
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7758725246230735e-06,
+      "loss": 0.0009,
+      "step": 5409
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7758307019928483e-06,
+      "loss": 0.0025,
+      "step": 5410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775788879362623e-06,
+      "loss": 3.8492,
+      "step": 5411
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7757470567323979e-06,
+      "loss": 0.0013,
+      "step": 5412
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7757052341021726e-06,
+      "loss": 0.0016,
+      "step": 5413
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7756634114719472e-06,
+      "loss": 0.0011,
+      "step": 5414
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775621588841722e-06,
+      "loss": 0.0011,
+      "step": 5415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7755797662114968e-06,
+      "loss": 0.0009,
+      "step": 5416
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7755379435812718e-06,
+      "loss": 0.1603,
+      "step": 5417
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7754961209510466e-06,
+      "loss": 0.0012,
+      "step": 5418
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7754542983208214e-06,
+      "loss": 0.0007,
+      "step": 5419
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7754124756905961e-06,
+      "loss": 0.0009,
+      "step": 5420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775370653060371e-06,
+      "loss": 0.0008,
+      "step": 5421
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7753288304301457e-06,
+      "loss": 0.0287,
+      "step": 5422
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7752870077999205e-06,
+      "loss": 0.0011,
+      "step": 5423
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7752451851696953e-06,
+      "loss": 0.0007,
+      "step": 5424
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77520336253947e-06,
+      "loss": 0.0008,
+      "step": 5425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7751615399092449e-06,
+      "loss": 1.8507,
+      "step": 5426
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7751197172790196e-06,
+      "loss": 0.0008,
+      "step": 5427
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7750778946487942e-06,
+      "loss": 0.0013,
+      "step": 5428
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775036072018569e-06,
+      "loss": 0.001,
+      "step": 5429
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7749942493883438e-06,
+      "loss": 0.0007,
+      "step": 5430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7749524267581188e-06,
+      "loss": 0.001,
+      "step": 5431
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7749106041278936e-06,
+      "loss": 1.5625,
+      "step": 5432
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7748687814976684e-06,
+      "loss": 0.0013,
+      "step": 5433
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7748269588674432e-06,
+      "loss": 0.5059,
+      "step": 5434
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774785136237218e-06,
+      "loss": 0.0069,
+      "step": 5435
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7747433136069927e-06,
+      "loss": 0.845,
+      "step": 5436
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7747014909767675e-06,
+      "loss": 0.0019,
+      "step": 5437
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7746596683465423e-06,
+      "loss": 0.0013,
+      "step": 5438
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774617845716317e-06,
+      "loss": 0.0012,
+      "step": 5439
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7745760230860919e-06,
+      "loss": 0.0024,
+      "step": 5440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7745342004558664e-06,
+      "loss": 0.2146,
+      "step": 5441
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7744923778256412e-06,
+      "loss": 0.0007,
+      "step": 5442
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774450555195416e-06,
+      "loss": 0.0016,
+      "step": 5443
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7744087325651908e-06,
+      "loss": 3.6413,
+      "step": 5444
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7743669099349656e-06,
+      "loss": 0.8761,
+      "step": 5445
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7743250873047406e-06,
+      "loss": 0.0018,
+      "step": 5446
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7742832646745154e-06,
+      "loss": 0.005,
+      "step": 5447
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7742414420442902e-06,
+      "loss": 0.1193,
+      "step": 5448
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774199619414065e-06,
+      "loss": 1.5555,
+      "step": 5449
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7741577967838397e-06,
+      "loss": 0.0036,
+      "step": 5450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7741159741536145e-06,
+      "loss": 0.751,
+      "step": 5451
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7740741515233893e-06,
+      "loss": 0.0008,
+      "step": 5452
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774032328893164e-06,
+      "loss": 0.005,
+      "step": 5453
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7739905062629389e-06,
+      "loss": 1.7986,
+      "step": 5454
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7739486836327135e-06,
+      "loss": 0.0102,
+      "step": 5455
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7739068610024882e-06,
+      "loss": 0.0015,
+      "step": 5456
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773865038372263e-06,
+      "loss": 0.4508,
+      "step": 5457
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7738232157420378e-06,
+      "loss": 0.0024,
+      "step": 5458
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7737813931118126e-06,
+      "loss": 1.0347,
+      "step": 5459
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7737395704815874e-06,
+      "loss": 0.0013,
+      "step": 5460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7736977478513624e-06,
+      "loss": 0.0012,
+      "step": 5461
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7736559252211372e-06,
+      "loss": 0.0008,
+      "step": 5462
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773614102590912e-06,
+      "loss": 0.0009,
+      "step": 5463
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7735722799606867e-06,
+      "loss": 0.0014,
+      "step": 5464
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7735304573304615e-06,
+      "loss": 0.0007,
+      "step": 5465
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7734886347002363e-06,
+      "loss": 2.4016,
+      "step": 5466
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773446812070011e-06,
+      "loss": 0.001,
+      "step": 5467
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7734049894397857e-06,
+      "loss": 0.0122,
+      "step": 5468
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7733631668095605e-06,
+      "loss": 0.0007,
+      "step": 5469
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7733213441793352e-06,
+      "loss": 0.0137,
+      "step": 5470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77327952154911e-06,
+      "loss": 0.0016,
+      "step": 5471
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7732376989188848e-06,
+      "loss": 0.0022,
+      "step": 5472
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7731958762886596e-06,
+      "loss": 0.4425,
+      "step": 5473
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7731540536584344e-06,
+      "loss": 0.0006,
+      "step": 5474
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7731122310282094e-06,
+      "loss": 0.0016,
+      "step": 5475
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7730704083979842e-06,
+      "loss": 0.5149,
+      "step": 5476
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773028585767759e-06,
+      "loss": 0.0032,
+      "step": 5477
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7729867631375338e-06,
+      "loss": 0.001,
+      "step": 5478
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7729449405073085e-06,
+      "loss": 0.0011,
+      "step": 5479
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7729031178770833e-06,
+      "loss": 0.0044,
+      "step": 5480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7728612952468581e-06,
+      "loss": 0.0047,
+      "step": 5481
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7728194726166327e-06,
+      "loss": 0.0727,
+      "step": 5482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7727776499864075e-06,
+      "loss": 0.0008,
+      "step": 5483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7727358273561823e-06,
+      "loss": 0.001,
+      "step": 5484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772694004725957e-06,
+      "loss": 0.0007,
+      "step": 5485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7726521820957318e-06,
+      "loss": 0.0009,
+      "step": 5486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7726103594655066e-06,
+      "loss": 0.0823,
+      "step": 5487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7725685368352814e-06,
+      "loss": 0.3868,
+      "step": 5488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7725267142050562e-06,
+      "loss": 0.0007,
+      "step": 5489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7724848915748312e-06,
+      "loss": 0.0021,
+      "step": 5490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772443068944606e-06,
+      "loss": 0.004,
+      "step": 5491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7724012463143808e-06,
+      "loss": 0.001,
+      "step": 5492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7723594236841555e-06,
+      "loss": 0.0006,
+      "step": 5493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7723176010539303e-06,
+      "loss": 1.6169,
+      "step": 5494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7722757784237051e-06,
+      "loss": 0.0765,
+      "step": 5495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7722339557934797e-06,
+      "loss": 0.0011,
+      "step": 5496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7721921331632545e-06,
+      "loss": 0.0007,
+      "step": 5497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7721503105330293e-06,
+      "loss": 0.0054,
+      "step": 5498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772108487902804e-06,
+      "loss": 0.0007,
+      "step": 5499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7720666652725788e-06,
+      "loss": 0.0013,
+      "step": 5500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7720248426423536e-06,
+      "loss": 2.8197,
+      "step": 5501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7719830200121284e-06,
+      "loss": 0.0009,
+      "step": 5502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7719411973819032e-06,
+      "loss": 2.0886,
+      "step": 5503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771899374751678e-06,
+      "loss": 0.0068,
+      "step": 5504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771857552121453e-06,
+      "loss": 0.8174,
+      "step": 5505
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7718157294912278e-06,
+      "loss": 0.0011,
+      "step": 5506
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7717739068610026e-06,
+      "loss": 0.0027,
+      "step": 5507
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7717320842307773e-06,
+      "loss": 0.0013,
+      "step": 5508
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.771690261600552e-06,
+      "loss": 0.0009,
+      "step": 5509
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7716484389703267e-06,
+      "loss": 0.0008,
+      "step": 5510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7716066163401015e-06,
+      "loss": 0.0046,
+      "step": 5511
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7715647937098763e-06,
+      "loss": 0.0014,
+      "step": 5512
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.771522971079651e-06,
+      "loss": 0.0009,
+      "step": 5513
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7714811484494258e-06,
+      "loss": 0.0026,
+      "step": 5514
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7714393258192006e-06,
+      "loss": 3.3621,
+      "step": 5515
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7713975031889754e-06,
+      "loss": 0.0039,
+      "step": 5516
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7713556805587502e-06,
+      "loss": 0.001,
+      "step": 5517
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.771313857928525e-06,
+      "loss": 3.567,
+      "step": 5518
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7712720352983e-06,
+      "loss": 0.002,
+      "step": 5519
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7712302126680748e-06,
+      "loss": 0.001,
+      "step": 5520
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7711883900378496e-06,
+      "loss": 0.0007,
+      "step": 5521
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7711465674076244e-06,
+      "loss": 0.0008,
+      "step": 5522
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.771104744777399e-06,
+      "loss": 0.004,
+      "step": 5523
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7710629221471737e-06,
+      "loss": 0.0011,
+      "step": 5524
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7710210995169485e-06,
+      "loss": 0.0009,
+      "step": 5525
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7709792768867233e-06,
+      "loss": 0.0013,
+      "step": 5526
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.770937454256498e-06,
+      "loss": 0.001,
+      "step": 5527
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7708956316262729e-06,
+      "loss": 0.0189,
+      "step": 5528
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7708538089960476e-06,
+      "loss": 0.0023,
+      "step": 5529
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7708119863658224e-06,
+      "loss": 0.0114,
+      "step": 5530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7707701637355972e-06,
+      "loss": 0.001,
+      "step": 5531
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.770728341105372e-06,
+      "loss": 0.1648,
+      "step": 5532
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7706865184751468e-06,
+      "loss": 0.0006,
+      "step": 5533
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7706446958449218e-06,
+      "loss": 0.0014,
+      "step": 5534
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7706028732146966e-06,
+      "loss": 0.0016,
+      "step": 5535
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7705610505844711e-06,
+      "loss": 0.0281,
+      "step": 5536
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.770519227954246e-06,
+      "loss": 0.0015,
+      "step": 5537
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7704774053240207e-06,
+      "loss": 0.0014,
+      "step": 5538
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7704355826937955e-06,
+      "loss": 0.0007,
+      "step": 5539
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7703937600635703e-06,
+      "loss": 0.0041,
+      "step": 5540
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.770351937433345e-06,
+      "loss": 0.0016,
+      "step": 5541
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7703101148031199e-06,
+      "loss": 0.001,
+      "step": 5542
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7702682921728947e-06,
+      "loss": 0.0066,
+      "step": 5543
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7702264695426694e-06,
+      "loss": 0.3061,
+      "step": 5544
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7701846469124442e-06,
+      "loss": 0.0006,
+      "step": 5545
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.770142824282219e-06,
+      "loss": 0.0009,
+      "step": 5546
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7701010016519938e-06,
+      "loss": 0.0007,
+      "step": 5547
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7700591790217686e-06,
+      "loss": 2.5691,
+      "step": 5548
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7700173563915436e-06,
+      "loss": 0.0134,
+      "step": 5549
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7699755337613182e-06,
+      "loss": 1.743,
+      "step": 5550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769933711131093e-06,
+      "loss": 3.5577,
+      "step": 5551
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7698918885008677e-06,
+      "loss": 0.0013,
+      "step": 5552
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7698500658706425e-06,
+      "loss": 0.0081,
+      "step": 5553
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7698082432404173e-06,
+      "loss": 0.0007,
+      "step": 5554
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769766420610192e-06,
+      "loss": 0.0008,
+      "step": 5555
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7697245979799669e-06,
+      "loss": 3.6518,
+      "step": 5556
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7696827753497417e-06,
+      "loss": 0.0007,
+      "step": 5557
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7696409527195164e-06,
+      "loss": 0.001,
+      "step": 5558
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7695991300892912e-06,
+      "loss": 1.6936,
+      "step": 5559
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769557307459066e-06,
+      "loss": 0.0017,
+      "step": 5560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7695154848288408e-06,
+      "loss": 0.04,
+      "step": 5561
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7694736621986156e-06,
+      "loss": 0.0007,
+      "step": 5562
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7694318395683904e-06,
+      "loss": 0.0011,
+      "step": 5563
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7693900169381652e-06,
+      "loss": 0.0059,
+      "step": 5564
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76934819430794e-06,
+      "loss": 1.4209,
+      "step": 5565
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7693063716777147e-06,
+      "loss": 0.0074,
+      "step": 5566
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7692645490474895e-06,
+      "loss": 0.0055,
+      "step": 5567
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7692227264172643e-06,
+      "loss": 0.6663,
+      "step": 5568
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769180903787039e-06,
+      "loss": 0.0034,
+      "step": 5569
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7691390811568139e-06,
+      "loss": 2.0583,
+      "step": 5570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7690972585265887e-06,
+      "loss": 4.1745,
+      "step": 5571
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7690554358963635e-06,
+      "loss": 0.0008,
+      "step": 5572
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7690136132661382e-06,
+      "loss": 0.0015,
+      "step": 5573
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768971790635913e-06,
+      "loss": 0.0006,
+      "step": 5574
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7689299680056878e-06,
+      "loss": 2.449,
+      "step": 5575
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7688881453754624e-06,
+      "loss": 0.0006,
+      "step": 5576
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7688463227452372e-06,
+      "loss": 0.0009,
+      "step": 5577
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7688045001150122e-06,
+      "loss": 0.001,
+      "step": 5578
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768762677484787e-06,
+      "loss": 0.0594,
+      "step": 5579
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7687208548545617e-06,
+      "loss": 0.0053,
+      "step": 5580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7686790322243365e-06,
+      "loss": 0.4036,
+      "step": 5581
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7686372095941113e-06,
+      "loss": 0.0009,
+      "step": 5582
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768595386963886e-06,
+      "loss": 0.0008,
+      "step": 5583
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7685535643336609e-06,
+      "loss": 1.491,
+      "step": 5584
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7685117417034357e-06,
+      "loss": 0.0461,
+      "step": 5585
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7684699190732105e-06,
+      "loss": 0.001,
+      "step": 5586
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7684280964429853e-06,
+      "loss": 0.0019,
+      "step": 5587
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76838627381276e-06,
+      "loss": 0.0009,
+      "step": 5588
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7683444511825348e-06,
+      "loss": 0.5403,
+      "step": 5589
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7683026285523094e-06,
+      "loss": 0.6452,
+      "step": 5590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7682608059220842e-06,
+      "loss": 0.0012,
+      "step": 5591
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768218983291859e-06,
+      "loss": 0.0464,
+      "step": 5592
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768177160661634e-06,
+      "loss": 0.0029,
+      "step": 5593
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7681353380314088e-06,
+      "loss": 0.0012,
+      "step": 5594
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7680935154011835e-06,
+      "loss": 2.6488,
+      "step": 5595
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7680516927709583e-06,
+      "loss": 0.4247,
+      "step": 5596
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7680098701407331e-06,
+      "loss": 0.0012,
+      "step": 5597
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767968047510508e-06,
+      "loss": 0.0015,
+      "step": 5598
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7679262248802827e-06,
+      "loss": 0.0012,
+      "step": 5599
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7678844022500575e-06,
+      "loss": 0.0035,
+      "step": 5600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7678425796198323e-06,
+      "loss": 0.0011,
+      "step": 5601
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767800756989607e-06,
+      "loss": 0.0029,
+      "step": 5602
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7677589343593818e-06,
+      "loss": 0.0243,
+      "step": 5603
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7677171117291564e-06,
+      "loss": 0.0011,
+      "step": 5604
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7676752890989312e-06,
+      "loss": 0.0011,
+      "step": 5605
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767633466468706e-06,
+      "loss": 0.0124,
+      "step": 5606
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767591643838481e-06,
+      "loss": 0.0012,
+      "step": 5607
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7675498212082558e-06,
+      "loss": 0.0013,
+      "step": 5608
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7675079985780305e-06,
+      "loss": 0.0011,
+      "step": 5609
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7674661759478053e-06,
+      "loss": 0.0017,
+      "step": 5610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7674243533175801e-06,
+      "loss": 1.5826,
+      "step": 5611
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767382530687355e-06,
+      "loss": 0.3469,
+      "step": 5612
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7673407080571297e-06,
+      "loss": 3.1421,
+      "step": 5613
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7672988854269045e-06,
+      "loss": 3.3526,
+      "step": 5614
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7672570627966793e-06,
+      "loss": 0.2116,
+      "step": 5615
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767215240166454e-06,
+      "loss": 0.0071,
+      "step": 5616
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7671734175362286e-06,
+      "loss": 2.7058,
+      "step": 5617
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7671315949060034e-06,
+      "loss": 0.0011,
+      "step": 5618
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7670897722757782e-06,
+      "loss": 0.0007,
+      "step": 5619
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767047949645553e-06,
+      "loss": 0.001,
+      "step": 5620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7670061270153278e-06,
+      "loss": 0.0011,
+      "step": 5621
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7669643043851028e-06,
+      "loss": 0.0061,
+      "step": 5622
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7669224817548776e-06,
+      "loss": 0.001,
+      "step": 5623
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7668806591246523e-06,
+      "loss": 0.0019,
+      "step": 5624
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7668388364944271e-06,
+      "loss": 0.0905,
+      "step": 5625
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766797013864202e-06,
+      "loss": 0.0023,
+      "step": 5626
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7667551912339767e-06,
+      "loss": 0.4365,
+      "step": 5627
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7667133686037515e-06,
+      "loss": 0.0016,
+      "step": 5628
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7666715459735263e-06,
+      "loss": 0.0112,
+      "step": 5629
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766629723343301e-06,
+      "loss": 0.0017,
+      "step": 5630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7665879007130756e-06,
+      "loss": 0.0015,
+      "step": 5631
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7665460780828504e-06,
+      "loss": 0.3086,
+      "step": 5632
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7665042554526252e-06,
+      "loss": 0.0012,
+      "step": 5633
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7664624328224e-06,
+      "loss": 0.0014,
+      "step": 5634
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7664206101921748e-06,
+      "loss": 1.5042,
+      "step": 5635
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7663787875619498e-06,
+      "loss": 0.0134,
+      "step": 5636
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7663369649317246e-06,
+      "loss": 0.1676,
+      "step": 5637
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7662951423014994e-06,
+      "loss": 0.0019,
+      "step": 5638
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7662533196712741e-06,
+      "loss": 1.4909,
+      "step": 5639
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766211497041049e-06,
+      "loss": 0.0013,
+      "step": 5640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7661696744108237e-06,
+      "loss": 0.0011,
+      "step": 5641
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7661278517805985e-06,
+      "loss": 2.1924,
+      "step": 5642
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7660860291503733e-06,
+      "loss": 0.0017,
+      "step": 5643
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7660442065201479e-06,
+      "loss": 0.0012,
+      "step": 5644
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7660023838899226e-06,
+      "loss": 0.0142,
+      "step": 5645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7659605612596974e-06,
+      "loss": 2.3974,
+      "step": 5646
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7659187386294722e-06,
+      "loss": 0.0018,
+      "step": 5647
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.765876915999247e-06,
+      "loss": 0.0011,
+      "step": 5648
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7658350933690218e-06,
+      "loss": 0.7283,
+      "step": 5649
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7657932707387966e-06,
+      "loss": 1.1509,
+      "step": 5650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7657514481085716e-06,
+      "loss": 0.0015,
+      "step": 5651
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7657096254783464e-06,
+      "loss": 0.0012,
+      "step": 5652
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7656678028481211e-06,
+      "loss": 0.0012,
+      "step": 5653
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.765625980217896e-06,
+      "loss": 0.0091,
+      "step": 5654
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7655841575876707e-06,
+      "loss": 0.0013,
+      "step": 5655
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7655423349574455e-06,
+      "loss": 0.0061,
+      "step": 5656
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7655005123272203e-06,
+      "loss": 0.0044,
+      "step": 5657
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7654586896969949e-06,
+      "loss": 0.0008,
+      "step": 5658
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7654168670667697e-06,
+      "loss": 0.0016,
+      "step": 5659
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7653750444365444e-06,
+      "loss": 0.0021,
+      "step": 5660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7653332218063192e-06,
+      "loss": 1.0829,
+      "step": 5661
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.765291399176094e-06,
+      "loss": 0.0012,
+      "step": 5662
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7652495765458688e-06,
+      "loss": 0.0021,
+      "step": 5663
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7652077539156436e-06,
+      "loss": 0.003,
+      "step": 5664
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7651659312854184e-06,
+      "loss": 0.0029,
+      "step": 5665
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7651241086551934e-06,
+      "loss": 0.0018,
+      "step": 5666
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7650822860249682e-06,
+      "loss": 0.0011,
+      "step": 5667
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.765040463394743e-06,
+      "loss": 3.5301,
+      "step": 5668
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7649986407645177e-06,
+      "loss": 0.1012,
+      "step": 5669
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7649568181342925e-06,
+      "loss": 0.0069,
+      "step": 5670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.764914995504067e-06,
+      "loss": 0.7587,
+      "step": 5671
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7648731728738419e-06,
+      "loss": 0.004,
+      "step": 5672
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7648313502436167e-06,
+      "loss": 0.0013,
+      "step": 5673
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7647895276133914e-06,
+      "loss": 0.0016,
+      "step": 5674
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7647477049831662e-06,
+      "loss": 0.0009,
+      "step": 5675
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.764705882352941e-06,
+      "loss": 0.1461,
+      "step": 5676
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7646640597227158e-06,
+      "loss": 0.0014,
+      "step": 5677
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7646222370924906e-06,
+      "loss": 0.3253,
+      "step": 5678
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7645804144622654e-06,
+      "loss": 0.0019,
+      "step": 5679
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7645385918320404e-06,
+      "loss": 0.0018,
+      "step": 5680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7644967692018152e-06,
+      "loss": 3.7066,
+      "step": 5681
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.76445494657159e-06,
+      "loss": 0.0106,
+      "step": 5682
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7644131239413647e-06,
+      "loss": 0.0029,
+      "step": 5683
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7643713013111395e-06,
+      "loss": 0.1207,
+      "step": 5684
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.764329478680914e-06,
+      "loss": 0.0009,
+      "step": 5685
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7642876560506889e-06,
+      "loss": 0.4365,
+      "step": 5686
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7642458334204637e-06,
+      "loss": 0.0009,
+      "step": 5687
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7642040107902385e-06,
+      "loss": 0.0202,
+      "step": 5688
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7641621881600132e-06,
+      "loss": 0.0015,
+      "step": 5689
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.764120365529788e-06,
+      "loss": 0.0016,
+      "step": 5690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7640785428995628e-06,
+      "loss": 0.0019,
+      "step": 5691
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7640367202693376e-06,
+      "loss": 1.8395,
+      "step": 5692
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7639948976391124e-06,
+      "loss": 1.4655,
+      "step": 5693
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7639530750088872e-06,
+      "loss": 1.6952,
+      "step": 5694
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7639112523786622e-06,
+      "loss": 0.0011,
+      "step": 5695
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763869429748437e-06,
+      "loss": 0.0011,
+      "step": 5696
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7638276071182117e-06,
+      "loss": 1.1312,
+      "step": 5697
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7637857844879865e-06,
+      "loss": 0.0016,
+      "step": 5698
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7637439618577611e-06,
+      "loss": 0.0028,
+      "step": 5699
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7637021392275359e-06,
+      "loss": 0.0009,
+      "step": 5700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7636603165973107e-06,
+      "loss": 0.0016,
+      "step": 5701
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7636184939670855e-06,
+      "loss": 0.0118,
+      "step": 5702
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7635766713368603e-06,
+      "loss": 0.0065,
+      "step": 5703
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763534848706635e-06,
+      "loss": 0.0021,
+      "step": 5704
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7634930260764098e-06,
+      "loss": 0.0026,
+      "step": 5705
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7634512034461846e-06,
+      "loss": 0.0012,
+      "step": 5706
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7634093808159594e-06,
+      "loss": 0.0017,
+      "step": 5707
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7633675581857342e-06,
+      "loss": 0.0037,
+      "step": 5708
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763325735555509e-06,
+      "loss": 0.001,
+      "step": 5709
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763283912925284e-06,
+      "loss": 0.0037,
+      "step": 5710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7632420902950588e-06,
+      "loss": 0.0017,
+      "step": 5711
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7632002676648333e-06,
+      "loss": 0.0018,
+      "step": 5712
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7631584450346081e-06,
+      "loss": 0.0008,
+      "step": 5713
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763116622404383e-06,
+      "loss": 0.001,
+      "step": 5714
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7630747997741577e-06,
+      "loss": 0.0017,
+      "step": 5715
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7630329771439325e-06,
+      "loss": 0.0025,
+      "step": 5716
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7629911545137073e-06,
+      "loss": 0.0022,
+      "step": 5717
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762949331883482e-06,
+      "loss": 0.004,
+      "step": 5718
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7629075092532568e-06,
+      "loss": 0.0037,
+      "step": 5719
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7628656866230316e-06,
+      "loss": 0.0065,
+      "step": 5720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7628238639928064e-06,
+      "loss": 0.0026,
+      "step": 5721
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7627820413625812e-06,
+      "loss": 0.0603,
+      "step": 5722
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762740218732356e-06,
+      "loss": 1.7782,
+      "step": 5723
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762698396102131e-06,
+      "loss": 0.0072,
+      "step": 5724
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7626565734719058e-06,
+      "loss": 2.1469,
+      "step": 5725
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7626147508416803e-06,
+      "loss": 0.823,
+      "step": 5726
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7625729282114551e-06,
+      "loss": 0.0661,
+      "step": 5727
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.76253110558123e-06,
+      "loss": 0.0017,
+      "step": 5728
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7624892829510047e-06,
+      "loss": 0.001,
+      "step": 5729
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7624474603207795e-06,
+      "loss": 1.5727,
+      "step": 5730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7624056376905543e-06,
+      "loss": 0.0014,
+      "step": 5731
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762363815060329e-06,
+      "loss": 0.006,
+      "step": 5732
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7623219924301038e-06,
+      "loss": 0.0022,
+      "step": 5733
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7622801697998786e-06,
+      "loss": 0.0063,
+      "step": 5734
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7622383471696534e-06,
+      "loss": 0.0012,
+      "step": 5735
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7621965245394282e-06,
+      "loss": 0.0148,
+      "step": 5736
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762154701909203e-06,
+      "loss": 0.002,
+      "step": 5737
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7621128792789778e-06,
+      "loss": 0.002,
+      "step": 5738
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7620710566487526e-06,
+      "loss": 0.0092,
+      "step": 5739
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7620292340185273e-06,
+      "loss": 0.0013,
+      "step": 5740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7619874113883021e-06,
+      "loss": 0.2252,
+      "step": 5741
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761945588758077e-06,
+      "loss": 0.0026,
+      "step": 5742
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7619037661278517e-06,
+      "loss": 0.1153,
+      "step": 5743
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7618619434976265e-06,
+      "loss": 0.0013,
+      "step": 5744
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7618201208674013e-06,
+      "loss": 0.1154,
+      "step": 5745
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761778298237176e-06,
+      "loss": 0.0013,
+      "step": 5746
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7617364756069509e-06,
+      "loss": 0.0088,
+      "step": 5747
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7616946529767256e-06,
+      "loss": 0.0009,
+      "step": 5748
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7616528303465004e-06,
+      "loss": 0.0048,
+      "step": 5749
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7616110077162752e-06,
+      "loss": 0.6985,
+      "step": 5750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.76156918508605e-06,
+      "loss": 0.0007,
+      "step": 5751
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7615273624558246e-06,
+      "loss": 0.0118,
+      "step": 5752
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7614855398255994e-06,
+      "loss": 0.0013,
+      "step": 5753
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7614437171953744e-06,
+      "loss": 0.0053,
+      "step": 5754
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7614018945651491e-06,
+      "loss": 0.0535,
+      "step": 5755
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761360071934924e-06,
+      "loss": 0.0009,
+      "step": 5756
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7613182493046987e-06,
+      "loss": 0.003,
+      "step": 5757
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7612764266744735e-06,
+      "loss": 0.0013,
+      "step": 5758
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7612346040442483e-06,
+      "loss": 0.2578,
+      "step": 5759
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761192781414023e-06,
+      "loss": 1.0383,
+      "step": 5760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7611509587837979e-06,
+      "loss": 1.6683,
+      "step": 5761
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7611091361535726e-06,
+      "loss": 0.0014,
+      "step": 5762
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7610673135233474e-06,
+      "loss": 0.0061,
+      "step": 5763
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7610254908931222e-06,
+      "loss": 0.0012,
+      "step": 5764
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760983668262897e-06,
+      "loss": 0.001,
+      "step": 5765
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7609418456326716e-06,
+      "loss": 0.0063,
+      "step": 5766
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7609000230024464e-06,
+      "loss": 2.593,
+      "step": 5767
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7608582003722214e-06,
+      "loss": 0.0064,
+      "step": 5768
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7608163777419962e-06,
+      "loss": 0.0013,
+      "step": 5769
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760774555111771e-06,
+      "loss": 3.7981,
+      "step": 5770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7607327324815457e-06,
+      "loss": 0.0008,
+      "step": 5771
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7606909098513205e-06,
+      "loss": 0.0014,
+      "step": 5772
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7606490872210953e-06,
+      "loss": 0.0028,
+      "step": 5773
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.76060726459087e-06,
+      "loss": 0.0008,
+      "step": 5774
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7605654419606449e-06,
+      "loss": 0.0009,
+      "step": 5775
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7605236193304197e-06,
+      "loss": 0.0771,
+      "step": 5776
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7604817967001944e-06,
+      "loss": 0.0008,
+      "step": 5777
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7604399740699692e-06,
+      "loss": 0.0041,
+      "step": 5778
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7603981514397438e-06,
+      "loss": 0.0011,
+      "step": 5779
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7603563288095186e-06,
+      "loss": 0.0011,
+      "step": 5780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7603145061792934e-06,
+      "loss": 0.0011,
+      "step": 5781
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7602726835490682e-06,
+      "loss": 0.0023,
+      "step": 5782
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7602308609188432e-06,
+      "loss": 2.0797,
+      "step": 5783
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760189038288618e-06,
+      "loss": 0.0009,
+      "step": 5784
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7601472156583927e-06,
+      "loss": 1.9995,
+      "step": 5785
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7601053930281675e-06,
+      "loss": 0.0015,
+      "step": 5786
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7600635703979423e-06,
+      "loss": 0.0009,
+      "step": 5787
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760021747767717e-06,
+      "loss": 0.0012,
+      "step": 5788
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7599799251374919e-06,
+      "loss": 0.0014,
+      "step": 5789
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7599381025072667e-06,
+      "loss": 1.0281,
+      "step": 5790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7598962798770414e-06,
+      "loss": 0.0015,
+      "step": 5791
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7598544572468162e-06,
+      "loss": 3.1211,
+      "step": 5792
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7598126346165908e-06,
+      "loss": 0.0428,
+      "step": 5793
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7597708119863656e-06,
+      "loss": 0.002,
+      "step": 5794
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7597289893561404e-06,
+      "loss": 0.0012,
+      "step": 5795
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7596871667259152e-06,
+      "loss": 0.0023,
+      "step": 5796
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.75964534409569e-06,
+      "loss": 0.0014,
+      "step": 5797
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759603521465465e-06,
+      "loss": 1.8837,
+      "step": 5798
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7595616988352397e-06,
+      "loss": 0.3807,
+      "step": 5799
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7595198762050145e-06,
+      "loss": 0.0009,
+      "step": 5800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7594780535747893e-06,
+      "loss": 0.0021,
+      "step": 5801
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759436230944564e-06,
+      "loss": 0.0008,
+      "step": 5802
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7593944083143389e-06,
+      "loss": 0.0018,
+      "step": 5803
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7593525856841137e-06,
+      "loss": 0.5944,
+      "step": 5804
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7593107630538885e-06,
+      "loss": 0.0168,
+      "step": 5805
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7592689404236632e-06,
+      "loss": 2.4246,
+      "step": 5806
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7592271177934378e-06,
+      "loss": 0.0013,
+      "step": 5807
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7591852951632126e-06,
+      "loss": 0.0092,
+      "step": 5808
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7591434725329874e-06,
+      "loss": 0.3985,
+      "step": 5809
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7591016499027622e-06,
+      "loss": 0.0012,
+      "step": 5810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759059827272537e-06,
+      "loss": 0.0011,
+      "step": 5811
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759018004642312e-06,
+      "loss": 1.2136,
+      "step": 5812
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7589761820120867e-06,
+      "loss": 0.0018,
+      "step": 5813
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7589343593818615e-06,
+      "loss": 0.0036,
+      "step": 5814
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7588925367516363e-06,
+      "loss": 0.0047,
+      "step": 5815
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7588507141214111e-06,
+      "loss": 3.941,
+      "step": 5816
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.758808891491186e-06,
+      "loss": 0.001,
+      "step": 5817
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7587670688609607e-06,
+      "loss": 0.0022,
+      "step": 5818
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7587252462307355e-06,
+      "loss": 0.0038,
+      "step": 5819
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.75868342360051e-06,
+      "loss": 0.0058,
+      "step": 5820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7586416009702848e-06,
+      "loss": 0.0027,
+      "step": 5821
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7585997783400596e-06,
+      "loss": 0.0025,
+      "step": 5822
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7585579557098344e-06,
+      "loss": 0.9763,
+      "step": 5823
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7585161330796092e-06,
+      "loss": 0.0015,
+      "step": 5824
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.758474310449384e-06,
+      "loss": 0.1976,
+      "step": 5825
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7584324878191588e-06,
+      "loss": 0.0016,
+      "step": 5826
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7583906651889338e-06,
+      "loss": 0.5762,
+      "step": 5827
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7583488425587085e-06,
+      "loss": 0.0222,
+      "step": 5828
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7583070199284833e-06,
+      "loss": 0.0067,
+      "step": 5829
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7582651972982581e-06,
+      "loss": 0.0014,
+      "step": 5830
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.758223374668033e-06,
+      "loss": 0.0073,
+      "step": 5831
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7581815520378077e-06,
+      "loss": 0.0117,
+      "step": 5832
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7581397294075825e-06,
+      "loss": 0.001,
+      "step": 5833
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.758097906777357e-06,
+      "loss": 0.0012,
+      "step": 5834
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7580560841471318e-06,
+      "loss": 0.0008,
+      "step": 5835
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7580142615169066e-06,
+      "loss": 0.0291,
+      "step": 5836
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7579724388866814e-06,
+      "loss": 0.0021,
+      "step": 5837
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7579306162564562e-06,
+      "loss": 1.5196,
+      "step": 5838
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.757888793626231e-06,
+      "loss": 0.1426,
+      "step": 5839
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7578469709960058e-06,
+      "loss": 0.001,
+      "step": 5840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7578051483657808e-06,
+      "loss": 0.0009,
+      "step": 5841
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7577633257355556e-06,
+      "loss": 0.842,
+      "step": 5842
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7577215031053303e-06,
+      "loss": 0.0803,
+      "step": 5843
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7576796804751051e-06,
+      "loss": 0.0061,
+      "step": 5844
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75763785784488e-06,
+      "loss": 0.0015,
+      "step": 5845
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7575960352146547e-06,
+      "loss": 0.8833,
+      "step": 5846
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7575542125844293e-06,
+      "loss": 0.0041,
+      "step": 5847
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.757512389954204e-06,
+      "loss": 0.0012,
+      "step": 5848
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7574705673239788e-06,
+      "loss": 0.0958,
+      "step": 5849
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7574287446937536e-06,
+      "loss": 0.0015,
+      "step": 5850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7573869220635284e-06,
+      "loss": 0.0125,
+      "step": 5851
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7573450994333032e-06,
+      "loss": 0.0014,
+      "step": 5852
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.757303276803078e-06,
+      "loss": 0.0029,
+      "step": 5853
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7572614541728528e-06,
+      "loss": 0.7371,
+      "step": 5854
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7572196315426276e-06,
+      "loss": 0.5708,
+      "step": 5855
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7571778089124026e-06,
+      "loss": 0.0008,
+      "step": 5856
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7571359862821773e-06,
+      "loss": 0.0009,
+      "step": 5857
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7570941636519521e-06,
+      "loss": 0.0022,
+      "step": 5858
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.757052341021727e-06,
+      "loss": 0.0009,
+      "step": 5859
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7570105183915017e-06,
+      "loss": 0.0019,
+      "step": 5860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7569686957612763e-06,
+      "loss": 0.0037,
+      "step": 5861
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756926873131051e-06,
+      "loss": 0.0063,
+      "step": 5862
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7568850505008259e-06,
+      "loss": 0.0015,
+      "step": 5863
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7568432278706006e-06,
+      "loss": 0.1152,
+      "step": 5864
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7568014052403754e-06,
+      "loss": 2.497,
+      "step": 5865
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7567595826101502e-06,
+      "loss": 0.0016,
+      "step": 5866
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756717759979925e-06,
+      "loss": 0.0032,
+      "step": 5867
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7566759373496998e-06,
+      "loss": 0.0008,
+      "step": 5868
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7566341147194746e-06,
+      "loss": 0.0046,
+      "step": 5869
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7565922920892494e-06,
+      "loss": 0.0297,
+      "step": 5870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7565504694590244e-06,
+      "loss": 0.1036,
+      "step": 5871
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7565086468287991e-06,
+      "loss": 0.0012,
+      "step": 5872
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756466824198574e-06,
+      "loss": 0.0025,
+      "step": 5873
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7564250015683485e-06,
+      "loss": 0.0063,
+      "step": 5874
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7563831789381233e-06,
+      "loss": 0.0041,
+      "step": 5875
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756341356307898e-06,
+      "loss": 0.0006,
+      "step": 5876
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7562995336776729e-06,
+      "loss": 0.0021,
+      "step": 5877
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7562577110474476e-06,
+      "loss": 0.0024,
+      "step": 5878
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7562158884172224e-06,
+      "loss": 0.0019,
+      "step": 5879
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7561740657869972e-06,
+      "loss": 0.0008,
+      "step": 5880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756132243156772e-06,
+      "loss": 0.0031,
+      "step": 5881
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7560904205265468e-06,
+      "loss": 0.0008,
+      "step": 5882
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7560485978963216e-06,
+      "loss": 3.5987,
+      "step": 5883
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7560067752660964e-06,
+      "loss": 0.0079,
+      "step": 5884
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7559649526358714e-06,
+      "loss": 0.0018,
+      "step": 5885
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7559231300056462e-06,
+      "loss": 0.002,
+      "step": 5886
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755881307375421e-06,
+      "loss": 0.0663,
+      "step": 5887
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7558394847451955e-06,
+      "loss": 0.0007,
+      "step": 5888
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7557976621149703e-06,
+      "loss": 0.004,
+      "step": 5889
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755755839484745e-06,
+      "loss": 3.1689,
+      "step": 5890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7557140168545199e-06,
+      "loss": 0.0105,
+      "step": 5891
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7556721942242947e-06,
+      "loss": 0.0008,
+      "step": 5892
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7556303715940694e-06,
+      "loss": 0.0013,
+      "step": 5893
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7555885489638442e-06,
+      "loss": 0.2425,
+      "step": 5894
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755546726333619e-06,
+      "loss": 0.0093,
+      "step": 5895
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7555049037033938e-06,
+      "loss": 1.8578,
+      "step": 5896
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7554630810731686e-06,
+      "loss": 0.0432,
+      "step": 5897
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7554212584429434e-06,
+      "loss": 1.2686,
+      "step": 5898
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7553794358127182e-06,
+      "loss": 0.5439,
+      "step": 5899
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7553376131824932e-06,
+      "loss": 0.0008,
+      "step": 5900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755295790552268e-06,
+      "loss": 0.0034,
+      "step": 5901
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7552539679220425e-06,
+      "loss": 2.8964,
+      "step": 5902
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7552121452918173e-06,
+      "loss": 1.8975,
+      "step": 5903
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755170322661592e-06,
+      "loss": 0.001,
+      "step": 5904
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7551285000313669e-06,
+      "loss": 0.0011,
+      "step": 5905
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7550866774011417e-06,
+      "loss": 0.0349,
+      "step": 5906
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7550448547709165e-06,
+      "loss": 0.0016,
+      "step": 5907
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7550030321406912e-06,
+      "loss": 0.1243,
+      "step": 5908
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754961209510466e-06,
+      "loss": 1.4844,
+      "step": 5909
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7549193868802408e-06,
+      "loss": 0.0028,
+      "step": 5910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7548775642500156e-06,
+      "loss": 0.0011,
+      "step": 5911
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7548357416197904e-06,
+      "loss": 0.0012,
+      "step": 5912
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7547939189895652e-06,
+      "loss": 0.0008,
+      "step": 5913
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75475209635934e-06,
+      "loss": 0.0009,
+      "step": 5914
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7547102737291147e-06,
+      "loss": 0.0012,
+      "step": 5915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7546684510988895e-06,
+      "loss": 1.4215,
+      "step": 5916
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7546266284686643e-06,
+      "loss": 0.0322,
+      "step": 5917
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754584805838439e-06,
+      "loss": 1.1394,
+      "step": 5918
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7545429832082139e-06,
+      "loss": 0.0042,
+      "step": 5919
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7545011605779887e-06,
+      "loss": 0.0011,
+      "step": 5920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7544593379477635e-06,
+      "loss": 0.0012,
+      "step": 5921
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7544175153175382e-06,
+      "loss": 2.7454,
+      "step": 5922
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754375692687313e-06,
+      "loss": 2.3463,
+      "step": 5923
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7543338700570878e-06,
+      "loss": 0.001,
+      "step": 5924
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7542920474268626e-06,
+      "loss": 0.0011,
+      "step": 5925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7542502247966374e-06,
+      "loss": 1.1948,
+      "step": 5926
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7542084021664122e-06,
+      "loss": 0.002,
+      "step": 5927
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7541665795361868e-06,
+      "loss": 0.0031,
+      "step": 5928
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7541247569059618e-06,
+      "loss": 0.6447,
+      "step": 5929
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7540829342757365e-06,
+      "loss": 0.0125,
+      "step": 5930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7540411116455113e-06,
+      "loss": 2.8092,
+      "step": 5931
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7539992890152861e-06,
+      "loss": 0.0019,
+      "step": 5932
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753957466385061e-06,
+      "loss": 0.0037,
+      "step": 5933
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7539156437548357e-06,
+      "loss": 2.5327,
+      "step": 5934
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7538738211246105e-06,
+      "loss": 0.0086,
+      "step": 5935
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7538319984943853e-06,
+      "loss": 0.0011,
+      "step": 5936
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75379017586416e-06,
+      "loss": 0.3948,
+      "step": 5937
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7537483532339348e-06,
+      "loss": 1.1913,
+      "step": 5938
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7537065306037096e-06,
+      "loss": 0.0014,
+      "step": 5939
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7536647079734844e-06,
+      "loss": 0.0017,
+      "step": 5940
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7536228853432592e-06,
+      "loss": 0.0086,
+      "step": 5941
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7535810627130338e-06,
+      "loss": 0.0015,
+      "step": 5942
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7535392400828085e-06,
+      "loss": 0.0027,
+      "step": 5943
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7534974174525835e-06,
+      "loss": 0.0383,
+      "step": 5944
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7534555948223583e-06,
+      "loss": 0.0016,
+      "step": 5945
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7534137721921331e-06,
+      "loss": 0.0015,
+      "step": 5946
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753371949561908e-06,
+      "loss": 0.0044,
+      "step": 5947
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7533301269316827e-06,
+      "loss": 0.0042,
+      "step": 5948
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7532883043014575e-06,
+      "loss": 0.0021,
+      "step": 5949
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7532464816712323e-06,
+      "loss": 0.0016,
+      "step": 5950
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753204659041007e-06,
+      "loss": 0.0039,
+      "step": 5951
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7531628364107818e-06,
+      "loss": 0.0078,
+      "step": 5952
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7531210137805566e-06,
+      "loss": 0.0044,
+      "step": 5953
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7530791911503314e-06,
+      "loss": 0.0021,
+      "step": 5954
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753037368520106e-06,
+      "loss": 0.8438,
+      "step": 5955
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7529955458898808e-06,
+      "loss": 1.8921,
+      "step": 5956
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7529537232596556e-06,
+      "loss": 0.0051,
+      "step": 5957
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7529119006294303e-06,
+      "loss": 0.0015,
+      "step": 5958
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7528700779992053e-06,
+      "loss": 0.0019,
+      "step": 5959
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7528282553689801e-06,
+      "loss": 1.7208,
+      "step": 5960
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752786432738755e-06,
+      "loss": 0.002,
+      "step": 5961
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7527446101085297e-06,
+      "loss": 0.0047,
+      "step": 5962
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7527027874783045e-06,
+      "loss": 0.2193,
+      "step": 5963
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7526609648480793e-06,
+      "loss": 0.0026,
+      "step": 5964
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752619142217854e-06,
+      "loss": 0.001,
+      "step": 5965
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7525773195876288e-06,
+      "loss": 0.002,
+      "step": 5966
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7525354969574036e-06,
+      "loss": 0.005,
+      "step": 5967
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7524936743271784e-06,
+      "loss": 0.0012,
+      "step": 5968
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752451851696953e-06,
+      "loss": 2.2109,
+      "step": 5969
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7524100290667278e-06,
+      "loss": 1.1756,
+      "step": 5970
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7523682064365026e-06,
+      "loss": 0.6833,
+      "step": 5971
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7523263838062774e-06,
+      "loss": 0.002,
+      "step": 5972
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7522845611760523e-06,
+      "loss": 0.0086,
+      "step": 5973
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7522427385458271e-06,
+      "loss": 2.3691,
+      "step": 5974
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752200915915602e-06,
+      "loss": 0.3784,
+      "step": 5975
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7521590932853767e-06,
+      "loss": 0.0044,
+      "step": 5976
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7521172706551515e-06,
+      "loss": 0.0055,
+      "step": 5977
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7520754480249263e-06,
+      "loss": 1.9496,
+      "step": 5978
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752033625394701e-06,
+      "loss": 0.0011,
+      "step": 5979
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7519918027644759e-06,
+      "loss": 0.0014,
+      "step": 5980
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7519499801342506e-06,
+      "loss": 0.0017,
+      "step": 5981
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7519081575040252e-06,
+      "loss": 0.0033,
+      "step": 5982
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7518663348738e-06,
+      "loss": 0.0018,
+      "step": 5983
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7518245122435748e-06,
+      "loss": 0.003,
+      "step": 5984
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7517826896133496e-06,
+      "loss": 0.0026,
+      "step": 5985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7517408669831244e-06,
+      "loss": 0.7762,
+      "step": 5986
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7516990443528991e-06,
+      "loss": 0.0041,
+      "step": 5987
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7516572217226741e-06,
+      "loss": 0.0015,
+      "step": 5988
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.751615399092449e-06,
+      "loss": 0.0036,
+      "step": 5989
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7515735764622237e-06,
+      "loss": 0.003,
+      "step": 5990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7515317538319985e-06,
+      "loss": 0.0017,
+      "step": 5991
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7514899312017733e-06,
+      "loss": 0.001,
+      "step": 5992
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.751448108571548e-06,
+      "loss": 0.3817,
+      "step": 5993
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7514062859413229e-06,
+      "loss": 0.5197,
+      "step": 5994
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7513644633110976e-06,
+      "loss": 0.0024,
+      "step": 5995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7513226406808722e-06,
+      "loss": 0.0023,
+      "step": 5996
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.751280818050647e-06,
+      "loss": 0.0008,
+      "step": 5997
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7512389954204218e-06,
+      "loss": 0.001,
+      "step": 5998
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7511971727901966e-06,
+      "loss": 0.001,
+      "step": 5999
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7511553501599714e-06,
+      "loss": 0.0022,
+      "step": 6000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7511135275297462e-06,
+      "loss": 0.0039,
+      "step": 6001
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.751071704899521e-06,
+      "loss": 1.1543,
+      "step": 6002
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.751029882269296e-06,
+      "loss": 0.0023,
+      "step": 6003
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7509880596390707e-06,
+      "loss": 0.001,
+      "step": 6004
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7509462370088455e-06,
+      "loss": 1.7214,
+      "step": 6005
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7509044143786203e-06,
+      "loss": 0.001,
+      "step": 6006
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.750862591748395e-06,
+      "loss": 0.0018,
+      "step": 6007
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7508207691181699e-06,
+      "loss": 0.2808,
+      "step": 6008
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7507789464879447e-06,
+      "loss": 1.9167,
+      "step": 6009
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7507371238577192e-06,
+      "loss": 0.0032,
+      "step": 6010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.750695301227494e-06,
+      "loss": 0.0025,
+      "step": 6011
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7506534785972688e-06,
+      "loss": 0.133,
+      "step": 6012
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7506116559670436e-06,
+      "loss": 0.0011,
+      "step": 6013
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7505698333368184e-06,
+      "loss": 2.4989,
+      "step": 6014
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7505280107065932e-06,
+      "loss": 0.0014,
+      "step": 6015
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.750486188076368e-06,
+      "loss": 0.0119,
+      "step": 6016
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.750444365446143e-06,
+      "loss": 0.0113,
+      "step": 6017
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7504025428159177e-06,
+      "loss": 0.0012,
+      "step": 6018
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7503607201856925e-06,
+      "loss": 0.0009,
+      "step": 6019
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7503188975554673e-06,
+      "loss": 0.0013,
+      "step": 6020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.750277074925242e-06,
+      "loss": 0.0026,
+      "step": 6021
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7502352522950169e-06,
+      "loss": 0.0023,
+      "step": 6022
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7501934296647915e-06,
+      "loss": 0.0103,
+      "step": 6023
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7501516070345662e-06,
+      "loss": 1.9268,
+      "step": 6024
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.750109784404341e-06,
+      "loss": 0.0021,
+      "step": 6025
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7500679617741158e-06,
+      "loss": 0.5922,
+      "step": 6026
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7500261391438906e-06,
+      "loss": 0.0033,
+      "step": 6027
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7499843165136654e-06,
+      "loss": 0.002,
+      "step": 6028
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7499424938834402e-06,
+      "loss": 3.7237,
+      "step": 6029
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749900671253215e-06,
+      "loss": 0.0009,
+      "step": 6030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7498588486229897e-06,
+      "loss": 0.0024,
+      "step": 6031
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7498170259927647e-06,
+      "loss": 0.0015,
+      "step": 6032
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7497752033625395e-06,
+      "loss": 0.0023,
+      "step": 6033
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7497333807323143e-06,
+      "loss": 0.0012,
+      "step": 6034
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749691558102089e-06,
+      "loss": 0.0603,
+      "step": 6035
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7496497354718639e-06,
+      "loss": 0.2306,
+      "step": 6036
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7496079128416385e-06,
+      "loss": 0.0017,
+      "step": 6037
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7495660902114132e-06,
+      "loss": 0.0012,
+      "step": 6038
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749524267581188e-06,
+      "loss": 2.2713,
+      "step": 6039
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7494824449509628e-06,
+      "loss": 3.093,
+      "step": 6040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7494406223207376e-06,
+      "loss": 0.001,
+      "step": 6041
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7493987996905124e-06,
+      "loss": 0.0009,
+      "step": 6042
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7493569770602872e-06,
+      "loss": 0.0056,
+      "step": 6043
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749315154430062e-06,
+      "loss": 0.0042,
+      "step": 6044
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7492733317998368e-06,
+      "loss": 0.0022,
+      "step": 6045
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7492315091696118e-06,
+      "loss": 0.3919,
+      "step": 6046
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7491896865393865e-06,
+      "loss": 0.0032,
+      "step": 6047
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7491478639091613e-06,
+      "loss": 0.0023,
+      "step": 6048
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7491060412789361e-06,
+      "loss": 0.0011,
+      "step": 6049
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7490642186487107e-06,
+      "loss": 0.0027,
+      "step": 6050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7490223960184855e-06,
+      "loss": 0.0018,
+      "step": 6051
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7489805733882603e-06,
+      "loss": 0.0018,
+      "step": 6052
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748938750758035e-06,
+      "loss": 0.0014,
+      "step": 6053
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7488969281278098e-06,
+      "loss": 0.0035,
+      "step": 6054
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7488551054975846e-06,
+      "loss": 0.0043,
+      "step": 6055
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7488132828673594e-06,
+      "loss": 0.4975,
+      "step": 6056
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7487714602371342e-06,
+      "loss": 0.0048,
+      "step": 6057
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748729637606909e-06,
+      "loss": 0.0021,
+      "step": 6058
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7486878149766838e-06,
+      "loss": 0.0059,
+      "step": 6059
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7486459923464585e-06,
+      "loss": 0.0015,
+      "step": 6060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7486041697162335e-06,
+      "loss": 2.0578,
+      "step": 6061
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7485623470860083e-06,
+      "loss": 0.4812,
+      "step": 6062
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7485205244557831e-06,
+      "loss": 0.0011,
+      "step": 6063
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7484787018255577e-06,
+      "loss": 0.0006,
+      "step": 6064
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7484368791953325e-06,
+      "loss": 0.0744,
+      "step": 6065
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7483950565651073e-06,
+      "loss": 0.001,
+      "step": 6066
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748353233934882e-06,
+      "loss": 0.0011,
+      "step": 6067
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7483114113046568e-06,
+      "loss": 0.0007,
+      "step": 6068
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7482695886744316e-06,
+      "loss": 0.7269,
+      "step": 6069
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7482277660442064e-06,
+      "loss": 0.443,
+      "step": 6070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7481859434139812e-06,
+      "loss": 0.0062,
+      "step": 6071
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748144120783756e-06,
+      "loss": 0.2935,
+      "step": 6072
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7481022981535308e-06,
+      "loss": 0.0019,
+      "step": 6073
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7480604755233056e-06,
+      "loss": 0.0033,
+      "step": 6074
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7480186528930803e-06,
+      "loss": 0.0021,
+      "step": 6075
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7479768302628553e-06,
+      "loss": 0.0012,
+      "step": 6076
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.74793500763263e-06,
+      "loss": 0.0012,
+      "step": 6077
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7478931850024047e-06,
+      "loss": 4.2417,
+      "step": 6078
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7478513623721795e-06,
+      "loss": 0.0034,
+      "step": 6079
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7478095397419543e-06,
+      "loss": 0.0155,
+      "step": 6080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747767717111729e-06,
+      "loss": 0.0026,
+      "step": 6081
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7477258944815038e-06,
+      "loss": 0.0056,
+      "step": 6082
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7476840718512786e-06,
+      "loss": 1.6076,
+      "step": 6083
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7476422492210534e-06,
+      "loss": 0.0033,
+      "step": 6084
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7476004265908282e-06,
+      "loss": 2.6617,
+      "step": 6085
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747558603960603e-06,
+      "loss": 0.0041,
+      "step": 6086
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7475167813303778e-06,
+      "loss": 0.3789,
+      "step": 6087
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7474749587001526e-06,
+      "loss": 0.0066,
+      "step": 6088
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7474331360699274e-06,
+      "loss": 0.0661,
+      "step": 6089
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7473913134397024e-06,
+      "loss": 0.001,
+      "step": 6090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747349490809477e-06,
+      "loss": 0.001,
+      "step": 6091
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7473076681792517e-06,
+      "loss": 0.3438,
+      "step": 6092
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7472658455490265e-06,
+      "loss": 0.0011,
+      "step": 6093
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7472240229188013e-06,
+      "loss": 0.0034,
+      "step": 6094
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747182200288576e-06,
+      "loss": 0.0012,
+      "step": 6095
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7471403776583509e-06,
+      "loss": 0.0051,
+      "step": 6096
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7470985550281256e-06,
+      "loss": 0.0014,
+      "step": 6097
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7470567323979004e-06,
+      "loss": 0.0039,
+      "step": 6098
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7470149097676752e-06,
+      "loss": 1.082,
+      "step": 6099
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.74697308713745e-06,
+      "loss": 1.8557,
+      "step": 6100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7469312645072248e-06,
+      "loss": 0.2615,
+      "step": 6101
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7468894418769996e-06,
+      "loss": 0.0026,
+      "step": 6102
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7468476192467744e-06,
+      "loss": 1.5343,
+      "step": 6103
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746805796616549e-06,
+      "loss": 0.0289,
+      "step": 6104
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746763973986324e-06,
+      "loss": 0.0019,
+      "step": 6105
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7467221513560987e-06,
+      "loss": 0.1596,
+      "step": 6106
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7466803287258735e-06,
+      "loss": 0.0017,
+      "step": 6107
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7466385060956483e-06,
+      "loss": 0.0022,
+      "step": 6108
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746596683465423e-06,
+      "loss": 0.0008,
+      "step": 6109
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7465548608351979e-06,
+      "loss": 0.006,
+      "step": 6110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7465130382049727e-06,
+      "loss": 2.6489,
+      "step": 6111
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7464712155747474e-06,
+      "loss": 0.0043,
+      "step": 6112
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7464293929445222e-06,
+      "loss": 1.9835,
+      "step": 6113
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746387570314297e-06,
+      "loss": 0.0015,
+      "step": 6114
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7463457476840718e-06,
+      "loss": 1.4448,
+      "step": 6115
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7463039250538466e-06,
+      "loss": 0.0021,
+      "step": 6116
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7462621024236214e-06,
+      "loss": 0.0054,
+      "step": 6117
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746220279793396e-06,
+      "loss": 0.1576,
+      "step": 6118
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7461784571631707e-06,
+      "loss": 0.0011,
+      "step": 6119
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7461366345329457e-06,
+      "loss": 0.0007,
+      "step": 6120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7460948119027205e-06,
+      "loss": 0.0025,
+      "step": 6121
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7460529892724953e-06,
+      "loss": 0.0015,
+      "step": 6122
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.74601116664227e-06,
+      "loss": 0.0013,
+      "step": 6123
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7459693440120449e-06,
+      "loss": 0.0015,
+      "step": 6124
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7459275213818197e-06,
+      "loss": 0.709,
+      "step": 6125
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7458856987515944e-06,
+      "loss": 0.003,
+      "step": 6126
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7458438761213692e-06,
+      "loss": 0.0008,
+      "step": 6127
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745802053491144e-06,
+      "loss": 0.0012,
+      "step": 6128
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7457602308609188e-06,
+      "loss": 0.0008,
+      "step": 6129
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7457184082306936e-06,
+      "loss": 0.2177,
+      "step": 6130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7456765856004682e-06,
+      "loss": 0.0007,
+      "step": 6131
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745634762970243e-06,
+      "loss": 0.0023,
+      "step": 6132
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7455929403400177e-06,
+      "loss": 0.0012,
+      "step": 6133
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7455511177097927e-06,
+      "loss": 0.0014,
+      "step": 6134
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7455092950795675e-06,
+      "loss": 0.001,
+      "step": 6135
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7454674724493423e-06,
+      "loss": 0.0069,
+      "step": 6136
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745425649819117e-06,
+      "loss": 0.0119,
+      "step": 6137
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7453838271888919e-06,
+      "loss": 0.0296,
+      "step": 6138
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7453420045586667e-06,
+      "loss": 0.0016,
+      "step": 6139
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7453001819284415e-06,
+      "loss": 0.0008,
+      "step": 6140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7452583592982162e-06,
+      "loss": 0.0013,
+      "step": 6141
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745216536667991e-06,
+      "loss": 0.6465,
+      "step": 6142
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7451747140377658e-06,
+      "loss": 0.0074,
+      "step": 6143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7451328914075406e-06,
+      "loss": 0.5876,
+      "step": 6144
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7450910687773152e-06,
+      "loss": 1.2571,
+      "step": 6145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.74504924614709e-06,
+      "loss": 0.0009,
+      "step": 6146
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7450074235168647e-06,
+      "loss": 0.0008,
+      "step": 6147
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7449656008866395e-06,
+      "loss": 0.0048,
+      "step": 6148
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7449237782564145e-06,
+      "loss": 0.0423,
+      "step": 6149
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7448819556261893e-06,
+      "loss": 1.9618,
+      "step": 6150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.744840132995964e-06,
+      "loss": 0.7643,
+      "step": 6151
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7447983103657389e-06,
+      "loss": 0.001,
+      "step": 6152
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7447564877355137e-06,
+      "loss": 0.0587,
+      "step": 6153
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7447146651052885e-06,
+      "loss": 0.0011,
+      "step": 6154
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7446728424750633e-06,
+      "loss": 0.0014,
+      "step": 6155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.744631019844838e-06,
+      "loss": 0.6737,
+      "step": 6156
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7445891972146128e-06,
+      "loss": 0.0025,
+      "step": 6157
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7445473745843874e-06,
+      "loss": 0.0119,
+      "step": 6158
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7445055519541622e-06,
+      "loss": 0.0101,
+      "step": 6159
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.744463729323937e-06,
+      "loss": 0.0017,
+      "step": 6160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7444219066937118e-06,
+      "loss": 0.0074,
+      "step": 6161
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7443800840634865e-06,
+      "loss": 1.7918,
+      "step": 6162
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7443382614332613e-06,
+      "loss": 0.0015,
+      "step": 6163
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7442964388030363e-06,
+      "loss": 0.0118,
+      "step": 6164
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7442546161728111e-06,
+      "loss": 0.1726,
+      "step": 6165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.744212793542586e-06,
+      "loss": 0.0016,
+      "step": 6166
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7441709709123607e-06,
+      "loss": 0.0013,
+      "step": 6167
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7441291482821355e-06,
+      "loss": 0.1829,
+      "step": 6168
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7440873256519103e-06,
+      "loss": 0.2011,
+      "step": 6169
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.744045503021685e-06,
+      "loss": 0.0038,
+      "step": 6170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7440036803914598e-06,
+      "loss": 4.1228,
+      "step": 6171
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7439618577612344e-06,
+      "loss": 3.2932,
+      "step": 6172
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7439200351310092e-06,
+      "loss": 0.001,
+      "step": 6173
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743878212500784e-06,
+      "loss": 0.0015,
+      "step": 6174
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7438363898705588e-06,
+      "loss": 0.0011,
+      "step": 6175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7437945672403336e-06,
+      "loss": 0.002,
+      "step": 6176
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7437527446101083e-06,
+      "loss": 0.0008,
+      "step": 6177
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7437109219798833e-06,
+      "loss": 0.0015,
+      "step": 6178
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7436690993496581e-06,
+      "loss": 3.7973,
+      "step": 6179
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743627276719433e-06,
+      "loss": 0.6,
+      "step": 6180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7435854540892077e-06,
+      "loss": 0.3533,
+      "step": 6181
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7435436314589825e-06,
+      "loss": 0.0012,
+      "step": 6182
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7435018088287573e-06,
+      "loss": 0.3707,
+      "step": 6183
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743459986198532e-06,
+      "loss": 1.7777,
+      "step": 6184
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7434181635683066e-06,
+      "loss": 0.2784,
+      "step": 6185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7433763409380814e-06,
+      "loss": 0.7162,
+      "step": 6186
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7433345183078562e-06,
+      "loss": 0.001,
+      "step": 6187
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743292695677631e-06,
+      "loss": 0.0036,
+      "step": 6188
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7432508730474058e-06,
+      "loss": 0.0017,
+      "step": 6189
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7432090504171806e-06,
+      "loss": 0.0051,
+      "step": 6190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7431672277869553e-06,
+      "loss": 0.0018,
+      "step": 6191
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7431254051567301e-06,
+      "loss": 0.0051,
+      "step": 6192
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7430835825265051e-06,
+      "loss": 0.0011,
+      "step": 6193
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.74304175989628e-06,
+      "loss": 0.008,
+      "step": 6194
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7429999372660547e-06,
+      "loss": 0.001,
+      "step": 6195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7429581146358295e-06,
+      "loss": 0.0733,
+      "step": 6196
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7429162920056043e-06,
+      "loss": 0.0008,
+      "step": 6197
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742874469375379e-06,
+      "loss": 0.0007,
+      "step": 6198
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7428326467451536e-06,
+      "loss": 1.4412,
+      "step": 6199
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7427908241149284e-06,
+      "loss": 0.0013,
+      "step": 6200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7427490014847032e-06,
+      "loss": 0.0027,
+      "step": 6201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742707178854478e-06,
+      "loss": 0.0069,
+      "step": 6202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7426653562242528e-06,
+      "loss": 0.0006,
+      "step": 6203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7426235335940276e-06,
+      "loss": 0.0008,
+      "step": 6204
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7425817109638024e-06,
+      "loss": 0.0049,
+      "step": 6205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7425398883335771e-06,
+      "loss": 2.0725,
+      "step": 6206
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7424980657033521e-06,
+      "loss": 0.002,
+      "step": 6207
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742456243073127e-06,
+      "loss": 0.0113,
+      "step": 6208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7424144204429017e-06,
+      "loss": 0.0009,
+      "step": 6209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7423725978126765e-06,
+      "loss": 0.0008,
+      "step": 6210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7423307751824513e-06,
+      "loss": 0.0011,
+      "step": 6211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742288952552226e-06,
+      "loss": 0.0021,
+      "step": 6212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7422471299220006e-06,
+      "loss": 0.0066,
+      "step": 6213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7422053072917754e-06,
+      "loss": 0.0018,
+      "step": 6214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7421634846615502e-06,
+      "loss": 1.2195,
+      "step": 6215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742121662031325e-06,
+      "loss": 1.637,
+      "step": 6216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7420798394010998e-06,
+      "loss": 1.0229,
+      "step": 6217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7420380167708746e-06,
+      "loss": 0.001,
+      "step": 6218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7419961941406494e-06,
+      "loss": 0.0212,
+      "step": 6219
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7419543715104241e-06,
+      "loss": 0.0012,
+      "step": 6220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741912548880199e-06,
+      "loss": 0.0011,
+      "step": 6221
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741870726249974e-06,
+      "loss": 1.0158,
+      "step": 6222
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7418289036197487e-06,
+      "loss": 0.0106,
+      "step": 6223
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7417870809895235e-06,
+      "loss": 0.0017,
+      "step": 6224
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7417452583592983e-06,
+      "loss": 0.4024,
+      "step": 6225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7417034357290729e-06,
+      "loss": 0.0012,
+      "step": 6226
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7416616130988477e-06,
+      "loss": 0.0029,
+      "step": 6227
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7416197904686224e-06,
+      "loss": 0.0019,
+      "step": 6228
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7415779678383972e-06,
+      "loss": 0.0036,
+      "step": 6229
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741536145208172e-06,
+      "loss": 0.0101,
+      "step": 6230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7414943225779468e-06,
+      "loss": 3.2918,
+      "step": 6231
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7414524999477216e-06,
+      "loss": 0.0018,
+      "step": 6232
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7414106773174964e-06,
+      "loss": 0.0039,
+      "step": 6233
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7413688546872712e-06,
+      "loss": 1.4453,
+      "step": 6234
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741327032057046e-06,
+      "loss": 0.002,
+      "step": 6235
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7412852094268207e-06,
+      "loss": 0.0087,
+      "step": 6236
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7412433867965957e-06,
+      "loss": 0.001,
+      "step": 6237
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7412015641663705e-06,
+      "loss": 0.5196,
+      "step": 6238
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7411597415361453e-06,
+      "loss": 1.6915,
+      "step": 6239
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7411179189059199e-06,
+      "loss": 0.4926,
+      "step": 6240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7410760962756947e-06,
+      "loss": 0.0016,
+      "step": 6241
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7410342736454694e-06,
+      "loss": 0.0016,
+      "step": 6242
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7409924510152442e-06,
+      "loss": 0.6424,
+      "step": 6243
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740950628385019e-06,
+      "loss": 0.0008,
+      "step": 6244
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7409088057547938e-06,
+      "loss": 0.006,
+      "step": 6245
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7408669831245686e-06,
+      "loss": 0.0011,
+      "step": 6246
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7408251604943434e-06,
+      "loss": 0.0008,
+      "step": 6247
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7407833378641182e-06,
+      "loss": 0.0019,
+      "step": 6248
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740741515233893e-06,
+      "loss": 0.0026,
+      "step": 6249
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7406996926036677e-06,
+      "loss": 2.0191,
+      "step": 6250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7406578699734427e-06,
+      "loss": 0.4645,
+      "step": 6251
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7406160473432175e-06,
+      "loss": 0.0015,
+      "step": 6252
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740574224712992e-06,
+      "loss": 0.0086,
+      "step": 6253
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7405324020827669e-06,
+      "loss": 0.5366,
+      "step": 6254
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7404905794525417e-06,
+      "loss": 0.0725,
+      "step": 6255
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7404487568223165e-06,
+      "loss": 0.0013,
+      "step": 6256
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7404069341920912e-06,
+      "loss": 0.0676,
+      "step": 6257
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740365111561866e-06,
+      "loss": 0.0019,
+      "step": 6258
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7403232889316408e-06,
+      "loss": 0.001,
+      "step": 6259
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7402814663014156e-06,
+      "loss": 0.2317,
+      "step": 6260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7402396436711904e-06,
+      "loss": 0.0018,
+      "step": 6261
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7401978210409652e-06,
+      "loss": 0.0155,
+      "step": 6262
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.74015599841074e-06,
+      "loss": 0.0008,
+      "step": 6263
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7401141757805147e-06,
+      "loss": 0.0006,
+      "step": 6264
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7400723531502895e-06,
+      "loss": 0.001,
+      "step": 6265
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7400305305200645e-06,
+      "loss": 0.0022,
+      "step": 6266
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7399887078898391e-06,
+      "loss": 0.0008,
+      "step": 6267
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7399468852596139e-06,
+      "loss": 0.0012,
+      "step": 6268
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7399050626293887e-06,
+      "loss": 0.4563,
+      "step": 6269
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7398632399991635e-06,
+      "loss": 0.027,
+      "step": 6270
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7398214173689383e-06,
+      "loss": 0.058,
+      "step": 6271
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739779594738713e-06,
+      "loss": 0.0007,
+      "step": 6272
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7397377721084878e-06,
+      "loss": 0.0027,
+      "step": 6273
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7396959494782626e-06,
+      "loss": 0.0014,
+      "step": 6274
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7396541268480374e-06,
+      "loss": 0.407,
+      "step": 6275
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7396123042178122e-06,
+      "loss": 0.7571,
+      "step": 6276
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739570481587587e-06,
+      "loss": 0.01,
+      "step": 6277
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7395286589573618e-06,
+      "loss": 0.0027,
+      "step": 6278
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7394868363271365e-06,
+      "loss": 0.0038,
+      "step": 6279
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7394450136969111e-06,
+      "loss": 0.0017,
+      "step": 6280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7394031910666861e-06,
+      "loss": 0.0011,
+      "step": 6281
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739361368436461e-06,
+      "loss": 0.0025,
+      "step": 6282
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7393195458062357e-06,
+      "loss": 0.0009,
+      "step": 6283
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7392777231760105e-06,
+      "loss": 0.0018,
+      "step": 6284
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7392359005457853e-06,
+      "loss": 1.7021,
+      "step": 6285
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.73919407791556e-06,
+      "loss": 0.007,
+      "step": 6286
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7391522552853348e-06,
+      "loss": 0.008,
+      "step": 6287
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7391104326551096e-06,
+      "loss": 0.0015,
+      "step": 6288
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7390686100248844e-06,
+      "loss": 0.0006,
+      "step": 6289
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7390267873946592e-06,
+      "loss": 0.6851,
+      "step": 6290
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738984964764434e-06,
+      "loss": 0.0024,
+      "step": 6291
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7389431421342088e-06,
+      "loss": 0.0046,
+      "step": 6292
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7389013195039833e-06,
+      "loss": 0.0011,
+      "step": 6293
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7388594968737581e-06,
+      "loss": 0.6379,
+      "step": 6294
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7388176742435331e-06,
+      "loss": 0.2298,
+      "step": 6295
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738775851613308e-06,
+      "loss": 0.0068,
+      "step": 6296
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7387340289830827e-06,
+      "loss": 1.7927,
+      "step": 6297
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7386922063528575e-06,
+      "loss": 0.0008,
+      "step": 6298
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7386503837226323e-06,
+      "loss": 0.0006,
+      "step": 6299
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738608561092407e-06,
+      "loss": 3.611,
+      "step": 6300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7385667384621818e-06,
+      "loss": 0.0075,
+      "step": 6301
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7385249158319566e-06,
+      "loss": 0.0028,
+      "step": 6302
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7384830932017314e-06,
+      "loss": 0.0007,
+      "step": 6303
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7384412705715062e-06,
+      "loss": 3.288,
+      "step": 6304
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.738399447941281e-06,
+      "loss": 0.6763,
+      "step": 6305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7383576253110558e-06,
+      "loss": 0.0161,
+      "step": 6306
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7383158026808303e-06,
+      "loss": 0.0005,
+      "step": 6307
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7382739800506051e-06,
+      "loss": 0.0067,
+      "step": 6308
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.73823215742038e-06,
+      "loss": 0.2167,
+      "step": 6309
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.738190334790155e-06,
+      "loss": 1.7261,
+      "step": 6310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7381485121599297e-06,
+      "loss": 0.6111,
+      "step": 6311
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7381066895297045e-06,
+      "loss": 0.001,
+      "step": 6312
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7380648668994793e-06,
+      "loss": 0.8585,
+      "step": 6313
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.738023044269254e-06,
+      "loss": 0.3224,
+      "step": 6314
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7379812216390289e-06,
+      "loss": 0.1044,
+      "step": 6315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7379393990088036e-06,
+      "loss": 0.0053,
+      "step": 6316
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7378975763785784e-06,
+      "loss": 0.5967,
+      "step": 6317
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7378557537483532e-06,
+      "loss": 0.6151,
+      "step": 6318
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.737813931118128e-06,
+      "loss": 0.6792,
+      "step": 6319
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7377721084879028e-06,
+      "loss": 1.632,
+      "step": 6320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7377302858576774e-06,
+      "loss": 0.7484,
+      "step": 6321
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7376884632274521e-06,
+      "loss": 0.296,
+      "step": 6322
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.737646640597227e-06,
+      "loss": 0.0288,
+      "step": 6323
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7376048179670017e-06,
+      "loss": 0.0008,
+      "step": 6324
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7375629953367767e-06,
+      "loss": 0.001,
+      "step": 6325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7375211727065515e-06,
+      "loss": 0.0018,
+      "step": 6326
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7374793500763263e-06,
+      "loss": 0.0013,
+      "step": 6327
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.737437527446101e-06,
+      "loss": 0.001,
+      "step": 6328
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7373957048158759e-06,
+      "loss": 0.0025,
+      "step": 6329
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7373538821856506e-06,
+      "loss": 0.0033,
+      "step": 6330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7373120595554254e-06,
+      "loss": 0.0008,
+      "step": 6331
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7372702369252002e-06,
+      "loss": 0.0173,
+      "step": 6332
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.737228414294975e-06,
+      "loss": 1.67,
+      "step": 6333
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7371865916647496e-06,
+      "loss": 0.0025,
+      "step": 6334
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7371447690345244e-06,
+      "loss": 0.3292,
+      "step": 6335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7371029464042992e-06,
+      "loss": 0.0031,
+      "step": 6336
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.737061123774074e-06,
+      "loss": 0.0075,
+      "step": 6337
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7370193011438487e-06,
+      "loss": 0.0084,
+      "step": 6338
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7369774785136237e-06,
+      "loss": 2.7652,
+      "step": 6339
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7369356558833985e-06,
+      "loss": 0.007,
+      "step": 6340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7368938332531733e-06,
+      "loss": 0.0021,
+      "step": 6341
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.736852010622948e-06,
+      "loss": 0.6108,
+      "step": 6342
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7368101879927229e-06,
+      "loss": 0.0012,
+      "step": 6343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7367683653624977e-06,
+      "loss": 1.6098,
+      "step": 6344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7367265427322724e-06,
+      "loss": 0.0014,
+      "step": 6345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7366847201020472e-06,
+      "loss": 0.0008,
+      "step": 6346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.736642897471822e-06,
+      "loss": 0.0579,
+      "step": 6347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7366010748415966e-06,
+      "loss": 0.2328,
+      "step": 6348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7365592522113714e-06,
+      "loss": 0.0132,
+      "step": 6349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7365174295811462e-06,
+      "loss": 0.0471,
+      "step": 6350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.736475606950921e-06,
+      "loss": 0.008,
+      "step": 6351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7364337843206957e-06,
+      "loss": 0.0008,
+      "step": 6352
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7363919616904705e-06,
+      "loss": 0.0032,
+      "step": 6353
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7363501390602455e-06,
+      "loss": 0.0048,
+      "step": 6354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7363083164300203e-06,
+      "loss": 2.7784,
+      "step": 6355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.736266493799795e-06,
+      "loss": 3.0107,
+      "step": 6356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7362246711695699e-06,
+      "loss": 0.0026,
+      "step": 6357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7361828485393447e-06,
+      "loss": 0.0019,
+      "step": 6358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7361410259091194e-06,
+      "loss": 0.0023,
+      "step": 6359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7360992032788942e-06,
+      "loss": 0.0052,
+      "step": 6360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7360573806486688e-06,
+      "loss": 2.4873,
+      "step": 6361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7360155580184436e-06,
+      "loss": 0.0444,
+      "step": 6362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7359737353882184e-06,
+      "loss": 0.0015,
+      "step": 6363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7359319127579932e-06,
+      "loss": 0.0008,
+      "step": 6364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735890090127768e-06,
+      "loss": 0.001,
+      "step": 6365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7358482674975427e-06,
+      "loss": 2.4006,
+      "step": 6366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7358064448673175e-06,
+      "loss": 0.0019,
+      "step": 6367
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7357646222370923e-06,
+      "loss": 0.0031,
+      "step": 6368
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7357227996068673e-06,
+      "loss": 0.0016,
+      "step": 6369
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735680976976642e-06,
+      "loss": 0.0015,
+      "step": 6370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7356391543464169e-06,
+      "loss": 0.0012,
+      "step": 6371
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7355973317161917e-06,
+      "loss": 0.0055,
+      "step": 6372
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7355555090859665e-06,
+      "loss": 0.0012,
+      "step": 6373
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7355136864557412e-06,
+      "loss": 0.0018,
+      "step": 6374
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7354718638255158e-06,
+      "loss": 0.0007,
+      "step": 6375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7354300411952906e-06,
+      "loss": 0.0007,
+      "step": 6376
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7353882185650654e-06,
+      "loss": 0.0021,
+      "step": 6377
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7353463959348402e-06,
+      "loss": 0.0065,
+      "step": 6378
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735304573304615e-06,
+      "loss": 0.2145,
+      "step": 6379
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7352627506743898e-06,
+      "loss": 0.0011,
+      "step": 6380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7352209280441645e-06,
+      "loss": 0.7005,
+      "step": 6381
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7351791054139393e-06,
+      "loss": 0.0072,
+      "step": 6382
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7351372827837143e-06,
+      "loss": 0.0101,
+      "step": 6383
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7350954601534891e-06,
+      "loss": 0.134,
+      "step": 6384
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7350536375232639e-06,
+      "loss": 0.0043,
+      "step": 6385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7350118148930387e-06,
+      "loss": 2.2388,
+      "step": 6386
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7349699922628135e-06,
+      "loss": 0.0054,
+      "step": 6387
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734928169632588e-06,
+      "loss": 0.0542,
+      "step": 6388
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7348863470023628e-06,
+      "loss": 0.0037,
+      "step": 6389
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7348445243721376e-06,
+      "loss": 0.422,
+      "step": 6390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7348027017419124e-06,
+      "loss": 0.0009,
+      "step": 6391
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7347608791116872e-06,
+      "loss": 0.0016,
+      "step": 6392
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734719056481462e-06,
+      "loss": 0.0021,
+      "step": 6393
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7346772338512368e-06,
+      "loss": 0.0014,
+      "step": 6394
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7346354112210115e-06,
+      "loss": 0.2837,
+      "step": 6395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7345935885907863e-06,
+      "loss": 0.002,
+      "step": 6396
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7345517659605611e-06,
+      "loss": 0.002,
+      "step": 6397
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7345099433303361e-06,
+      "loss": 0.0014,
+      "step": 6398
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734468120700111e-06,
+      "loss": 0.0014,
+      "step": 6399
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7344262980698857e-06,
+      "loss": 0.0029,
+      "step": 6400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7343844754396605e-06,
+      "loss": 0.0009,
+      "step": 6401
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734342652809435e-06,
+      "loss": 0.1249,
+      "step": 6402
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7343008301792098e-06,
+      "loss": 2.3598,
+      "step": 6403
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7342590075489846e-06,
+      "loss": 0.0008,
+      "step": 6404
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7342171849187594e-06,
+      "loss": 0.2029,
+      "step": 6405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7341753622885342e-06,
+      "loss": 0.0031,
+      "step": 6406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734133539658309e-06,
+      "loss": 0.0011,
+      "step": 6407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7340917170280838e-06,
+      "loss": 0.0028,
+      "step": 6408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7340498943978586e-06,
+      "loss": 1.0139,
+      "step": 6409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7340080717676333e-06,
+      "loss": 0.0044,
+      "step": 6410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7339662491374081e-06,
+      "loss": 0.0015,
+      "step": 6411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7339244265071831e-06,
+      "loss": 0.002,
+      "step": 6412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733882603876958e-06,
+      "loss": 0.0014,
+      "step": 6413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7338407812467327e-06,
+      "loss": 0.0344,
+      "step": 6414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7337989586165075e-06,
+      "loss": 0.001,
+      "step": 6415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733757135986282e-06,
+      "loss": 0.0029,
+      "step": 6416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7337153133560568e-06,
+      "loss": 0.0006,
+      "step": 6417
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7336734907258316e-06,
+      "loss": 0.0011,
+      "step": 6418
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7336316680956064e-06,
+      "loss": 0.0479,
+      "step": 6419
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7335898454653812e-06,
+      "loss": 0.1092,
+      "step": 6420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733548022835156e-06,
+      "loss": 0.1462,
+      "step": 6421
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7335062002049308e-06,
+      "loss": 1.3735,
+      "step": 6422
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7334643775747056e-06,
+      "loss": 0.7168,
+      "step": 6423
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7334225549444803e-06,
+      "loss": 0.001,
+      "step": 6424
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7333807323142551e-06,
+      "loss": 3.9686,
+      "step": 6425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.73333890968403e-06,
+      "loss": 0.0008,
+      "step": 6426
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733297087053805e-06,
+      "loss": 0.001,
+      "step": 6427
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7332552644235797e-06,
+      "loss": 0.0039,
+      "step": 6428
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7332134417933543e-06,
+      "loss": 0.0083,
+      "step": 6429
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733171619163129e-06,
+      "loss": 0.0023,
+      "step": 6430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7331297965329039e-06,
+      "loss": 0.0029,
+      "step": 6431
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7330879739026786e-06,
+      "loss": 0.0016,
+      "step": 6432
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7330461512724534e-06,
+      "loss": 0.4835,
+      "step": 6433
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7330043286422282e-06,
+      "loss": 0.0045,
+      "step": 6434
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732962506012003e-06,
+      "loss": 0.5633,
+      "step": 6435
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7329206833817778e-06,
+      "loss": 0.0007,
+      "step": 6436
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7328788607515526e-06,
+      "loss": 0.0125,
+      "step": 6437
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7328370381213274e-06,
+      "loss": 0.0041,
+      "step": 6438
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7327952154911021e-06,
+      "loss": 0.0039,
+      "step": 6439
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732753392860877e-06,
+      "loss": 0.0014,
+      "step": 6440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7327115702306517e-06,
+      "loss": 0.0008,
+      "step": 6441
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7326697476004267e-06,
+      "loss": 0.0389,
+      "step": 6442
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7326279249702013e-06,
+      "loss": 1.9216,
+      "step": 6443
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732586102339976e-06,
+      "loss": 0.0023,
+      "step": 6444
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7325442797097509e-06,
+      "loss": 3.2177,
+      "step": 6445
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7325024570795256e-06,
+      "loss": 0.4628,
+      "step": 6446
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7324606344493004e-06,
+      "loss": 3.1642,
+      "step": 6447
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7324188118190752e-06,
+      "loss": 0.5444,
+      "step": 6448
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.73237698918885e-06,
+      "loss": 0.5248,
+      "step": 6449
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7323351665586248e-06,
+      "loss": 0.002,
+      "step": 6450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7322933439283996e-06,
+      "loss": 0.0734,
+      "step": 6451
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7322515212981744e-06,
+      "loss": 0.0026,
+      "step": 6452
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7322096986679492e-06,
+      "loss": 0.0146,
+      "step": 6453
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732167876037724e-06,
+      "loss": 0.0017,
+      "step": 6454
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7321260534074987e-06,
+      "loss": 0.0695,
+      "step": 6455
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7320842307772735e-06,
+      "loss": 3.2895,
+      "step": 6456
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7320424081470483e-06,
+      "loss": 1.619,
+      "step": 6457
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732000585516823e-06,
+      "loss": 0.0074,
+      "step": 6458
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7319587628865979e-06,
+      "loss": 0.0018,
+      "step": 6459
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7319169402563727e-06,
+      "loss": 1.2552,
+      "step": 6460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7318751176261474e-06,
+      "loss": 0.0019,
+      "step": 6461
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7318332949959222e-06,
+      "loss": 0.0025,
+      "step": 6462
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.731791472365697e-06,
+      "loss": 0.0047,
+      "step": 6463
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7317496497354718e-06,
+      "loss": 0.3909,
+      "step": 6464
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7317078271052466e-06,
+      "loss": 0.0038,
+      "step": 6465
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7316660044750214e-06,
+      "loss": 0.0088,
+      "step": 6466
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7316241818447962e-06,
+      "loss": 0.0016,
+      "step": 6467
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.731582359214571e-06,
+      "loss": 0.0091,
+      "step": 6468
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7315405365843455e-06,
+      "loss": 1.5758,
+      "step": 6469
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7314987139541203e-06,
+      "loss": 0.0011,
+      "step": 6470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7314568913238953e-06,
+      "loss": 0.1137,
+      "step": 6471
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.73141506869367e-06,
+      "loss": 0.2103,
+      "step": 6472
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7313732460634449e-06,
+      "loss": 0.3745,
+      "step": 6473
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7313314234332197e-06,
+      "loss": 0.0011,
+      "step": 6474
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7312896008029945e-06,
+      "loss": 4.1518,
+      "step": 6475
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7312477781727692e-06,
+      "loss": 0.3998,
+      "step": 6476
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.731205955542544e-06,
+      "loss": 0.0033,
+      "step": 6477
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7311641329123188e-06,
+      "loss": 0.0013,
+      "step": 6478
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7311223102820936e-06,
+      "loss": 0.5006,
+      "step": 6479
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7310804876518684e-06,
+      "loss": 0.1356,
+      "step": 6480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7310386650216432e-06,
+      "loss": 0.0019,
+      "step": 6481
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.730996842391418e-06,
+      "loss": 0.0036,
+      "step": 6482
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7309550197611925e-06,
+      "loss": 0.0027,
+      "step": 6483
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7309131971309673e-06,
+      "loss": 0.0102,
+      "step": 6484
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.730871374500742e-06,
+      "loss": 0.004,
+      "step": 6485
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.730829551870517e-06,
+      "loss": 0.3019,
+      "step": 6486
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7307877292402919e-06,
+      "loss": 0.0176,
+      "step": 6487
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7307459066100667e-06,
+      "loss": 1.4661,
+      "step": 6488
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7307040839798415e-06,
+      "loss": 0.0015,
+      "step": 6489
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7306622613496162e-06,
+      "loss": 1.3841,
+      "step": 6490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.730620438719391e-06,
+      "loss": 0.0048,
+      "step": 6491
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7305786160891658e-06,
+      "loss": 0.0013,
+      "step": 6492
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7305367934589406e-06,
+      "loss": 2.0465,
+      "step": 6493
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7304949708287154e-06,
+      "loss": 0.0029,
+      "step": 6494
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7304531481984902e-06,
+      "loss": 0.0013,
+      "step": 6495
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7304113255682648e-06,
+      "loss": 0.0014,
+      "step": 6496
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7303695029380395e-06,
+      "loss": 0.0011,
+      "step": 6497
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7303276803078143e-06,
+      "loss": 0.0022,
+      "step": 6498
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7302858576775891e-06,
+      "loss": 0.5076,
+      "step": 6499
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7302440350473641e-06,
+      "loss": 0.0016,
+      "step": 6500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.730202212417139e-06,
+      "loss": 0.0839,
+      "step": 6501
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7301603897869137e-06,
+      "loss": 0.0048,
+      "step": 6502
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7301185671566885e-06,
+      "loss": 0.0008,
+      "step": 6503
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7300767445264633e-06,
+      "loss": 0.0008,
+      "step": 6504
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.730034921896238e-06,
+      "loss": 0.2326,
+      "step": 6505
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7299930992660128e-06,
+      "loss": 2.0982,
+      "step": 6506
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7299512766357876e-06,
+      "loss": 0.0025,
+      "step": 6507
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7299094540055624e-06,
+      "loss": 0.0013,
+      "step": 6508
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7298676313753372e-06,
+      "loss": 0.0032,
+      "step": 6509
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7298258087451118e-06,
+      "loss": 0.0009,
+      "step": 6510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7297839861148865e-06,
+      "loss": 0.0017,
+      "step": 6511
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7297421634846613e-06,
+      "loss": 0.0199,
+      "step": 6512
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7297003408544361e-06,
+      "loss": 0.0042,
+      "step": 6513
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729658518224211e-06,
+      "loss": 1.0241,
+      "step": 6514
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729616695593986e-06,
+      "loss": 0.01,
+      "step": 6515
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7295748729637607e-06,
+      "loss": 0.0015,
+      "step": 6516
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7295330503335355e-06,
+      "loss": 0.0008,
+      "step": 6517
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7294912277033103e-06,
+      "loss": 0.027,
+      "step": 6518
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729449405073085e-06,
+      "loss": 0.2077,
+      "step": 6519
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7294075824428598e-06,
+      "loss": 0.0011,
+      "step": 6520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7293657598126346e-06,
+      "loss": 0.0039,
+      "step": 6521
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7293239371824094e-06,
+      "loss": 0.0015,
+      "step": 6522
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7292821145521842e-06,
+      "loss": 0.1881,
+      "step": 6523
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7292402919219588e-06,
+      "loss": 2.7236,
+      "step": 6524
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7291984692917336e-06,
+      "loss": 0.0013,
+      "step": 6525
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7291566466615083e-06,
+      "loss": 0.1718,
+      "step": 6526
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7291148240312831e-06,
+      "loss": 0.0078,
+      "step": 6527
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729073001401058e-06,
+      "loss": 1.5089,
+      "step": 6528
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7290311787708327e-06,
+      "loss": 0.0658,
+      "step": 6529
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7289893561406077e-06,
+      "loss": 0.0012,
+      "step": 6530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7289475335103825e-06,
+      "loss": 0.0014,
+      "step": 6531
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7289057108801573e-06,
+      "loss": 1.113,
+      "step": 6532
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728863888249932e-06,
+      "loss": 0.0176,
+      "step": 6533
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7288220656197068e-06,
+      "loss": 0.0024,
+      "step": 6534
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7287802429894816e-06,
+      "loss": 0.2528,
+      "step": 6535
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7287384203592564e-06,
+      "loss": 2.0253,
+      "step": 6536
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728696597729031e-06,
+      "loss": 0.0018,
+      "step": 6537
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7286547750988058e-06,
+      "loss": 0.0024,
+      "step": 6538
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7286129524685806e-06,
+      "loss": 0.0021,
+      "step": 6539
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7285711298383554e-06,
+      "loss": 0.0047,
+      "step": 6540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7285293072081301e-06,
+      "loss": 0.0065,
+      "step": 6541
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728487484577905e-06,
+      "loss": 0.0017,
+      "step": 6542
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7284456619476797e-06,
+      "loss": 0.0007,
+      "step": 6543
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7284038393174547e-06,
+      "loss": 0.0008,
+      "step": 6544
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7283620166872295e-06,
+      "loss": 0.0034,
+      "step": 6545
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7283201940570043e-06,
+      "loss": 0.0006,
+      "step": 6546
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728278371426779e-06,
+      "loss": 0.0008,
+      "step": 6547
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7282365487965539e-06,
+      "loss": 0.9323,
+      "step": 6548
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7281947261663286e-06,
+      "loss": 2.3948,
+      "step": 6549
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7281529035361034e-06,
+      "loss": 0.001,
+      "step": 6550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728111080905878e-06,
+      "loss": 0.0047,
+      "step": 6551
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7280692582756528e-06,
+      "loss": 0.0016,
+      "step": 6552
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7280274356454276e-06,
+      "loss": 0.5361,
+      "step": 6553
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7279856130152024e-06,
+      "loss": 0.0023,
+      "step": 6554
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7279437903849771e-06,
+      "loss": 0.0006,
+      "step": 6555
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727901967754752e-06,
+      "loss": 0.059,
+      "step": 6556
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7278601451245267e-06,
+      "loss": 0.0012,
+      "step": 6557
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7278183224943015e-06,
+      "loss": 0.0025,
+      "step": 6558
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7277764998640765e-06,
+      "loss": 2.5013,
+      "step": 6559
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7277346772338513e-06,
+      "loss": 0.0015,
+      "step": 6560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727692854603626e-06,
+      "loss": 0.0014,
+      "step": 6561
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7276510319734009e-06,
+      "loss": 0.2895,
+      "step": 6562
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7276092093431756e-06,
+      "loss": 0.0037,
+      "step": 6563
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7275673867129502e-06,
+      "loss": 0.0185,
+      "step": 6564
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727525564082725e-06,
+      "loss": 0.8245,
+      "step": 6565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7274837414524998e-06,
+      "loss": 0.0007,
+      "step": 6566
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7274419188222746e-06,
+      "loss": 0.0025,
+      "step": 6567
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7274000961920494e-06,
+      "loss": 0.0031,
+      "step": 6568
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7273582735618242e-06,
+      "loss": 0.0014,
+      "step": 6569
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727316450931599e-06,
+      "loss": 0.0008,
+      "step": 6570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7272746283013737e-06,
+      "loss": 0.0013,
+      "step": 6571
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7272328056711485e-06,
+      "loss": 0.0021,
+      "step": 6572
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7271909830409233e-06,
+      "loss": 0.0844,
+      "step": 6573
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7271491604106983e-06,
+      "loss": 0.0026,
+      "step": 6574
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727107337780473e-06,
+      "loss": 0.0014,
+      "step": 6575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7270655151502479e-06,
+      "loss": 0.0041,
+      "step": 6576
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7270236925200227e-06,
+      "loss": 0.0053,
+      "step": 6577
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7269818698897972e-06,
+      "loss": 0.0013,
+      "step": 6578
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726940047259572e-06,
+      "loss": 0.0234,
+      "step": 6579
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7268982246293468e-06,
+      "loss": 0.0201,
+      "step": 6580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7268564019991216e-06,
+      "loss": 0.0123,
+      "step": 6581
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7268145793688964e-06,
+      "loss": 0.0022,
+      "step": 6582
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7267727567386712e-06,
+      "loss": 0.0021,
+      "step": 6583
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726730934108446e-06,
+      "loss": 0.001,
+      "step": 6584
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7266891114782207e-06,
+      "loss": 0.0011,
+      "step": 6585
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7266472888479955e-06,
+      "loss": 0.3215,
+      "step": 6586
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7266054662177703e-06,
+      "loss": 0.0007,
+      "step": 6587
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7265636435875453e-06,
+      "loss": 0.0035,
+      "step": 6588
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.72652182095732e-06,
+      "loss": 3.6135,
+      "step": 6589
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7264799983270949e-06,
+      "loss": 0.9995,
+      "step": 6590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7264381756968695e-06,
+      "loss": 0.0289,
+      "step": 6591
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7263963530666442e-06,
+      "loss": 1.5324,
+      "step": 6592
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726354530436419e-06,
+      "loss": 0.0875,
+      "step": 6593
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7263127078061938e-06,
+      "loss": 0.0008,
+      "step": 6594
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7262708851759686e-06,
+      "loss": 0.1863,
+      "step": 6595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7262290625457434e-06,
+      "loss": 0.4934,
+      "step": 6596
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7261872399155182e-06,
+      "loss": 0.0174,
+      "step": 6597
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726145417285293e-06,
+      "loss": 0.0009,
+      "step": 6598
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7261035946550677e-06,
+      "loss": 0.0014,
+      "step": 6599
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7260617720248425e-06,
+      "loss": 0.0014,
+      "step": 6600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7260199493946173e-06,
+      "loss": 3.7236,
+      "step": 6601
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725978126764392e-06,
+      "loss": 0.0007,
+      "step": 6602
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725936304134167e-06,
+      "loss": 0.6084,
+      "step": 6603
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7258944815039419e-06,
+      "loss": 0.0023,
+      "step": 6604
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7258526588737165e-06,
+      "loss": 3.265,
+      "step": 6605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7258108362434912e-06,
+      "loss": 0.094,
+      "step": 6606
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725769013613266e-06,
+      "loss": 1.2585,
+      "step": 6607
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7257271909830408e-06,
+      "loss": 0.0063,
+      "step": 6608
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7256853683528156e-06,
+      "loss": 2.3362,
+      "step": 6609
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7256435457225904e-06,
+      "loss": 3.3235,
+      "step": 6610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7256017230923652e-06,
+      "loss": 0.0047,
+      "step": 6611
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.72555990046214e-06,
+      "loss": 0.0015,
+      "step": 6612
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7255180778319148e-06,
+      "loss": 0.0013,
+      "step": 6613
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7254762552016895e-06,
+      "loss": 0.0032,
+      "step": 6614
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7254344325714643e-06,
+      "loss": 0.002,
+      "step": 6615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7253926099412391e-06,
+      "loss": 0.0093,
+      "step": 6616
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7253507873110141e-06,
+      "loss": 0.0174,
+      "step": 6617
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725308964680789e-06,
+      "loss": 0.0024,
+      "step": 6618
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7252671420505635e-06,
+      "loss": 0.5259,
+      "step": 6619
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7252253194203383e-06,
+      "loss": 0.002,
+      "step": 6620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725183496790113e-06,
+      "loss": 0.0008,
+      "step": 6621
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7251416741598878e-06,
+      "loss": 1.6082,
+      "step": 6622
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7250998515296626e-06,
+      "loss": 0.0007,
+      "step": 6623
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7250580288994374e-06,
+      "loss": 0.005,
+      "step": 6624
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7250162062692122e-06,
+      "loss": 0.0221,
+      "step": 6625
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.724974383638987e-06,
+      "loss": 0.0358,
+      "step": 6626
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7249325610087618e-06,
+      "loss": 1.731,
+      "step": 6627
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7248907383785365e-06,
+      "loss": 0.1082,
+      "step": 6628
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7248489157483113e-06,
+      "loss": 0.0014,
+      "step": 6629
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7248070931180861e-06,
+      "loss": 0.0553,
+      "step": 6630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.724765270487861e-06,
+      "loss": 0.0023,
+      "step": 6631
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7247234478576357e-06,
+      "loss": 0.0031,
+      "step": 6632
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7246816252274105e-06,
+      "loss": 0.6761,
+      "step": 6633
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7246398025971853e-06,
+      "loss": 0.0015,
+      "step": 6634
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.72459797996696e-06,
+      "loss": 0.1246,
+      "step": 6635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7245561573367348e-06,
+      "loss": 0.6606,
+      "step": 6636
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7245143347065096e-06,
+      "loss": 0.0016,
+      "step": 6637
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7244725120762844e-06,
+      "loss": 2.4782,
+      "step": 6638
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7244306894460592e-06,
+      "loss": 1.4208,
+      "step": 6639
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.724388866815834e-06,
+      "loss": 0.0009,
+      "step": 6640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7243470441856088e-06,
+      "loss": 0.0381,
+      "step": 6641
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7243052215553836e-06,
+      "loss": 0.0009,
+      "step": 6642
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7242633989251583e-06,
+      "loss": 1.5036,
+      "step": 6643
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7242215762949331e-06,
+      "loss": 0.0044,
+      "step": 6644
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7241797536647077e-06,
+      "loss": 0.0309,
+      "step": 6645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7241379310344825e-06,
+      "loss": 0.0044,
+      "step": 6646
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7240961084042575e-06,
+      "loss": 0.0024,
+      "step": 6647
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7240542857740323e-06,
+      "loss": 0.0017,
+      "step": 6648
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.724012463143807e-06,
+      "loss": 0.0035,
+      "step": 6649
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7239706405135818e-06,
+      "loss": 1.3807,
+      "step": 6650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7239288178833566e-06,
+      "loss": 0.0156,
+      "step": 6651
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7238869952531314e-06,
+      "loss": 2.5097,
+      "step": 6652
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7238451726229062e-06,
+      "loss": 0.9332,
+      "step": 6653
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.723803349992681e-06,
+      "loss": 0.0057,
+      "step": 6654
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7237615273624558e-06,
+      "loss": 0.0052,
+      "step": 6655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7237197047322306e-06,
+      "loss": 2.302,
+      "step": 6656
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7236778821020054e-06,
+      "loss": 0.0141,
+      "step": 6657
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7236360594717801e-06,
+      "loss": 0.0011,
+      "step": 6658
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7235942368415547e-06,
+      "loss": 1.162,
+      "step": 6659
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7235524142113295e-06,
+      "loss": 0.0146,
+      "step": 6660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7235105915811045e-06,
+      "loss": 0.0165,
+      "step": 6661
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7234687689508793e-06,
+      "loss": 0.0037,
+      "step": 6662
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.723426946320654e-06,
+      "loss": 1.9333,
+      "step": 6663
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7233851236904289e-06,
+      "loss": 0.0015,
+      "step": 6664
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7233433010602036e-06,
+      "loss": 0.0027,
+      "step": 6665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7233014784299784e-06,
+      "loss": 0.0008,
+      "step": 6666
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7232596557997532e-06,
+      "loss": 1.2497,
+      "step": 6667
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.723217833169528e-06,
+      "loss": 0.0014,
+      "step": 6668
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7231760105393028e-06,
+      "loss": 0.0115,
+      "step": 6669
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7231341879090776e-06,
+      "loss": 0.0177,
+      "step": 6670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7230923652788524e-06,
+      "loss": 1.0399,
+      "step": 6671
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.723050542648627e-06,
+      "loss": 0.0035,
+      "step": 6672
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7230087200184017e-06,
+      "loss": 0.004,
+      "step": 6673
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7229668973881765e-06,
+      "loss": 0.008,
+      "step": 6674
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7229250747579513e-06,
+      "loss": 0.0031,
+      "step": 6675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7228832521277263e-06,
+      "loss": 0.0009,
+      "step": 6676
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722841429497501e-06,
+      "loss": 1.6571,
+      "step": 6677
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7227996068672759e-06,
+      "loss": 0.0061,
+      "step": 6678
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7227577842370507e-06,
+      "loss": 0.0075,
+      "step": 6679
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7227159616068254e-06,
+      "loss": 0.0595,
+      "step": 6680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7226741389766002e-06,
+      "loss": 0.3312,
+      "step": 6681
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722632316346375e-06,
+      "loss": 0.283,
+      "step": 6682
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7225904937161498e-06,
+      "loss": 0.001,
+      "step": 6683
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7225486710859246e-06,
+      "loss": 0.0017,
+      "step": 6684
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7225068484556994e-06,
+      "loss": 0.0062,
+      "step": 6685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722465025825474e-06,
+      "loss": 2.0396,
+      "step": 6686
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7224232031952487e-06,
+      "loss": 1.5431,
+      "step": 6687
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7223813805650235e-06,
+      "loss": 0.0099,
+      "step": 6688
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7223395579347983e-06,
+      "loss": 2.1911,
+      "step": 6689
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722297735304573e-06,
+      "loss": 0.0009,
+      "step": 6690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722255912674348e-06,
+      "loss": 0.0012,
+      "step": 6691
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7222140900441229e-06,
+      "loss": 0.0008,
+      "step": 6692
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7221722674138977e-06,
+      "loss": 0.0022,
+      "step": 6693
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7221304447836724e-06,
+      "loss": 0.0054,
+      "step": 6694
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7220886221534472e-06,
+      "loss": 0.2006,
+      "step": 6695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722046799523222e-06,
+      "loss": 0.0071,
+      "step": 6696
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7220049768929968e-06,
+      "loss": 0.0029,
+      "step": 6697
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7219631542627716e-06,
+      "loss": 0.0018,
+      "step": 6698
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7219213316325462e-06,
+      "loss": 0.0015,
+      "step": 6699
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721879509002321e-06,
+      "loss": 0.0007,
+      "step": 6700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7218376863720957e-06,
+      "loss": 0.002,
+      "step": 6701
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7217958637418705e-06,
+      "loss": 0.0107,
+      "step": 6702
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7217540411116453e-06,
+      "loss": 0.5933,
+      "step": 6703
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.72171221848142e-06,
+      "loss": 1.3473,
+      "step": 6704
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721670395851195e-06,
+      "loss": 0.0074,
+      "step": 6705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7216285732209699e-06,
+      "loss": 0.2277,
+      "step": 6706
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7215867505907447e-06,
+      "loss": 0.0015,
+      "step": 6707
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7215449279605195e-06,
+      "loss": 0.0008,
+      "step": 6708
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7215031053302942e-06,
+      "loss": 0.002,
+      "step": 6709
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721461282700069e-06,
+      "loss": 0.0012,
+      "step": 6710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7214194600698438e-06,
+      "loss": 0.0015,
+      "step": 6711
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7213776374396186e-06,
+      "loss": 0.0016,
+      "step": 6712
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7213358148093932e-06,
+      "loss": 0.0009,
+      "step": 6713
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721293992179168e-06,
+      "loss": 0.0468,
+      "step": 6714
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7212521695489427e-06,
+      "loss": 0.0233,
+      "step": 6715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7212103469187175e-06,
+      "loss": 2.5214,
+      "step": 6716
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7211685242884923e-06,
+      "loss": 0.001,
+      "step": 6717
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7211267016582671e-06,
+      "loss": 0.6338,
+      "step": 6718
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7210848790280419e-06,
+      "loss": 0.0022,
+      "step": 6719
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7210430563978169e-06,
+      "loss": 0.0007,
+      "step": 6720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7210012337675917e-06,
+      "loss": 0.0013,
+      "step": 6721
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7209594111373665e-06,
+      "loss": 0.0008,
+      "step": 6722
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7209175885071412e-06,
+      "loss": 0.1962,
+      "step": 6723
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720875765876916e-06,
+      "loss": 0.0021,
+      "step": 6724
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7208339432466908e-06,
+      "loss": 0.0033,
+      "step": 6725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7207921206164656e-06,
+      "loss": 1.3515,
+      "step": 6726
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7207502979862402e-06,
+      "loss": 0.0256,
+      "step": 6727
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720708475356015e-06,
+      "loss": 0.4161,
+      "step": 6728
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7206666527257898e-06,
+      "loss": 0.001,
+      "step": 6729
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7206248300955645e-06,
+      "loss": 0.0028,
+      "step": 6730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7205830074653393e-06,
+      "loss": 0.0014,
+      "step": 6731
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7205411848351141e-06,
+      "loss": 0.9021,
+      "step": 6732
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720499362204889e-06,
+      "loss": 0.0012,
+      "step": 6733
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7204575395746637e-06,
+      "loss": 0.0014,
+      "step": 6734
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7204157169444387e-06,
+      "loss": 1.7637,
+      "step": 6735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7203738943142135e-06,
+      "loss": 0.0054,
+      "step": 6736
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7203320716839883e-06,
+      "loss": 0.0011,
+      "step": 6737
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720290249053763e-06,
+      "loss": 0.0021,
+      "step": 6738
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7202484264235378e-06,
+      "loss": 0.0009,
+      "step": 6739
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7202066037933124e-06,
+      "loss": 0.0027,
+      "step": 6740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7201647811630872e-06,
+      "loss": 0.0094,
+      "step": 6741
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720122958532862e-06,
+      "loss": 0.0035,
+      "step": 6742
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7200811359026368e-06,
+      "loss": 0.0048,
+      "step": 6743
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7200393132724116e-06,
+      "loss": 0.0007,
+      "step": 6744
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7199974906421863e-06,
+      "loss": 1.1709,
+      "step": 6745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7199556680119611e-06,
+      "loss": 0.2865,
+      "step": 6746
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719913845381736e-06,
+      "loss": 0.0624,
+      "step": 6747
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7198720227515107e-06,
+      "loss": 0.0009,
+      "step": 6748
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7198302001212857e-06,
+      "loss": 0.0006,
+      "step": 6749
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7197883774910605e-06,
+      "loss": 0.3239,
+      "step": 6750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7197465548608353e-06,
+      "loss": 0.0131,
+      "step": 6751
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.71970473223061e-06,
+      "loss": 0.2215,
+      "step": 6752
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7196629096003848e-06,
+      "loss": 2.4357,
+      "step": 6753
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7196210869701594e-06,
+      "loss": 0.0524,
+      "step": 6754
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7195792643399342e-06,
+      "loss": 0.0322,
+      "step": 6755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719537441709709e-06,
+      "loss": 0.0018,
+      "step": 6756
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7194956190794838e-06,
+      "loss": 0.0006,
+      "step": 6757
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7194537964492586e-06,
+      "loss": 0.0008,
+      "step": 6758
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7194119738190333e-06,
+      "loss": 2.2397,
+      "step": 6759
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7193701511888081e-06,
+      "loss": 0.0036,
+      "step": 6760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719328328558583e-06,
+      "loss": 0.3142,
+      "step": 6761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7192865059283577e-06,
+      "loss": 0.0846,
+      "step": 6762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7192446832981325e-06,
+      "loss": 0.007,
+      "step": 6763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7192028606679075e-06,
+      "loss": 0.0019,
+      "step": 6764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7191610380376823e-06,
+      "loss": 0.0392,
+      "step": 6765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719119215407457e-06,
+      "loss": 0.0009,
+      "step": 6766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7190773927772316e-06,
+      "loss": 0.0011,
+      "step": 6767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7190355701470064e-06,
+      "loss": 0.0159,
+      "step": 6768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7189937475167812e-06,
+      "loss": 0.0008,
+      "step": 6769
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718951924886556e-06,
+      "loss": 2.0011,
+      "step": 6770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7189101022563308e-06,
+      "loss": 0.0022,
+      "step": 6771
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7188682796261056e-06,
+      "loss": 2.4591,
+      "step": 6772
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7188264569958804e-06,
+      "loss": 0.0092,
+      "step": 6773
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7187846343656551e-06,
+      "loss": 3.0734,
+      "step": 6774
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.71874281173543e-06,
+      "loss": 0.001,
+      "step": 6775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7187009891052047e-06,
+      "loss": 0.0018,
+      "step": 6776
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7186591664749795e-06,
+      "loss": 0.0015,
+      "step": 6777
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7186173438447543e-06,
+      "loss": 0.025,
+      "step": 6778
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7185755212145293e-06,
+      "loss": 0.0063,
+      "step": 6779
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718533698584304e-06,
+      "loss": 0.944,
+      "step": 6780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7184918759540786e-06,
+      "loss": 0.0013,
+      "step": 6781
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7184500533238534e-06,
+      "loss": 0.0061,
+      "step": 6782
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7184082306936282e-06,
+      "loss": 0.0026,
+      "step": 6783
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.718366408063403e-06,
+      "loss": 0.4262,
+      "step": 6784
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7183245854331778e-06,
+      "loss": 0.0008,
+      "step": 6785
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7182827628029526e-06,
+      "loss": 0.0005,
+      "step": 6786
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7182409401727274e-06,
+      "loss": 0.0078,
+      "step": 6787
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7181991175425021e-06,
+      "loss": 0.5445,
+      "step": 6788
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.718157294912277e-06,
+      "loss": 0.0036,
+      "step": 6789
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7181154722820517e-06,
+      "loss": 0.0039,
+      "step": 6790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7180736496518265e-06,
+      "loss": 0.0016,
+      "step": 6791
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7180318270216013e-06,
+      "loss": 0.0082,
+      "step": 6792
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7179900043913763e-06,
+      "loss": 0.8521,
+      "step": 6793
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7179481817611509e-06,
+      "loss": 0.0008,
+      "step": 6794
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7179063591309257e-06,
+      "loss": 0.0015,
+      "step": 6795
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7178645365007004e-06,
+      "loss": 0.0007,
+      "step": 6796
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7178227138704752e-06,
+      "loss": 0.0281,
+      "step": 6797
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.71778089124025e-06,
+      "loss": 0.7075,
+      "step": 6798
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7177390686100248e-06,
+      "loss": 0.0079,
+      "step": 6799
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7176972459797996e-06,
+      "loss": 0.0012,
+      "step": 6800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7176554233495744e-06,
+      "loss": 0.0044,
+      "step": 6801
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7176136007193492e-06,
+      "loss": 0.0005,
+      "step": 6802
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.717571778089124e-06,
+      "loss": 1.0101,
+      "step": 6803
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7175299554588987e-06,
+      "loss": 0.0011,
+      "step": 6804
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7174881328286735e-06,
+      "loss": 2.266,
+      "step": 6805
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7174463101984483e-06,
+      "loss": 0.0017,
+      "step": 6806
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7174044875682229e-06,
+      "loss": 0.139,
+      "step": 6807
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7173626649379979e-06,
+      "loss": 0.6683,
+      "step": 6808
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7173208423077727e-06,
+      "loss": 0.0009,
+      "step": 6809
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7172790196775474e-06,
+      "loss": 0.001,
+      "step": 6810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7172371970473222e-06,
+      "loss": 0.7182,
+      "step": 6811
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.717195374417097e-06,
+      "loss": 0.0009,
+      "step": 6812
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7171535517868718e-06,
+      "loss": 3.3536,
+      "step": 6813
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7171117291566466e-06,
+      "loss": 0.0037,
+      "step": 6814
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7170699065264214e-06,
+      "loss": 0.0011,
+      "step": 6815
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7170280838961962e-06,
+      "loss": 1.4889,
+      "step": 6816
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.716986261265971e-06,
+      "loss": 0.0007,
+      "step": 6817
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7169444386357457e-06,
+      "loss": 0.0189,
+      "step": 6818
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7169026160055205e-06,
+      "loss": 0.8262,
+      "step": 6819
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7168607933752953e-06,
+      "loss": 0.5151,
+      "step": 6820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7168189707450699e-06,
+      "loss": 0.0032,
+      "step": 6821
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7167771481148449e-06,
+      "loss": 0.0251,
+      "step": 6822
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7167353254846197e-06,
+      "loss": 0.0801,
+      "step": 6823
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7166935028543945e-06,
+      "loss": 0.0005,
+      "step": 6824
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7166516802241692e-06,
+      "loss": 0.0007,
+      "step": 6825
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.716609857593944e-06,
+      "loss": 0.6242,
+      "step": 6826
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7165680349637188e-06,
+      "loss": 0.001,
+      "step": 6827
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7165262123334936e-06,
+      "loss": 0.0065,
+      "step": 6828
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7164843897032684e-06,
+      "loss": 0.3692,
+      "step": 6829
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7164425670730432e-06,
+      "loss": 0.0013,
+      "step": 6830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.716400744442818e-06,
+      "loss": 0.0005,
+      "step": 6831
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7163589218125927e-06,
+      "loss": 0.014,
+      "step": 6832
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7163170991823675e-06,
+      "loss": 0.0006,
+      "step": 6833
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7162752765521423e-06,
+      "loss": 0.5771,
+      "step": 6834
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.716233453921917e-06,
+      "loss": 0.1695,
+      "step": 6835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7161916312916917e-06,
+      "loss": 0.0027,
+      "step": 6836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7161498086614667e-06,
+      "loss": 0.0446,
+      "step": 6837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7161079860312415e-06,
+      "loss": 0.0018,
+      "step": 6838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7160661634010163e-06,
+      "loss": 0.002,
+      "step": 6839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.716024340770791e-06,
+      "loss": 0.0009,
+      "step": 6840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7159825181405658e-06,
+      "loss": 0.0006,
+      "step": 6841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7159406955103406e-06,
+      "loss": 3.0867,
+      "step": 6842
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7158988728801154e-06,
+      "loss": 0.0016,
+      "step": 6843
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7158570502498902e-06,
+      "loss": 0.0006,
+      "step": 6844
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715815227619665e-06,
+      "loss": 0.0012,
+      "step": 6845
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7157734049894398e-06,
+      "loss": 0.0006,
+      "step": 6846
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7157315823592145e-06,
+      "loss": 0.0035,
+      "step": 6847
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7156897597289891e-06,
+      "loss": 3.4215,
+      "step": 6848
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715647937098764e-06,
+      "loss": 2.3289,
+      "step": 6849
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7156061144685387e-06,
+      "loss": 0.0033,
+      "step": 6850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7155642918383135e-06,
+      "loss": 0.0006,
+      "step": 6851
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7155224692080885e-06,
+      "loss": 0.0107,
+      "step": 6852
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7154806465778633e-06,
+      "loss": 0.0031,
+      "step": 6853
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715438823947638e-06,
+      "loss": 0.0035,
+      "step": 6854
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7153970013174128e-06,
+      "loss": 0.4202,
+      "step": 6855
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7153551786871876e-06,
+      "loss": 0.3421,
+      "step": 6856
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7153133560569624e-06,
+      "loss": 0.0133,
+      "step": 6857
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7152715334267372e-06,
+      "loss": 0.7197,
+      "step": 6858
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715229710796512e-06,
+      "loss": 0.0013,
+      "step": 6859
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7151878881662868e-06,
+      "loss": 0.001,
+      "step": 6860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7151460655360616e-06,
+      "loss": 0.0008,
+      "step": 6861
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7151042429058361e-06,
+      "loss": 0.0014,
+      "step": 6862
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715062420275611e-06,
+      "loss": 0.7724,
+      "step": 6863
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7150205976453857e-06,
+      "loss": 0.2063,
+      "step": 6864
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7149787750151605e-06,
+      "loss": 0.0031,
+      "step": 6865
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7149369523849355e-06,
+      "loss": 0.0024,
+      "step": 6866
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7148951297547103e-06,
+      "loss": 0.0011,
+      "step": 6867
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714853307124485e-06,
+      "loss": 0.4308,
+      "step": 6868
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7148114844942598e-06,
+      "loss": 0.0008,
+      "step": 6869
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7147696618640346e-06,
+      "loss": 0.0115,
+      "step": 6870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7147278392338094e-06,
+      "loss": 2.9268,
+      "step": 6871
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7146860166035842e-06,
+      "loss": 0.0012,
+      "step": 6872
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714644193973359e-06,
+      "loss": 0.0008,
+      "step": 6873
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7146023713431338e-06,
+      "loss": 0.0023,
+      "step": 6874
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7145605487129083e-06,
+      "loss": 3.2267,
+      "step": 6875
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7145187260826831e-06,
+      "loss": 0.0014,
+      "step": 6876
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714476903452458e-06,
+      "loss": 0.0014,
+      "step": 6877
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7144350808222327e-06,
+      "loss": 0.0038,
+      "step": 6878
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7143932581920075e-06,
+      "loss": 0.0016,
+      "step": 6879
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7143514355617823e-06,
+      "loss": 0.001,
+      "step": 6880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7143096129315573e-06,
+      "loss": 0.7153,
+      "step": 6881
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714267790301332e-06,
+      "loss": 0.0033,
+      "step": 6882
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7142259676711069e-06,
+      "loss": 0.2099,
+      "step": 6883
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7141841450408816e-06,
+      "loss": 1.0594,
+      "step": 6884
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7141423224106564e-06,
+      "loss": 0.0006,
+      "step": 6885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7141004997804312e-06,
+      "loss": 0.0009,
+      "step": 6886
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714058677150206e-06,
+      "loss": 0.0673,
+      "step": 6887
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7140168545199808e-06,
+      "loss": 0.0697,
+      "step": 6888
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7139750318897554e-06,
+      "loss": 0.0007,
+      "step": 6889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7139332092595301e-06,
+      "loss": 1.5192,
+      "step": 6890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713891386629305e-06,
+      "loss": 0.0044,
+      "step": 6891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7138495639990797e-06,
+      "loss": 0.0037,
+      "step": 6892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7138077413688545e-06,
+      "loss": 0.0028,
+      "step": 6893
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7137659187386293e-06,
+      "loss": 0.0027,
+      "step": 6894
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713724096108404e-06,
+      "loss": 2.1863,
+      "step": 6895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713682273478179e-06,
+      "loss": 0.425,
+      "step": 6896
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7136404508479539e-06,
+      "loss": 0.0031,
+      "step": 6897
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7135986282177286e-06,
+      "loss": 0.0011,
+      "step": 6898
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7135568055875034e-06,
+      "loss": 0.0008,
+      "step": 6899
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7135149829572782e-06,
+      "loss": 0.0024,
+      "step": 6900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713473160327053e-06,
+      "loss": 0.0018,
+      "step": 6901
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7134313376968276e-06,
+      "loss": 0.0009,
+      "step": 6902
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7133895150666024e-06,
+      "loss": 0.0071,
+      "step": 6903
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7133476924363772e-06,
+      "loss": 0.0029,
+      "step": 6904
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713305869806152e-06,
+      "loss": 0.0032,
+      "step": 6905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7132640471759267e-06,
+      "loss": 0.0013,
+      "step": 6906
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7132222245457015e-06,
+      "loss": 0.0014,
+      "step": 6907
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7131804019154763e-06,
+      "loss": 0.0053,
+      "step": 6908
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713138579285251e-06,
+      "loss": 0.0269,
+      "step": 6909
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713096756655026e-06,
+      "loss": 0.0007,
+      "step": 6910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7130549340248009e-06,
+      "loss": 0.0024,
+      "step": 6911
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7130131113945757e-06,
+      "loss": 0.0052,
+      "step": 6912
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7129712887643504e-06,
+      "loss": 0.0097,
+      "step": 6913
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7129294661341252e-06,
+      "loss": 1.3144,
+      "step": 6914
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7128876435039e-06,
+      "loss": 0.0012,
+      "step": 6915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7128458208736746e-06,
+      "loss": 0.0018,
+      "step": 6916
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7128039982434494e-06,
+      "loss": 3.3416,
+      "step": 6917
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7127621756132242e-06,
+      "loss": 0.002,
+      "step": 6918
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712720352982999e-06,
+      "loss": 0.0006,
+      "step": 6919
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7126785303527737e-06,
+      "loss": 1.9617,
+      "step": 6920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7126367077225485e-06,
+      "loss": 0.0009,
+      "step": 6921
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7125948850923233e-06,
+      "loss": 0.0009,
+      "step": 6922
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712553062462098e-06,
+      "loss": 0.1468,
+      "step": 6923
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7125112398318729e-06,
+      "loss": 1.8185,
+      "step": 6924
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7124694172016479e-06,
+      "loss": 0.0005,
+      "step": 6925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7124275945714227e-06,
+      "loss": 0.0008,
+      "step": 6926
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7123857719411974e-06,
+      "loss": 2.6373,
+      "step": 6927
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7123439493109722e-06,
+      "loss": 1.33,
+      "step": 6928
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712302126680747e-06,
+      "loss": 0.2636,
+      "step": 6929
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7122603040505216e-06,
+      "loss": 0.0012,
+      "step": 6930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7122184814202964e-06,
+      "loss": 0.4555,
+      "step": 6931
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7121766587900712e-06,
+      "loss": 0.0053,
+      "step": 6932
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712134836159846e-06,
+      "loss": 0.0008,
+      "step": 6933
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7120930135296207e-06,
+      "loss": 0.001,
+      "step": 6934
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7120511908993955e-06,
+      "loss": 0.3559,
+      "step": 6935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7120093682691703e-06,
+      "loss": 0.0015,
+      "step": 6936
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.711967545638945e-06,
+      "loss": 0.0106,
+      "step": 6937
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7119257230087199e-06,
+      "loss": 1.8432,
+      "step": 6938
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7118839003784947e-06,
+      "loss": 1.9697,
+      "step": 6939
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7118420777482697e-06,
+      "loss": 0.0012,
+      "step": 6940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7118002551180445e-06,
+      "loss": 0.0012,
+      "step": 6941
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7117584324878192e-06,
+      "loss": 0.0009,
+      "step": 6942
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7117166098575938e-06,
+      "loss": 0.0042,
+      "step": 6943
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7116747872273686e-06,
+      "loss": 0.001,
+      "step": 6944
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7116329645971434e-06,
+      "loss": 0.1895,
+      "step": 6945
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7115911419669182e-06,
+      "loss": 2.4685,
+      "step": 6946
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.711549319336693e-06,
+      "loss": 0.3581,
+      "step": 6947
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7115074967064678e-06,
+      "loss": 0.0815,
+      "step": 6948
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7114656740762425e-06,
+      "loss": 0.0019,
+      "step": 6949
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7114238514460173e-06,
+      "loss": 0.4163,
+      "step": 6950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7113820288157921e-06,
+      "loss": 0.0036,
+      "step": 6951
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.711340206185567e-06,
+      "loss": 0.1096,
+      "step": 6952
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7112983835553417e-06,
+      "loss": 0.003,
+      "step": 6953
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7112565609251167e-06,
+      "loss": 0.001,
+      "step": 6954
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7112147382948915e-06,
+      "loss": 0.0016,
+      "step": 6955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7111729156646663e-06,
+      "loss": 0.2272,
+      "step": 6956
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7111310930344408e-06,
+      "loss": 0.0016,
+      "step": 6957
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7110892704042156e-06,
+      "loss": 0.0101,
+      "step": 6958
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7110474477739904e-06,
+      "loss": 0.0032,
+      "step": 6959
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7110056251437652e-06,
+      "loss": 0.0011,
+      "step": 6960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.71096380251354e-06,
+      "loss": 0.0009,
+      "step": 6961
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7109219798833148e-06,
+      "loss": 0.0022,
+      "step": 6962
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7108801572530895e-06,
+      "loss": 0.6559,
+      "step": 6963
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7108383346228643e-06,
+      "loss": 0.001,
+      "step": 6964
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7107965119926391e-06,
+      "loss": 0.0009,
+      "step": 6965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.710754689362414e-06,
+      "loss": 0.3192,
+      "step": 6966
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7107128667321887e-06,
+      "loss": 0.0026,
+      "step": 6967
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7106710441019635e-06,
+      "loss": 0.0059,
+      "step": 6968
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7106292214717385e-06,
+      "loss": 0.3196,
+      "step": 6969
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.710587398841513e-06,
+      "loss": 0.0024,
+      "step": 6970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7105455762112878e-06,
+      "loss": 0.0007,
+      "step": 6971
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7105037535810626e-06,
+      "loss": 0.0008,
+      "step": 6972
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7104619309508374e-06,
+      "loss": 0.0015,
+      "step": 6973
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7104201083206122e-06,
+      "loss": 0.0006,
+      "step": 6974
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.710378285690387e-06,
+      "loss": 0.0112,
+      "step": 6975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7103364630601618e-06,
+      "loss": 0.767,
+      "step": 6976
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7102946404299366e-06,
+      "loss": 0.0011,
+      "step": 6977
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7102528177997113e-06,
+      "loss": 0.0597,
+      "step": 6978
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7102109951694861e-06,
+      "loss": 0.4945,
+      "step": 6979
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.710169172539261e-06,
+      "loss": 0.0006,
+      "step": 6980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7101273499090357e-06,
+      "loss": 0.0047,
+      "step": 6981
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7100855272788105e-06,
+      "loss": 0.0068,
+      "step": 6982
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.710043704648585e-06,
+      "loss": 0.0018,
+      "step": 6983
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.71000188201836e-06,
+      "loss": 0.0018,
+      "step": 6984
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7099600593881348e-06,
+      "loss": 0.0093,
+      "step": 6985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7099182367579096e-06,
+      "loss": 0.0053,
+      "step": 6986
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7098764141276844e-06,
+      "loss": 0.0023,
+      "step": 6987
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7098345914974592e-06,
+      "loss": 0.0008,
+      "step": 6988
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709792768867234e-06,
+      "loss": 2.1701,
+      "step": 6989
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7097509462370088e-06,
+      "loss": 0.0041,
+      "step": 6990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7097091236067836e-06,
+      "loss": 2.108,
+      "step": 6991
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7096673009765583e-06,
+      "loss": 1.5101,
+      "step": 6992
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7096254783463331e-06,
+      "loss": 0.0008,
+      "step": 6993
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709583655716108e-06,
+      "loss": 1.0645,
+      "step": 6994
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7095418330858827e-06,
+      "loss": 0.0034,
+      "step": 6995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7095000104556575e-06,
+      "loss": 0.0601,
+      "step": 6996
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709458187825432e-06,
+      "loss": 0.0023,
+      "step": 6997
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709416365195207e-06,
+      "loss": 0.6064,
+      "step": 6998
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7093745425649819e-06,
+      "loss": 0.1746,
+      "step": 6999
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7093327199347566e-06,
+      "loss": 0.0122,
+      "step": 7000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7092908973045314e-06,
+      "loss": 0.0061,
+      "step": 7001
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7092490746743062e-06,
+      "loss": 1.6308,
+      "step": 7002
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709207252044081e-06,
+      "loss": 0.0027,
+      "step": 7003
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7091654294138558e-06,
+      "loss": 0.4518,
+      "step": 7004
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7091236067836306e-06,
+      "loss": 0.001,
+      "step": 7005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7090817841534054e-06,
+      "loss": 0.0006,
+      "step": 7006
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7090399615231801e-06,
+      "loss": 0.0023,
+      "step": 7007
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708998138892955e-06,
+      "loss": 0.0011,
+      "step": 7008
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7089563162627297e-06,
+      "loss": 1.8963,
+      "step": 7009
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7089144936325043e-06,
+      "loss": 0.0186,
+      "step": 7010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708872671002279e-06,
+      "loss": 0.0008,
+      "step": 7011
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7088308483720539e-06,
+      "loss": 0.0006,
+      "step": 7012
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7087890257418289e-06,
+      "loss": 0.0015,
+      "step": 7013
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7087472031116036e-06,
+      "loss": 0.4116,
+      "step": 7014
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7087053804813784e-06,
+      "loss": 0.0011,
+      "step": 7015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7086635578511532e-06,
+      "loss": 0.0006,
+      "step": 7016
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708621735220928e-06,
+      "loss": 0.0011,
+      "step": 7017
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7085799125907028e-06,
+      "loss": 0.6849,
+      "step": 7018
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7085380899604776e-06,
+      "loss": 0.0005,
+      "step": 7019
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7084962673302524e-06,
+      "loss": 0.0029,
+      "step": 7020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7084544447000272e-06,
+      "loss": 0.001,
+      "step": 7021
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708412622069802e-06,
+      "loss": 0.0916,
+      "step": 7022
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7083707994395767e-06,
+      "loss": 0.0005,
+      "step": 7023
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7083289768093513e-06,
+      "loss": 0.0006,
+      "step": 7024
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708287154179126e-06,
+      "loss": 2.8724,
+      "step": 7025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7082453315489009e-06,
+      "loss": 0.0008,
+      "step": 7026
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7082035089186759e-06,
+      "loss": 0.0039,
+      "step": 7027
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7081616862884507e-06,
+      "loss": 0.0013,
+      "step": 7028
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7081198636582254e-06,
+      "loss": 0.0008,
+      "step": 7029
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7080780410280002e-06,
+      "loss": 0.0008,
+      "step": 7030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708036218397775e-06,
+      "loss": 0.0016,
+      "step": 7031
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7079943957675498e-06,
+      "loss": 0.0874,
+      "step": 7032
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7079525731373246e-06,
+      "loss": 0.0007,
+      "step": 7033
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7079107505070994e-06,
+      "loss": 0.0081,
+      "step": 7034
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7078689278768742e-06,
+      "loss": 0.001,
+      "step": 7035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707827105246649e-06,
+      "loss": 0.0187,
+      "step": 7036
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7077852826164237e-06,
+      "loss": 0.0005,
+      "step": 7037
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7077434599861983e-06,
+      "loss": 0.0046,
+      "step": 7038
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707701637355973e-06,
+      "loss": 1.5897,
+      "step": 7039
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7076598147257479e-06,
+      "loss": 0.0009,
+      "step": 7040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7076179920955227e-06,
+      "loss": 0.0009,
+      "step": 7041
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7075761694652977e-06,
+      "loss": 0.6229,
+      "step": 7042
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7075343468350725e-06,
+      "loss": 0.001,
+      "step": 7043
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7074925242048472e-06,
+      "loss": 0.0993,
+      "step": 7044
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707450701574622e-06,
+      "loss": 0.5204,
+      "step": 7045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7074088789443968e-06,
+      "loss": 0.0005,
+      "step": 7046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7073670563141716e-06,
+      "loss": 0.0202,
+      "step": 7047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7073252336839464e-06,
+      "loss": 0.0094,
+      "step": 7048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7072834110537212e-06,
+      "loss": 0.1633,
+      "step": 7049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707241588423496e-06,
+      "loss": 0.001,
+      "step": 7050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7071997657932705e-06,
+      "loss": 3.5803,
+      "step": 7051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7071579431630453e-06,
+      "loss": 0.0006,
+      "step": 7052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.70711612053282e-06,
+      "loss": 0.0005,
+      "step": 7053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7070742979025949e-06,
+      "loss": 0.0008,
+      "step": 7054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7070324752723697e-06,
+      "loss": 0.0013,
+      "step": 7055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7069906526421445e-06,
+      "loss": 0.0007,
+      "step": 7056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7069488300119195e-06,
+      "loss": 0.1862,
+      "step": 7057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7069070073816942e-06,
+      "loss": 0.0006,
+      "step": 7058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706865184751469e-06,
+      "loss": 0.0009,
+      "step": 7059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7068233621212438e-06,
+      "loss": 0.0009,
+      "step": 7060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7067815394910186e-06,
+      "loss": 0.0014,
+      "step": 7061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7067397168607934e-06,
+      "loss": 0.0559,
+      "step": 7062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7066978942305682e-06,
+      "loss": 0.4065,
+      "step": 7063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706656071600343e-06,
+      "loss": 0.4701,
+      "step": 7064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7066142489701175e-06,
+      "loss": 0.0095,
+      "step": 7065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7065724263398923e-06,
+      "loss": 0.0734,
+      "step": 7066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7065306037096671e-06,
+      "loss": 0.0011,
+      "step": 7067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706488781079442e-06,
+      "loss": 0.0134,
+      "step": 7068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7064469584492167e-06,
+      "loss": 1.4945,
+      "step": 7069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7064051358189915e-06,
+      "loss": 0.0051,
+      "step": 7070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7063633131887665e-06,
+      "loss": 0.0007,
+      "step": 7071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7063214905585413e-06,
+      "loss": 0.0007,
+      "step": 7072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706279667928316e-06,
+      "loss": 2.9914,
+      "step": 7073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7062378452980908e-06,
+      "loss": 0.0007,
+      "step": 7074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7061960226678656e-06,
+      "loss": 0.0027,
+      "step": 7075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7061542000376404e-06,
+      "loss": 0.0009,
+      "step": 7076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7061123774074152e-06,
+      "loss": 0.0113,
+      "step": 7077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7060705547771898e-06,
+      "loss": 1.4427,
+      "step": 7078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7060287321469645e-06,
+      "loss": 1.33,
+      "step": 7079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7059869095167393e-06,
+      "loss": 1.1199,
+      "step": 7080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7059450868865141e-06,
+      "loss": 0.0405,
+      "step": 7081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705903264256289e-06,
+      "loss": 0.7685,
+      "step": 7082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7058614416260637e-06,
+      "loss": 0.0052,
+      "step": 7083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7058196189958385e-06,
+      "loss": 0.0008,
+      "step": 7084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7057777963656133e-06,
+      "loss": 1.926,
+      "step": 7085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7057359737353883e-06,
+      "loss": 0.0007,
+      "step": 7086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705694151105163e-06,
+      "loss": 0.0013,
+      "step": 7087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7056523284749378e-06,
+      "loss": 0.0012,
+      "step": 7088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7056105058447126e-06,
+      "loss": 0.0007,
+      "step": 7089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7055686832144874e-06,
+      "loss": 0.003,
+      "step": 7090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7055268605842622e-06,
+      "loss": 0.0007,
+      "step": 7091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7054850379540368e-06,
+      "loss": 0.0046,
+      "step": 7092
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7054432153238116e-06,
+      "loss": 0.0024,
+      "step": 7093
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7054013926935863e-06,
+      "loss": 0.0009,
+      "step": 7094
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7053595700633611e-06,
+      "loss": 0.0023,
+      "step": 7095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705317747433136e-06,
+      "loss": 0.0028,
+      "step": 7096
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7052759248029107e-06,
+      "loss": 0.0245,
+      "step": 7097
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7052341021726855e-06,
+      "loss": 0.0014,
+      "step": 7098
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7051922795424603e-06,
+      "loss": 0.0016,
+      "step": 7099
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705150456912235e-06,
+      "loss": 0.0039,
+      "step": 7100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.70510863428201e-06,
+      "loss": 0.7199,
+      "step": 7101
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7050668116517848e-06,
+      "loss": 0.0012,
+      "step": 7102
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7050249890215596e-06,
+      "loss": 0.0038,
+      "step": 7103
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7049831663913344e-06,
+      "loss": 0.0014,
+      "step": 7104
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.704941343761109e-06,
+      "loss": 0.0325,
+      "step": 7105
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7048995211308838e-06,
+      "loss": 0.0006,
+      "step": 7106
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7048576985006586e-06,
+      "loss": 0.0033,
+      "step": 7107
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7048158758704334e-06,
+      "loss": 0.0008,
+      "step": 7108
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7047740532402081e-06,
+      "loss": 0.001,
+      "step": 7109
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.704732230609983e-06,
+      "loss": 0.0023,
+      "step": 7110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7046904079797577e-06,
+      "loss": 0.0006,
+      "step": 7111
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7046485853495325e-06,
+      "loss": 0.0009,
+      "step": 7112
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7046067627193073e-06,
+      "loss": 0.0008,
+      "step": 7113
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.704564940089082e-06,
+      "loss": 0.0017,
+      "step": 7114
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.704523117458857e-06,
+      "loss": 0.0029,
+      "step": 7115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7044812948286319e-06,
+      "loss": 0.0009,
+      "step": 7116
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7044394721984066e-06,
+      "loss": 0.0017,
+      "step": 7117
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7043976495681814e-06,
+      "loss": 0.001,
+      "step": 7118
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.704355826937956e-06,
+      "loss": 0.0005,
+      "step": 7119
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7043140043077308e-06,
+      "loss": 0.001,
+      "step": 7120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7042721816775056e-06,
+      "loss": 1.3441,
+      "step": 7121
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7042303590472804e-06,
+      "loss": 0.0016,
+      "step": 7122
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7041885364170551e-06,
+      "loss": 0.0013,
+      "step": 7123
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.70414671378683e-06,
+      "loss": 0.0011,
+      "step": 7124
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7041048911566047e-06,
+      "loss": 0.0257,
+      "step": 7125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7040630685263795e-06,
+      "loss": 2.2784,
+      "step": 7126
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7040212458961543e-06,
+      "loss": 0.0172,
+      "step": 7127
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.703979423265929e-06,
+      "loss": 0.0022,
+      "step": 7128
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7039376006357039e-06,
+      "loss": 0.0015,
+      "step": 7129
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7038957780054789e-06,
+      "loss": 0.0007,
+      "step": 7130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7038539553752536e-06,
+      "loss": 0.0007,
+      "step": 7131
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7038121327450284e-06,
+      "loss": 1.3074,
+      "step": 7132
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.703770310114803e-06,
+      "loss": 0.0105,
+      "step": 7133
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7037284874845778e-06,
+      "loss": 0.0007,
+      "step": 7134
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7036866648543526e-06,
+      "loss": 0.0004,
+      "step": 7135
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7036448422241274e-06,
+      "loss": 0.0008,
+      "step": 7136
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7036030195939022e-06,
+      "loss": 0.0914,
+      "step": 7137
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.703561196963677e-06,
+      "loss": 0.001,
+      "step": 7138
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7035193743334517e-06,
+      "loss": 0.0028,
+      "step": 7139
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7034775517032265e-06,
+      "loss": 0.0012,
+      "step": 7140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7034357290730013e-06,
+      "loss": 0.0006,
+      "step": 7141
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.703393906442776e-06,
+      "loss": 1.8631,
+      "step": 7142
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7033520838125509e-06,
+      "loss": 0.0019,
+      "step": 7143
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7033102611823257e-06,
+      "loss": 0.0668,
+      "step": 7144
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7032684385521007e-06,
+      "loss": 0.8215,
+      "step": 7145
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7032266159218752e-06,
+      "loss": 0.0008,
+      "step": 7146
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.70318479329165e-06,
+      "loss": 0.0011,
+      "step": 7147
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7031429706614248e-06,
+      "loss": 0.689,
+      "step": 7148
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7031011480311996e-06,
+      "loss": 0.0011,
+      "step": 7149
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7030593254009744e-06,
+      "loss": 0.1509,
+      "step": 7150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7030175027707492e-06,
+      "loss": 0.001,
+      "step": 7151
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702975680140524e-06,
+      "loss": 0.1365,
+      "step": 7152
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7029338575102987e-06,
+      "loss": 0.2826,
+      "step": 7153
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7028920348800735e-06,
+      "loss": 1.2351,
+      "step": 7154
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7028502122498483e-06,
+      "loss": 0.2141,
+      "step": 7155
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702808389619623e-06,
+      "loss": 0.1169,
+      "step": 7156
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7027665669893979e-06,
+      "loss": 0.0174,
+      "step": 7157
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7027247443591727e-06,
+      "loss": 0.6825,
+      "step": 7158
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7026829217289477e-06,
+      "loss": 0.0022,
+      "step": 7159
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7026410990987222e-06,
+      "loss": 0.0009,
+      "step": 7160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702599276468497e-06,
+      "loss": 0.0005,
+      "step": 7161
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7025574538382718e-06,
+      "loss": 0.0017,
+      "step": 7162
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7025156312080466e-06,
+      "loss": 0.0012,
+      "step": 7163
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7024738085778214e-06,
+      "loss": 0.8437,
+      "step": 7164
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7024319859475962e-06,
+      "loss": 0.7514,
+      "step": 7165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702390163317371e-06,
+      "loss": 0.0011,
+      "step": 7166
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7023483406871457e-06,
+      "loss": 0.0012,
+      "step": 7167
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7023065180569205e-06,
+      "loss": 0.3006,
+      "step": 7168
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7022646954266953e-06,
+      "loss": 0.0016,
+      "step": 7169
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.70222287279647e-06,
+      "loss": 0.016,
+      "step": 7170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7021810501662449e-06,
+      "loss": 0.0006,
+      "step": 7171
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7021392275360197e-06,
+      "loss": 0.0063,
+      "step": 7172
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7020974049057943e-06,
+      "loss": 0.0005,
+      "step": 7173
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7020555822755692e-06,
+      "loss": 0.0027,
+      "step": 7174
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702013759645344e-06,
+      "loss": 0.0013,
+      "step": 7175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7019719370151188e-06,
+      "loss": 0.0025,
+      "step": 7176
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7019301143848936e-06,
+      "loss": 0.0022,
+      "step": 7177
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7018882917546684e-06,
+      "loss": 0.428,
+      "step": 7178
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7018464691244432e-06,
+      "loss": 0.0044,
+      "step": 7179
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701804646494218e-06,
+      "loss": 0.0025,
+      "step": 7180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7017628238639928e-06,
+      "loss": 0.0005,
+      "step": 7181
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7017210012337675e-06,
+      "loss": 3.6266,
+      "step": 7182
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7016791786035423e-06,
+      "loss": 3.2004,
+      "step": 7183
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7016373559733171e-06,
+      "loss": 1.1156,
+      "step": 7184
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701595533343092e-06,
+      "loss": 0.0008,
+      "step": 7185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7015537107128665e-06,
+      "loss": 0.0008,
+      "step": 7186
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7015118880826413e-06,
+      "loss": 1.6965,
+      "step": 7187
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701470065452416e-06,
+      "loss": 0.0008,
+      "step": 7188
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701428242822191e-06,
+      "loss": 0.4946,
+      "step": 7189
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7013864201919658e-06,
+      "loss": 0.0093,
+      "step": 7190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7013445975617406e-06,
+      "loss": 3.6218,
+      "step": 7191
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7013027749315154e-06,
+      "loss": 3.0571,
+      "step": 7192
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7012609523012902e-06,
+      "loss": 0.0005,
+      "step": 7193
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701219129671065e-06,
+      "loss": 0.0007,
+      "step": 7194
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7011773070408398e-06,
+      "loss": 0.0109,
+      "step": 7195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7011354844106145e-06,
+      "loss": 0.0011,
+      "step": 7196
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7010936617803893e-06,
+      "loss": 0.0012,
+      "step": 7197
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7010518391501641e-06,
+      "loss": 0.0049,
+      "step": 7198
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701010016519939e-06,
+      "loss": 0.0228,
+      "step": 7199
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7009681938897135e-06,
+      "loss": 0.0012,
+      "step": 7200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7009263712594883e-06,
+      "loss": 0.0085,
+      "step": 7201
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700884548629263e-06,
+      "loss": 0.0824,
+      "step": 7202
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700842725999038e-06,
+      "loss": 0.0007,
+      "step": 7203
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7008009033688128e-06,
+      "loss": 0.0021,
+      "step": 7204
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7007590807385876e-06,
+      "loss": 0.1825,
+      "step": 7205
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7007172581083624e-06,
+      "loss": 0.0066,
+      "step": 7206
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7006754354781372e-06,
+      "loss": 0.0011,
+      "step": 7207
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700633612847912e-06,
+      "loss": 0.0384,
+      "step": 7208
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7005917902176868e-06,
+      "loss": 0.0009,
+      "step": 7209
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7005499675874616e-06,
+      "loss": 0.0019,
+      "step": 7210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7005081449572363e-06,
+      "loss": 1.2316,
+      "step": 7211
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7004663223270111e-06,
+      "loss": 0.0013,
+      "step": 7212
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7004244996967857e-06,
+      "loss": 0.0007,
+      "step": 7213
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7003826770665605e-06,
+      "loss": 0.0055,
+      "step": 7214
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7003408544363353e-06,
+      "loss": 0.0017,
+      "step": 7215
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.70029903180611e-06,
+      "loss": 0.3053,
+      "step": 7216
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7002572091758848e-06,
+      "loss": 0.0009,
+      "step": 7217
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7002153865456598e-06,
+      "loss": 0.0009,
+      "step": 7218
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7001735639154346e-06,
+      "loss": 0.0007,
+      "step": 7219
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7001317412852094e-06,
+      "loss": 0.0019,
+      "step": 7220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7000899186549842e-06,
+      "loss": 0.0017,
+      "step": 7221
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700048096024759e-06,
+      "loss": 0.007,
+      "step": 7222
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7000062733945338e-06,
+      "loss": 0.0139,
+      "step": 7223
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6999644507643086e-06,
+      "loss": 0.0006,
+      "step": 7224
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6999226281340834e-06,
+      "loss": 0.7036,
+      "step": 7225
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6998808055038581e-06,
+      "loss": 0.0012,
+      "step": 7226
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6998389828736327e-06,
+      "loss": 0.0016,
+      "step": 7227
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6997971602434075e-06,
+      "loss": 0.0033,
+      "step": 7228
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6997553376131823e-06,
+      "loss": 0.0008,
+      "step": 7229
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699713514982957e-06,
+      "loss": 0.0024,
+      "step": 7230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6996716923527319e-06,
+      "loss": 0.3176,
+      "step": 7231
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6996298697225069e-06,
+      "loss": 0.0021,
+      "step": 7232
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6995880470922816e-06,
+      "loss": 0.0009,
+      "step": 7233
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6995462244620564e-06,
+      "loss": 0.7606,
+      "step": 7234
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6995044018318312e-06,
+      "loss": 0.0011,
+      "step": 7235
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699462579201606e-06,
+      "loss": 0.0025,
+      "step": 7236
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6994207565713808e-06,
+      "loss": 0.0014,
+      "step": 7237
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6993789339411556e-06,
+      "loss": 0.005,
+      "step": 7238
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6993371113109304e-06,
+      "loss": 0.0012,
+      "step": 7239
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6992952886807051e-06,
+      "loss": 0.0874,
+      "step": 7240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6992534660504797e-06,
+      "loss": 0.0016,
+      "step": 7241
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6992116434202545e-06,
+      "loss": 0.0041,
+      "step": 7242
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6991698207900293e-06,
+      "loss": 0.002,
+      "step": 7243
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699127998159804e-06,
+      "loss": 0.002,
+      "step": 7244
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6990861755295789e-06,
+      "loss": 0.0016,
+      "step": 7245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6990443528993537e-06,
+      "loss": 0.0004,
+      "step": 7246
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6990025302691287e-06,
+      "loss": 0.0008,
+      "step": 7247
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6989607076389034e-06,
+      "loss": 0.0027,
+      "step": 7248
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6989188850086782e-06,
+      "loss": 0.002,
+      "step": 7249
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698877062378453e-06,
+      "loss": 0.0011,
+      "step": 7250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6988352397482278e-06,
+      "loss": 0.023,
+      "step": 7251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6987934171180026e-06,
+      "loss": 0.0008,
+      "step": 7252
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6987515944877774e-06,
+      "loss": 0.0008,
+      "step": 7253
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698709771857552e-06,
+      "loss": 0.0014,
+      "step": 7254
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6986679492273267e-06,
+      "loss": 0.0371,
+      "step": 7255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6986261265971015e-06,
+      "loss": 0.0009,
+      "step": 7256
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6985843039668763e-06,
+      "loss": 1.9819,
+      "step": 7257
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698542481336651e-06,
+      "loss": 0.0036,
+      "step": 7258
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6985006587064259e-06,
+      "loss": 0.0049,
+      "step": 7259
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6984588360762007e-06,
+      "loss": 0.1604,
+      "step": 7260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6984170134459754e-06,
+      "loss": 0.0011,
+      "step": 7261
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6983751908157504e-06,
+      "loss": 0.005,
+      "step": 7262
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6983333681855252e-06,
+      "loss": 0.435,
+      "step": 7263
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6982915455553e-06,
+      "loss": 0.0008,
+      "step": 7264
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6982497229250748e-06,
+      "loss": 0.0015,
+      "step": 7265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6982079002948496e-06,
+      "loss": 0.0006,
+      "step": 7266
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6981660776646244e-06,
+      "loss": 0.0017,
+      "step": 7267
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.698124255034399e-06,
+      "loss": 1.506,
+      "step": 7268
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6980824324041737e-06,
+      "loss": 2.3891,
+      "step": 7269
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6980406097739485e-06,
+      "loss": 0.0059,
+      "step": 7270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6979987871437233e-06,
+      "loss": 0.0006,
+      "step": 7271
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.697956964513498e-06,
+      "loss": 0.0008,
+      "step": 7272
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6979151418832729e-06,
+      "loss": 0.0071,
+      "step": 7273
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6978733192530477e-06,
+      "loss": 0.0009,
+      "step": 7274
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6978314966228225e-06,
+      "loss": 0.0008,
+      "step": 7275
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6977896739925975e-06,
+      "loss": 0.0007,
+      "step": 7276
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6977478513623722e-06,
+      "loss": 0.0009,
+      "step": 7277
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.697706028732147e-06,
+      "loss": 0.8499,
+      "step": 7278
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6976642061019218e-06,
+      "loss": 0.0009,
+      "step": 7279
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6976223834716966e-06,
+      "loss": 0.0011,
+      "step": 7280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6975805608414712e-06,
+      "loss": 0.0007,
+      "step": 7281
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.697538738211246e-06,
+      "loss": 0.0006,
+      "step": 7282
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6974969155810207e-06,
+      "loss": 0.0188,
+      "step": 7283
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6974550929507955e-06,
+      "loss": 0.0013,
+      "step": 7284
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6974132703205703e-06,
+      "loss": 1.5181,
+      "step": 7285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6973714476903451e-06,
+      "loss": 0.3314,
+      "step": 7286
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6973296250601199e-06,
+      "loss": 0.0005,
+      "step": 7287
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6972878024298947e-06,
+      "loss": 0.0036,
+      "step": 7288
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6972459797996695e-06,
+      "loss": 0.001,
+      "step": 7289
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6972041571694443e-06,
+      "loss": 0.001,
+      "step": 7290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6971623345392192e-06,
+      "loss": 0.0017,
+      "step": 7291
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.697120511908994e-06,
+      "loss": 0.0016,
+      "step": 7292
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6970786892787688e-06,
+      "loss": 0.0309,
+      "step": 7293
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6970368666485436e-06,
+      "loss": 1.5888,
+      "step": 7294
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6969950440183182e-06,
+      "loss": 2.1587,
+      "step": 7295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696953221388093e-06,
+      "loss": 0.0028,
+      "step": 7296
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6969113987578678e-06,
+      "loss": 0.159,
+      "step": 7297
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6968695761276425e-06,
+      "loss": 0.0021,
+      "step": 7298
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6968277534974173e-06,
+      "loss": 0.0007,
+      "step": 7299
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6967859308671921e-06,
+      "loss": 0.0018,
+      "step": 7300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696744108236967e-06,
+      "loss": 0.0006,
+      "step": 7301
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6967022856067417e-06,
+      "loss": 0.511,
+      "step": 7302
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6966604629765165e-06,
+      "loss": 0.6177,
+      "step": 7303
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6966186403462913e-06,
+      "loss": 0.0023,
+      "step": 7304
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696576817716066e-06,
+      "loss": 0.0014,
+      "step": 7305
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696534995085841e-06,
+      "loss": 0.0016,
+      "step": 7306
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6964931724556158e-06,
+      "loss": 0.0055,
+      "step": 7307
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6964513498253904e-06,
+      "loss": 1.0544,
+      "step": 7308
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6964095271951652e-06,
+      "loss": 0.2358,
+      "step": 7309
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.69636770456494e-06,
+      "loss": 0.1415,
+      "step": 7310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6963258819347148e-06,
+      "loss": 3.2084,
+      "step": 7311
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6962840593044896e-06,
+      "loss": 0.2574,
+      "step": 7312
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6962422366742643e-06,
+      "loss": 0.0012,
+      "step": 7313
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6962004140440391e-06,
+      "loss": 0.0013,
+      "step": 7314
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696158591413814e-06,
+      "loss": 0.0006,
+      "step": 7315
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6961167687835887e-06,
+      "loss": 0.0035,
+      "step": 7316
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6960749461533635e-06,
+      "loss": 0.4951,
+      "step": 7317
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6960331235231383e-06,
+      "loss": 0.7203,
+      "step": 7318
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695991300892913e-06,
+      "loss": 2.2907,
+      "step": 7319
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695949478262688e-06,
+      "loss": 0.0017,
+      "step": 7320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6959076556324628e-06,
+      "loss": 1.2508,
+      "step": 7321
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6958658330022374e-06,
+      "loss": 0.0008,
+      "step": 7322
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6958240103720122e-06,
+      "loss": 0.0023,
+      "step": 7323
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695782187741787e-06,
+      "loss": 0.0006,
+      "step": 7324
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6957403651115618e-06,
+      "loss": 4.2885,
+      "step": 7325
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6956985424813366e-06,
+      "loss": 0.0034,
+      "step": 7326
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6956567198511113e-06,
+      "loss": 0.0016,
+      "step": 7327
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6956148972208861e-06,
+      "loss": 0.5249,
+      "step": 7328
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695573074590661e-06,
+      "loss": 0.0011,
+      "step": 7329
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6955312519604357e-06,
+      "loss": 0.0011,
+      "step": 7330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6954894293302105e-06,
+      "loss": 0.0424,
+      "step": 7331
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6954476066999853e-06,
+      "loss": 0.876,
+      "step": 7332
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.69540578406976e-06,
+      "loss": 0.0005,
+      "step": 7333
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6953639614395349e-06,
+      "loss": 0.5622,
+      "step": 7334
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6953221388093098e-06,
+      "loss": 0.0009,
+      "step": 7335
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6952803161790844e-06,
+      "loss": 0.0234,
+      "step": 7336
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6952384935488592e-06,
+      "loss": 0.0024,
+      "step": 7337
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695196670918634e-06,
+      "loss": 0.03,
+      "step": 7338
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6951548482884088e-06,
+      "loss": 0.0056,
+      "step": 7339
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6951130256581836e-06,
+      "loss": 0.0079,
+      "step": 7340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6950712030279584e-06,
+      "loss": 0.9389,
+      "step": 7341
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6950293803977331e-06,
+      "loss": 0.001,
+      "step": 7342
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694987557767508e-06,
+      "loss": 0.0016,
+      "step": 7343
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6949457351372827e-06,
+      "loss": 0.725,
+      "step": 7344
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6949039125070575e-06,
+      "loss": 0.0014,
+      "step": 7345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6948620898768323e-06,
+      "loss": 2.1182,
+      "step": 7346
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694820267246607e-06,
+      "loss": 0.0022,
+      "step": 7347
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6947784446163819e-06,
+      "loss": 0.0023,
+      "step": 7348
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6947366219861564e-06,
+      "loss": 0.0016,
+      "step": 7349
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6946947993559314e-06,
+      "loss": 0.0007,
+      "step": 7350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6946529767257062e-06,
+      "loss": 0.0012,
+      "step": 7351
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694611154095481e-06,
+      "loss": 0.0015,
+      "step": 7352
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6945693314652558e-06,
+      "loss": 0.0031,
+      "step": 7353
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6945275088350306e-06,
+      "loss": 0.0007,
+      "step": 7354
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6944856862048054e-06,
+      "loss": 0.0009,
+      "step": 7355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6944438635745801e-06,
+      "loss": 0.0018,
+      "step": 7356
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694402040944355e-06,
+      "loss": 0.3658,
+      "step": 7357
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6943602183141297e-06,
+      "loss": 0.5747,
+      "step": 7358
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6943183956839045e-06,
+      "loss": 0.451,
+      "step": 7359
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6942765730536793e-06,
+      "loss": 1.9252,
+      "step": 7360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694234750423454e-06,
+      "loss": 0.607,
+      "step": 7361
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6941929277932287e-06,
+      "loss": 0.0026,
+      "step": 7362
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6941511051630034e-06,
+      "loss": 0.0064,
+      "step": 7363
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6941092825327784e-06,
+      "loss": 0.0066,
+      "step": 7364
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6940674599025532e-06,
+      "loss": 0.0021,
+      "step": 7365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694025637272328e-06,
+      "loss": 0.0015,
+      "step": 7366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6939838146421028e-06,
+      "loss": 0.5692,
+      "step": 7367
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6939419920118776e-06,
+      "loss": 0.0007,
+      "step": 7368
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6939001693816524e-06,
+      "loss": 0.0014,
+      "step": 7369
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6938583467514272e-06,
+      "loss": 0.0019,
+      "step": 7370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693816524121202e-06,
+      "loss": 1.9395,
+      "step": 7371
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6937747014909767e-06,
+      "loss": 0.001,
+      "step": 7372
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6937328788607515e-06,
+      "loss": 0.0019,
+      "step": 7373
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6936910562305263e-06,
+      "loss": 0.0009,
+      "step": 7374
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693649233600301e-06,
+      "loss": 0.0125,
+      "step": 7375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6936074109700757e-06,
+      "loss": 0.002,
+      "step": 7376
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6935655883398505e-06,
+      "loss": 0.0066,
+      "step": 7377
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6935237657096252e-06,
+      "loss": 1.6401,
+      "step": 7378
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6934819430794002e-06,
+      "loss": 0.0017,
+      "step": 7379
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693440120449175e-06,
+      "loss": 0.0015,
+      "step": 7380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6933982978189498e-06,
+      "loss": 0.0009,
+      "step": 7381
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6933564751887246e-06,
+      "loss": 3.5144,
+      "step": 7382
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6933146525584994e-06,
+      "loss": 0.0028,
+      "step": 7383
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6932728299282742e-06,
+      "loss": 0.0009,
+      "step": 7384
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693231007298049e-06,
+      "loss": 0.0021,
+      "step": 7385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6931891846678237e-06,
+      "loss": 0.0015,
+      "step": 7386
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6931473620375985e-06,
+      "loss": 0.0223,
+      "step": 7387
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6931055394073733e-06,
+      "loss": 0.0081,
+      "step": 7388
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6930637167771479e-06,
+      "loss": 0.0011,
+      "step": 7389
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6930218941469227e-06,
+      "loss": 0.0028,
+      "step": 7390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6929800715166975e-06,
+      "loss": 0.007,
+      "step": 7391
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6929382488864722e-06,
+      "loss": 1.9529,
+      "step": 7392
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6928964262562472e-06,
+      "loss": 0.0011,
+      "step": 7393
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692854603626022e-06,
+      "loss": 0.7638,
+      "step": 7394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6928127809957968e-06,
+      "loss": 0.0067,
+      "step": 7395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6927709583655716e-06,
+      "loss": 0.1743,
+      "step": 7396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6927291357353464e-06,
+      "loss": 0.4148,
+      "step": 7397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6926873131051212e-06,
+      "loss": 1.7709,
+      "step": 7398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692645490474896e-06,
+      "loss": 0.0129,
+      "step": 7399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6926036678446707e-06,
+      "loss": 1.4953,
+      "step": 7400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6925618452144455e-06,
+      "loss": 0.0015,
+      "step": 7401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6925200225842203e-06,
+      "loss": 0.001,
+      "step": 7402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692478199953995e-06,
+      "loss": 0.0006,
+      "step": 7403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6924363773237697e-06,
+      "loss": 1.2593,
+      "step": 7404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6923945546935445e-06,
+      "loss": 0.0012,
+      "step": 7405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6923527320633193e-06,
+      "loss": 0.003,
+      "step": 7406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692310909433094e-06,
+      "loss": 0.0039,
+      "step": 7407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692269086802869e-06,
+      "loss": 0.0025,
+      "step": 7408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6922272641726438e-06,
+      "loss": 0.001,
+      "step": 7409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6921854415424186e-06,
+      "loss": 0.0027,
+      "step": 7410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6921436189121934e-06,
+      "loss": 1.9738,
+      "step": 7411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6921017962819682e-06,
+      "loss": 1.1733,
+      "step": 7412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692059973651743e-06,
+      "loss": 0.0013,
+      "step": 7413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6920181510215178e-06,
+      "loss": 0.0018,
+      "step": 7414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6919763283912925e-06,
+      "loss": 0.0023,
+      "step": 7415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6919345057610671e-06,
+      "loss": 3.0559,
+      "step": 7416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691892683130842e-06,
+      "loss": 2.0655,
+      "step": 7417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6918508605006167e-06,
+      "loss": 0.0009,
+      "step": 7418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6918090378703915e-06,
+      "loss": 2.3882,
+      "step": 7419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6917672152401663e-06,
+      "loss": 0.0027,
+      "step": 7420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.691725392609941e-06,
+      "loss": 1.6035,
+      "step": 7421
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6916835699797158e-06,
+      "loss": 0.0014,
+      "step": 7422
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6916417473494908e-06,
+      "loss": 0.0014,
+      "step": 7423
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6915999247192656e-06,
+      "loss": 0.2363,
+      "step": 7424
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6915581020890404e-06,
+      "loss": 0.6591,
+      "step": 7425
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6915162794588152e-06,
+      "loss": 0.0039,
+      "step": 7426
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.69147445682859e-06,
+      "loss": 0.0105,
+      "step": 7427
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6914326341983648e-06,
+      "loss": 0.4092,
+      "step": 7428
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6913908115681396e-06,
+      "loss": 2.365,
+      "step": 7429
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6913489889379141e-06,
+      "loss": 0.0074,
+      "step": 7430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.691307166307689e-06,
+      "loss": 0.1328,
+      "step": 7431
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6912653436774637e-06,
+      "loss": 0.0009,
+      "step": 7432
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6912235210472385e-06,
+      "loss": 0.0022,
+      "step": 7433
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6911816984170133e-06,
+      "loss": 0.0013,
+      "step": 7434
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.691139875786788e-06,
+      "loss": 0.0015,
+      "step": 7435
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6910980531565628e-06,
+      "loss": 0.0036,
+      "step": 7436
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6910562305263378e-06,
+      "loss": 0.0022,
+      "step": 7437
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6910144078961126e-06,
+      "loss": 0.0022,
+      "step": 7438
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6909725852658874e-06,
+      "loss": 1.2786,
+      "step": 7439
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6909307626356622e-06,
+      "loss": 0.0062,
+      "step": 7440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.690888940005437e-06,
+      "loss": 0.2446,
+      "step": 7441
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6908471173752118e-06,
+      "loss": 0.0017,
+      "step": 7442
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6908052947449866e-06,
+      "loss": 1.372,
+      "step": 7443
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6907634721147611e-06,
+      "loss": 0.0019,
+      "step": 7444
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.690721649484536e-06,
+      "loss": 0.0006,
+      "step": 7445
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6906798268543107e-06,
+      "loss": 0.0042,
+      "step": 7446
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6906380042240855e-06,
+      "loss": 0.1559,
+      "step": 7447
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6905961815938603e-06,
+      "loss": 0.0116,
+      "step": 7448
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.690554358963635e-06,
+      "loss": 0.001,
+      "step": 7449
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6905125363334099e-06,
+      "loss": 0.0048,
+      "step": 7450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6904707137031846e-06,
+      "loss": 0.0036,
+      "step": 7451
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6904288910729596e-06,
+      "loss": 0.0145,
+      "step": 7452
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6903870684427344e-06,
+      "loss": 0.0015,
+      "step": 7453
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6903452458125092e-06,
+      "loss": 2.7101,
+      "step": 7454
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.690303423182284e-06,
+      "loss": 0.0031,
+      "step": 7455
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6902616005520588e-06,
+      "loss": 2.4668,
+      "step": 7456
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6902197779218334e-06,
+      "loss": 0.0014,
+      "step": 7457
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6901779552916081e-06,
+      "loss": 0.0009,
+      "step": 7458
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.690136132661383e-06,
+      "loss": 1.2251,
+      "step": 7459
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6900943100311577e-06,
+      "loss": 0.0022,
+      "step": 7460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6900524874009325e-06,
+      "loss": 0.003,
+      "step": 7461
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6900106647707073e-06,
+      "loss": 0.0124,
+      "step": 7462
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689968842140482e-06,
+      "loss": 0.0006,
+      "step": 7463
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6899270195102569e-06,
+      "loss": 0.1875,
+      "step": 7464
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6898851968800316e-06,
+      "loss": 0.0134,
+      "step": 7465
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6898433742498064e-06,
+      "loss": 0.016,
+      "step": 7466
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6898015516195814e-06,
+      "loss": 0.0017,
+      "step": 7467
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6897597289893562e-06,
+      "loss": 0.0033,
+      "step": 7468
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689717906359131e-06,
+      "loss": 0.004,
+      "step": 7469
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6896760837289058e-06,
+      "loss": 0.0007,
+      "step": 7470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6896342610986804e-06,
+      "loss": 0.0085,
+      "step": 7471
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6895924384684552e-06,
+      "loss": 0.1115,
+      "step": 7472
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.68955061583823e-06,
+      "loss": 0.0026,
+      "step": 7473
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6895087932080047e-06,
+      "loss": 3.2637,
+      "step": 7474
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6894669705777795e-06,
+      "loss": 0.0016,
+      "step": 7475
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6894251479475543e-06,
+      "loss": 0.0023,
+      "step": 7476
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689383325317329e-06,
+      "loss": 0.2823,
+      "step": 7477
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6893415026871039e-06,
+      "loss": 0.0012,
+      "step": 7478
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6892996800568787e-06,
+      "loss": 0.0044,
+      "step": 7479
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6892578574266534e-06,
+      "loss": 0.0316,
+      "step": 7480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6892160347964284e-06,
+      "loss": 1.7294,
+      "step": 7481
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6891742121662032e-06,
+      "loss": 3.1382,
+      "step": 7482
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689132389535978e-06,
+      "loss": 0.0023,
+      "step": 7483
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6890905669057526e-06,
+      "loss": 0.001,
+      "step": 7484
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6890487442755274e-06,
+      "loss": 0.0006,
+      "step": 7485
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6890069216453022e-06,
+      "loss": 0.0015,
+      "step": 7486
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688965099015077e-06,
+      "loss": 1.1901,
+      "step": 7487
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6889232763848517e-06,
+      "loss": 0.0029,
+      "step": 7488
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6888814537546265e-06,
+      "loss": 0.0062,
+      "step": 7489
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6888396311244013e-06,
+      "loss": 0.0008,
+      "step": 7490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688797808494176e-06,
+      "loss": 0.0035,
+      "step": 7491
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6887559858639509e-06,
+      "loss": 0.0011,
+      "step": 7492
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6887141632337257e-06,
+      "loss": 0.0669,
+      "step": 7493
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6886723406035005e-06,
+      "loss": 0.0012,
+      "step": 7494
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6886305179732752e-06,
+      "loss": 2.0268,
+      "step": 7495
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6885886953430502e-06,
+      "loss": 0.0018,
+      "step": 7496
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688546872712825e-06,
+      "loss": 0.0298,
+      "step": 7497
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6885050500825996e-06,
+      "loss": 2.0717,
+      "step": 7498
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6884632274523744e-06,
+      "loss": 0.726,
+      "step": 7499
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6884214048221492e-06,
+      "loss": 0.0016,
+      "step": 7500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688379582191924e-06,
+      "loss": 0.0016,
+      "step": 7501
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6883377595616987e-06,
+      "loss": 0.0019,
+      "step": 7502
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6882959369314735e-06,
+      "loss": 0.0008,
+      "step": 7503
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6882541143012483e-06,
+      "loss": 0.0006,
+      "step": 7504
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688212291671023e-06,
+      "loss": 0.058,
+      "step": 7505
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6881704690407979e-06,
+      "loss": 3.0443,
+      "step": 7506
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6881286464105727e-06,
+      "loss": 0.0009,
+      "step": 7507
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6880868237803475e-06,
+      "loss": 0.0013,
+      "step": 7508
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6880450011501222e-06,
+      "loss": 0.0023,
+      "step": 7509
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688003178519897e-06,
+      "loss": 0.0073,
+      "step": 7510
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6879613558896718e-06,
+      "loss": 0.0007,
+      "step": 7511
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6879195332594466e-06,
+      "loss": 0.5617,
+      "step": 7512
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6878777106292214e-06,
+      "loss": 0.002,
+      "step": 7513
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6878358879989962e-06,
+      "loss": 0.0009,
+      "step": 7514
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687794065368771e-06,
+      "loss": 0.0013,
+      "step": 7515
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6877522427385458e-06,
+      "loss": 0.0013,
+      "step": 7516
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6877104201083205e-06,
+      "loss": 0.0009,
+      "step": 7517
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6876685974780953e-06,
+      "loss": 0.0024,
+      "step": 7518
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6876267748478701e-06,
+      "loss": 0.0005,
+      "step": 7519
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687584952217645e-06,
+      "loss": 0.0054,
+      "step": 7520
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6875431295874197e-06,
+      "loss": 0.0022,
+      "step": 7521
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6875013069571945e-06,
+      "loss": 0.0011,
+      "step": 7522
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6874594843269693e-06,
+      "loss": 0.0023,
+      "step": 7523
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6874176616967438e-06,
+      "loss": 0.151,
+      "step": 7524
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6873758390665188e-06,
+      "loss": 1.5837,
+      "step": 7525
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6873340164362936e-06,
+      "loss": 0.0012,
+      "step": 7526
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6872921938060684e-06,
+      "loss": 0.0022,
+      "step": 7527
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6872503711758432e-06,
+      "loss": 0.0008,
+      "step": 7528
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687208548545618e-06,
+      "loss": 0.005,
+      "step": 7529
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6871667259153928e-06,
+      "loss": 0.0013,
+      "step": 7530
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6871249032851675e-06,
+      "loss": 0.0028,
+      "step": 7531
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6870830806549423e-06,
+      "loss": 0.0016,
+      "step": 7532
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6870412580247171e-06,
+      "loss": 0.0006,
+      "step": 7533
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686999435394492e-06,
+      "loss": 0.0027,
+      "step": 7534
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6869576127642667e-06,
+      "loss": 2.1138,
+      "step": 7535
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6869157901340415e-06,
+      "loss": 0.5599,
+      "step": 7536
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6868739675038163e-06,
+      "loss": 0.6615,
+      "step": 7537
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6868321448735908e-06,
+      "loss": 0.0203,
+      "step": 7538
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6867903222433656e-06,
+      "loss": 1.1988,
+      "step": 7539
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6867484996131406e-06,
+      "loss": 0.0061,
+      "step": 7540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6867066769829154e-06,
+      "loss": 2.1018,
+      "step": 7541
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6866648543526902e-06,
+      "loss": 0.0051,
+      "step": 7542
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686623031722465e-06,
+      "loss": 0.0014,
+      "step": 7543
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6865812090922398e-06,
+      "loss": 0.001,
+      "step": 7544
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6865393864620146e-06,
+      "loss": 0.0051,
+      "step": 7545
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6864975638317893e-06,
+      "loss": 0.0021,
+      "step": 7546
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6864557412015641e-06,
+      "loss": 0.0018,
+      "step": 7547
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686413918571339e-06,
+      "loss": 0.0024,
+      "step": 7548
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6863720959411137e-06,
+      "loss": 0.0005,
+      "step": 7549
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6863302733108885e-06,
+      "loss": 0.0288,
+      "step": 7550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6862884506806633e-06,
+      "loss": 2.0297,
+      "step": 7551
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6862466280504378e-06,
+      "loss": 0.0015,
+      "step": 7552
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6862048054202126e-06,
+      "loss": 1.5585,
+      "step": 7553
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6861629827899874e-06,
+      "loss": 0.5055,
+      "step": 7554
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6861211601597624e-06,
+      "loss": 1.7858,
+      "step": 7555
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6860793375295372e-06,
+      "loss": 0.0008,
+      "step": 7556
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686037514899312e-06,
+      "loss": 0.001,
+      "step": 7557
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6859956922690868e-06,
+      "loss": 0.0019,
+      "step": 7558
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6859538696388616e-06,
+      "loss": 0.001,
+      "step": 7559
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6859120470086363e-06,
+      "loss": 0.0022,
+      "step": 7560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6858702243784111e-06,
+      "loss": 0.001,
+      "step": 7561
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685828401748186e-06,
+      "loss": 0.0016,
+      "step": 7562
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6857865791179607e-06,
+      "loss": 0.0015,
+      "step": 7563
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6857447564877355e-06,
+      "loss": 1.5848,
+      "step": 7564
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.68570293385751e-06,
+      "loss": 0.0034,
+      "step": 7565
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6856611112272849e-06,
+      "loss": 0.001,
+      "step": 7566
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6856192885970596e-06,
+      "loss": 0.0008,
+      "step": 7567
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6855774659668344e-06,
+      "loss": 0.4592,
+      "step": 7568
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6855356433366094e-06,
+      "loss": 0.0034,
+      "step": 7569
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6854938207063842e-06,
+      "loss": 0.0439,
+      "step": 7570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685451998076159e-06,
+      "loss": 2.5599,
+      "step": 7571
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6854101754459338e-06,
+      "loss": 0.0011,
+      "step": 7572
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6853683528157086e-06,
+      "loss": 1.9436,
+      "step": 7573
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6853265301854834e-06,
+      "loss": 0.001,
+      "step": 7574
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6852847075552581e-06,
+      "loss": 0.0019,
+      "step": 7575
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685242884925033e-06,
+      "loss": 0.004,
+      "step": 7576
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6852010622948077e-06,
+      "loss": 0.0008,
+      "step": 7577
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6851592396645825e-06,
+      "loss": 0.0117,
+      "step": 7578
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685117417034357e-06,
+      "loss": 0.0006,
+      "step": 7579
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6850755944041319e-06,
+      "loss": 0.0025,
+      "step": 7580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6850337717739066e-06,
+      "loss": 0.0011,
+      "step": 7581
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6849919491436814e-06,
+      "loss": 0.0017,
+      "step": 7582
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6849501265134562e-06,
+      "loss": 0.0028,
+      "step": 7583
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6849083038832312e-06,
+      "loss": 0.0013,
+      "step": 7584
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.684866481253006e-06,
+      "loss": 0.0782,
+      "step": 7585
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6848246586227808e-06,
+      "loss": 0.0012,
+      "step": 7586
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6847828359925556e-06,
+      "loss": 0.454,
+      "step": 7587
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6847410133623304e-06,
+      "loss": 1.6557,
+      "step": 7588
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6846991907321052e-06,
+      "loss": 0.0005,
+      "step": 7589
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.68465736810188e-06,
+      "loss": 2.4511,
+      "step": 7590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6846155454716547e-06,
+      "loss": 1.5412,
+      "step": 7591
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6845737228414293e-06,
+      "loss": 3.0587,
+      "step": 7592
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.684531900211204e-06,
+      "loss": 0.001,
+      "step": 7593
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6844900775809789e-06,
+      "loss": 0.0011,
+      "step": 7594
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6844482549507537e-06,
+      "loss": 0.0007,
+      "step": 7595
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6844064323205284e-06,
+      "loss": 1.9583,
+      "step": 7596
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6843646096903032e-06,
+      "loss": 0.001,
+      "step": 7597
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6843227870600782e-06,
+      "loss": 0.0044,
+      "step": 7598
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.684280964429853e-06,
+      "loss": 0.0012,
+      "step": 7599
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6842391417996278e-06,
+      "loss": 0.0012,
+      "step": 7600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6841973191694026e-06,
+      "loss": 0.0095,
+      "step": 7601
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6841554965391774e-06,
+      "loss": 0.0114,
+      "step": 7602
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6841136739089522e-06,
+      "loss": 2.9026,
+      "step": 7603
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.684071851278727e-06,
+      "loss": 0.0016,
+      "step": 7604
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6840300286485017e-06,
+      "loss": 0.0027,
+      "step": 7605
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6839882060182763e-06,
+      "loss": 0.0011,
+      "step": 7606
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.683946383388051e-06,
+      "loss": 1.2595,
+      "step": 7607
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6839045607578259e-06,
+      "loss": 0.0085,
+      "step": 7608
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6838627381276007e-06,
+      "loss": 0.0014,
+      "step": 7609
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6838209154973755e-06,
+      "loss": 0.0353,
+      "step": 7610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6837790928671502e-06,
+      "loss": 0.001,
+      "step": 7611
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.683737270236925e-06,
+      "loss": 0.2694,
+      "step": 7612
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6836954476067e-06,
+      "loss": 0.024,
+      "step": 7613
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6836536249764748e-06,
+      "loss": 0.0021,
+      "step": 7614
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6836118023462496e-06,
+      "loss": 0.4375,
+      "step": 7615
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6835699797160244e-06,
+      "loss": 0.0021,
+      "step": 7616
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6835281570857992e-06,
+      "loss": 0.227,
+      "step": 7617
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.683486334455574e-06,
+      "loss": 0.0005,
+      "step": 7618
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6834445118253485e-06,
+      "loss": 0.5512,
+      "step": 7619
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6834026891951233e-06,
+      "loss": 0.2506,
+      "step": 7620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.683360866564898e-06,
+      "loss": 0.0008,
+      "step": 7621
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6833190439346729e-06,
+      "loss": 0.0022,
+      "step": 7622
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6832772213044477e-06,
+      "loss": 0.0019,
+      "step": 7623
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6832353986742225e-06,
+      "loss": 0.0014,
+      "step": 7624
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6831935760439972e-06,
+      "loss": 0.9951,
+      "step": 7625
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.683151753413772e-06,
+      "loss": 0.0026,
+      "step": 7626
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6831099307835468e-06,
+      "loss": 0.0041,
+      "step": 7627
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6830681081533218e-06,
+      "loss": 0.0042,
+      "step": 7628
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6830262855230966e-06,
+      "loss": 0.001,
+      "step": 7629
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6829844628928714e-06,
+      "loss": 0.0032,
+      "step": 7630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6829426402626462e-06,
+      "loss": 0.0018,
+      "step": 7631
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682900817632421e-06,
+      "loss": 0.0107,
+      "step": 7632
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6828589950021955e-06,
+      "loss": 1.514,
+      "step": 7633
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6828171723719703e-06,
+      "loss": 0.5069,
+      "step": 7634
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6827753497417451e-06,
+      "loss": 0.0016,
+      "step": 7635
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.68273352711152e-06,
+      "loss": 0.8422,
+      "step": 7636
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6826917044812947e-06,
+      "loss": 0.0009,
+      "step": 7637
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6826498818510695e-06,
+      "loss": 0.0026,
+      "step": 7638
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6826080592208443e-06,
+      "loss": 2.2262,
+      "step": 7639
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682566236590619e-06,
+      "loss": 0.0419,
+      "step": 7640
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6825244139603938e-06,
+      "loss": 0.5778,
+      "step": 7641
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6824825913301688e-06,
+      "loss": 0.0013,
+      "step": 7642
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6824407686999436e-06,
+      "loss": 0.0017,
+      "step": 7643
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6823989460697184e-06,
+      "loss": 0.0017,
+      "step": 7644
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6823571234394932e-06,
+      "loss": 0.0615,
+      "step": 7645
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682315300809268e-06,
+      "loss": 0.0047,
+      "step": 7646
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6822734781790425e-06,
+      "loss": 0.0007,
+      "step": 7647
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6822316555488173e-06,
+      "loss": 0.0038,
+      "step": 7648
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6821898329185921e-06,
+      "loss": 0.0006,
+      "step": 7649
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682148010288367e-06,
+      "loss": 0.0409,
+      "step": 7650
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6821061876581417e-06,
+      "loss": 1.8195,
+      "step": 7651
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6820643650279165e-06,
+      "loss": 0.0008,
+      "step": 7652
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6820225423976913e-06,
+      "loss": 0.4009,
+      "step": 7653
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681980719767466e-06,
+      "loss": 0.0014,
+      "step": 7654
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6819388971372408e-06,
+      "loss": 0.0023,
+      "step": 7655
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6818970745070156e-06,
+      "loss": 0.1144,
+      "step": 7656
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6818552518767906e-06,
+      "loss": 2.6049,
+      "step": 7657
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6818134292465654e-06,
+      "loss": 0.002,
+      "step": 7658
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6817716066163402e-06,
+      "loss": 0.0031,
+      "step": 7659
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6817297839861148e-06,
+      "loss": 0.0019,
+      "step": 7660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6816879613558896e-06,
+      "loss": 0.004,
+      "step": 7661
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6816461387256643e-06,
+      "loss": 0.4575,
+      "step": 7662
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6816043160954391e-06,
+      "loss": 0.001,
+      "step": 7663
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681562493465214e-06,
+      "loss": 0.1825,
+      "step": 7664
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6815206708349887e-06,
+      "loss": 0.0331,
+      "step": 7665
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6814788482047635e-06,
+      "loss": 0.0008,
+      "step": 7666
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6814370255745383e-06,
+      "loss": 0.0018,
+      "step": 7667
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681395202944313e-06,
+      "loss": 0.002,
+      "step": 7668
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6813533803140878e-06,
+      "loss": 0.4232,
+      "step": 7669
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6813115576838626e-06,
+      "loss": 0.9692,
+      "step": 7670
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6812697350536374e-06,
+      "loss": 0.0017,
+      "step": 7671
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6812279124234124e-06,
+      "loss": 0.0522,
+      "step": 7672
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6811860897931872e-06,
+      "loss": 0.0008,
+      "step": 7673
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6811442671629618e-06,
+      "loss": 0.5517,
+      "step": 7674
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6811024445327366e-06,
+      "loss": 3.0928,
+      "step": 7675
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6810606219025114e-06,
+      "loss": 0.0005,
+      "step": 7676
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6810187992722861e-06,
+      "loss": 0.0009,
+      "step": 7677
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680976976642061e-06,
+      "loss": 0.0017,
+      "step": 7678
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6809351540118357e-06,
+      "loss": 0.0006,
+      "step": 7679
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6808933313816105e-06,
+      "loss": 0.0038,
+      "step": 7680
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6808515087513853e-06,
+      "loss": 0.0007,
+      "step": 7681
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.68080968612116e-06,
+      "loss": 1.6716,
+      "step": 7682
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6807678634909349e-06,
+      "loss": 0.0012,
+      "step": 7683
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6807260408607096e-06,
+      "loss": 0.0025,
+      "step": 7684
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6806842182304844e-06,
+      "loss": 0.0029,
+      "step": 7685
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6806423956002594e-06,
+      "loss": 0.0007,
+      "step": 7686
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680600572970034e-06,
+      "loss": 0.0052,
+      "step": 7687
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6805587503398088e-06,
+      "loss": 0.0184,
+      "step": 7688
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6805169277095836e-06,
+      "loss": 1.166,
+      "step": 7689
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6804751050793584e-06,
+      "loss": 0.0005,
+      "step": 7690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6804332824491331e-06,
+      "loss": 0.4687,
+      "step": 7691
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680391459818908e-06,
+      "loss": 0.0604,
+      "step": 7692
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6803496371886827e-06,
+      "loss": 0.0005,
+      "step": 7693
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6803078145584575e-06,
+      "loss": 0.0008,
+      "step": 7694
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6802659919282323e-06,
+      "loss": 2.2247,
+      "step": 7695
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680224169298007e-06,
+      "loss": 0.0006,
+      "step": 7696
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6801823466677819e-06,
+      "loss": 0.041,
+      "step": 7697
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6801405240375567e-06,
+      "loss": 2.4517,
+      "step": 7698
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6800987014073314e-06,
+      "loss": 0.0006,
+      "step": 7699
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680056878777106e-06,
+      "loss": 0.0005,
+      "step": 7700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680015056146881e-06,
+      "loss": 0.0045,
+      "step": 7701
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6799732335166558e-06,
+      "loss": 0.2463,
+      "step": 7702
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6799314108864306e-06,
+      "loss": 0.0051,
+      "step": 7703
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6798895882562054e-06,
+      "loss": 0.0014,
+      "step": 7704
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6798477656259802e-06,
+      "loss": 0.265,
+      "step": 7705
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679805942995755e-06,
+      "loss": 0.0907,
+      "step": 7706
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6797641203655297e-06,
+      "loss": 0.0037,
+      "step": 7707
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6797222977353045e-06,
+      "loss": 0.0035,
+      "step": 7708
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6796804751050793e-06,
+      "loss": 0.4197,
+      "step": 7709
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679638652474854e-06,
+      "loss": 0.0009,
+      "step": 7710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6795968298446289e-06,
+      "loss": 0.0049,
+      "step": 7711
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6795550072144037e-06,
+      "loss": 0.0012,
+      "step": 7712
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6795131845841784e-06,
+      "loss": 1.7667,
+      "step": 7713
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679471361953953e-06,
+      "loss": 0.0007,
+      "step": 7714
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6794295393237278e-06,
+      "loss": 2.542,
+      "step": 7715
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6793877166935028e-06,
+      "loss": 0.1733,
+      "step": 7716
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6793458940632776e-06,
+      "loss": 0.0149,
+      "step": 7717
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6793040714330524e-06,
+      "loss": 0.0011,
+      "step": 7718
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6792622488028272e-06,
+      "loss": 0.0019,
+      "step": 7719
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679220426172602e-06,
+      "loss": 0.0016,
+      "step": 7720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6791786035423767e-06,
+      "loss": 0.5989,
+      "step": 7721
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6791367809121515e-06,
+      "loss": 0.477,
+      "step": 7722
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6790949582819263e-06,
+      "loss": 0.001,
+      "step": 7723
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679053135651701e-06,
+      "loss": 2.8162,
+      "step": 7724
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6790113130214759e-06,
+      "loss": 0.2528,
+      "step": 7725
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6789694903912507e-06,
+      "loss": 4.2833,
+      "step": 7726
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6789276677610252e-06,
+      "loss": 0.0008,
+      "step": 7727
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6788858451308e-06,
+      "loss": 0.0862,
+      "step": 7728
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6788440225005748e-06,
+      "loss": 0.0021,
+      "step": 7729
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6788021998703498e-06,
+      "loss": 3.021,
+      "step": 7730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6787603772401246e-06,
+      "loss": 0.0008,
+      "step": 7731
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6787185546098994e-06,
+      "loss": 0.0015,
+      "step": 7732
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6786767319796742e-06,
+      "loss": 0.0009,
+      "step": 7733
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678634909349449e-06,
+      "loss": 0.0048,
+      "step": 7734
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6785930867192237e-06,
+      "loss": 0.0008,
+      "step": 7735
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6785512640889985e-06,
+      "loss": 0.0006,
+      "step": 7736
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6785094414587733e-06,
+      "loss": 0.0012,
+      "step": 7737
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678467618828548e-06,
+      "loss": 0.0011,
+      "step": 7738
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6784257961983229e-06,
+      "loss": 0.0016,
+      "step": 7739
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6783839735680977e-06,
+      "loss": 0.6407,
+      "step": 7740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6783421509378723e-06,
+      "loss": 0.002,
+      "step": 7741
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.678300328307647e-06,
+      "loss": 0.2413,
+      "step": 7742
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6782585056774218e-06,
+      "loss": 0.0071,
+      "step": 7743
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6782166830471966e-06,
+      "loss": 3.2845,
+      "step": 7744
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6781748604169716e-06,
+      "loss": 0.0011,
+      "step": 7745
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6781330377867464e-06,
+      "loss": 2.4922,
+      "step": 7746
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6780912151565212e-06,
+      "loss": 0.0013,
+      "step": 7747
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.678049392526296e-06,
+      "loss": 0.0019,
+      "step": 7748
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6780075698960708e-06,
+      "loss": 0.0049,
+      "step": 7749
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6779657472658455e-06,
+      "loss": 0.0014,
+      "step": 7750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6779239246356203e-06,
+      "loss": 0.0509,
+      "step": 7751
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6778821020053951e-06,
+      "loss": 0.0012,
+      "step": 7752
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67784027937517e-06,
+      "loss": 0.0024,
+      "step": 7753
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6777984567449447e-06,
+      "loss": 0.0987,
+      "step": 7754
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6777566341147193e-06,
+      "loss": 0.0007,
+      "step": 7755
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.677714811484494e-06,
+      "loss": 0.0082,
+      "step": 7756
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6776729888542688e-06,
+      "loss": 0.0188,
+      "step": 7757
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6776311662240436e-06,
+      "loss": 0.0017,
+      "step": 7758
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6775893435938184e-06,
+      "loss": 0.0202,
+      "step": 7759
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6775475209635934e-06,
+      "loss": 0.0015,
+      "step": 7760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6775056983333682e-06,
+      "loss": 0.8081,
+      "step": 7761
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.677463875703143e-06,
+      "loss": 0.0609,
+      "step": 7762
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6774220530729178e-06,
+      "loss": 2.473,
+      "step": 7763
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6773802304426925e-06,
+      "loss": 0.0023,
+      "step": 7764
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6773384078124673e-06,
+      "loss": 1.9446,
+      "step": 7765
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6772965851822421e-06,
+      "loss": 1.7902,
+      "step": 7766
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.677254762552017e-06,
+      "loss": 0.002,
+      "step": 7767
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6772129399217915e-06,
+      "loss": 0.0082,
+      "step": 7768
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6771711172915663e-06,
+      "loss": 0.0007,
+      "step": 7769
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.677129294661341e-06,
+      "loss": 0.0094,
+      "step": 7770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6770874720311158e-06,
+      "loss": 0.002,
+      "step": 7771
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6770456494008906e-06,
+      "loss": 0.0015,
+      "step": 7772
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6770038267706654e-06,
+      "loss": 0.0026,
+      "step": 7773
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6769620041404404e-06,
+      "loss": 1.4456,
+      "step": 7774
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6769201815102152e-06,
+      "loss": 1.2846,
+      "step": 7775
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67687835887999e-06,
+      "loss": 0.0015,
+      "step": 7776
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6768365362497648e-06,
+      "loss": 0.0013,
+      "step": 7777
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6767947136195396e-06,
+      "loss": 0.009,
+      "step": 7778
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6767528909893143e-06,
+      "loss": 0.0008,
+      "step": 7779
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6767110683590891e-06,
+      "loss": 0.7678,
+      "step": 7780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.676669245728864e-06,
+      "loss": 1.5102,
+      "step": 7781
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6766274230986385e-06,
+      "loss": 0.0022,
+      "step": 7782
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6765856004684133e-06,
+      "loss": 0.0038,
+      "step": 7783
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.676543777838188e-06,
+      "loss": 0.0047,
+      "step": 7784
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6765019552079628e-06,
+      "loss": 0.0007,
+      "step": 7785
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6764601325777376e-06,
+      "loss": 0.0019,
+      "step": 7786
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6764183099475124e-06,
+      "loss": 0.0024,
+      "step": 7787
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6763764873172872e-06,
+      "loss": 0.0114,
+      "step": 7788
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6763346646870622e-06,
+      "loss": 0.0073,
+      "step": 7789
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.676292842056837e-06,
+      "loss": 0.0018,
+      "step": 7790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6762510194266118e-06,
+      "loss": 0.0067,
+      "step": 7791
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6762091967963866e-06,
+      "loss": 0.0013,
+      "step": 7792
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6761673741661614e-06,
+      "loss": 3.1843,
+      "step": 7793
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6761255515359361e-06,
+      "loss": 0.0032,
+      "step": 7794
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6760837289057107e-06,
+      "loss": 0.002,
+      "step": 7795
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6760419062754855e-06,
+      "loss": 0.0012,
+      "step": 7796
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6760000836452603e-06,
+      "loss": 0.0126,
+      "step": 7797
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675958261015035e-06,
+      "loss": 0.0116,
+      "step": 7798
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6759164383848099e-06,
+      "loss": 0.303,
+      "step": 7799
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6758746157545846e-06,
+      "loss": 0.3446,
+      "step": 7800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6758327931243594e-06,
+      "loss": 0.0016,
+      "step": 7801
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6757909704941342e-06,
+      "loss": 0.0102,
+      "step": 7802
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6757491478639092e-06,
+      "loss": 0.2476,
+      "step": 7803
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675707325233684e-06,
+      "loss": 1.1109,
+      "step": 7804
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6756655026034588e-06,
+      "loss": 0.002,
+      "step": 7805
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6756236799732336e-06,
+      "loss": 0.0054,
+      "step": 7806
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6755818573430084e-06,
+      "loss": 0.0014,
+      "step": 7807
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6755400347127831e-06,
+      "loss": 0.0943,
+      "step": 7808
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6754982120825577e-06,
+      "loss": 0.0748,
+      "step": 7809
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6754563894523325e-06,
+      "loss": 0.0016,
+      "step": 7810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6754145668221073e-06,
+      "loss": 0.2303,
+      "step": 7811
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675372744191882e-06,
+      "loss": 0.2381,
+      "step": 7812
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6753309215616569e-06,
+      "loss": 0.001,
+      "step": 7813
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6752890989314317e-06,
+      "loss": 0.0024,
+      "step": 7814
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6752472763012064e-06,
+      "loss": 0.004,
+      "step": 7815
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6752054536709812e-06,
+      "loss": 0.0036,
+      "step": 7816
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675163631040756e-06,
+      "loss": 4.1178,
+      "step": 7817
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675121808410531e-06,
+      "loss": 0.0147,
+      "step": 7818
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6750799857803058e-06,
+      "loss": 0.004,
+      "step": 7819
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6750381631500806e-06,
+      "loss": 0.0017,
+      "step": 7820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6749963405198554e-06,
+      "loss": 2.0841,
+      "step": 7821
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67495451788963e-06,
+      "loss": 0.006,
+      "step": 7822
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6749126952594047e-06,
+      "loss": 0.0019,
+      "step": 7823
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6748708726291795e-06,
+      "loss": 1.6666,
+      "step": 7824
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6748290499989543e-06,
+      "loss": 0.0293,
+      "step": 7825
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674787227368729e-06,
+      "loss": 0.0018,
+      "step": 7826
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6747454047385039e-06,
+      "loss": 0.0009,
+      "step": 7827
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6747035821082787e-06,
+      "loss": 2.5057,
+      "step": 7828
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6746617594780534e-06,
+      "loss": 0.0007,
+      "step": 7829
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6746199368478282e-06,
+      "loss": 0.0013,
+      "step": 7830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674578114217603e-06,
+      "loss": 0.001,
+      "step": 7831
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6745362915873778e-06,
+      "loss": 0.0007,
+      "step": 7832
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6744944689571528e-06,
+      "loss": 0.9875,
+      "step": 7833
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6744526463269276e-06,
+      "loss": 0.0014,
+      "step": 7834
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6744108236967024e-06,
+      "loss": 0.0013,
+      "step": 7835
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674369001066477e-06,
+      "loss": 0.0015,
+      "step": 7836
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6743271784362517e-06,
+      "loss": 0.0014,
+      "step": 7837
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6742853558060265e-06,
+      "loss": 0.0014,
+      "step": 7838
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6742435331758013e-06,
+      "loss": 0.0006,
+      "step": 7839
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674201710545576e-06,
+      "loss": 0.0026,
+      "step": 7840
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6741598879153509e-06,
+      "loss": 0.0316,
+      "step": 7841
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6741180652851257e-06,
+      "loss": 0.055,
+      "step": 7842
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6740762426549005e-06,
+      "loss": 0.0049,
+      "step": 7843
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6740344200246752e-06,
+      "loss": 2.94,
+      "step": 7844
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67399259739445e-06,
+      "loss": 0.004,
+      "step": 7845
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6739507747642248e-06,
+      "loss": 0.0059,
+      "step": 7846
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6739089521339998e-06,
+      "loss": 0.6577,
+      "step": 7847
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6738671295037746e-06,
+      "loss": 0.0468,
+      "step": 7848
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6738253068735494e-06,
+      "loss": 0.0146,
+      "step": 7849
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673783484243324e-06,
+      "loss": 0.1459,
+      "step": 7850
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6737416616130987e-06,
+      "loss": 0.0017,
+      "step": 7851
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6736998389828735e-06,
+      "loss": 0.0059,
+      "step": 7852
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6736580163526483e-06,
+      "loss": 0.0023,
+      "step": 7853
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6736161937224231e-06,
+      "loss": 2.8265,
+      "step": 7854
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6735743710921979e-06,
+      "loss": 0.0546,
+      "step": 7855
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6735325484619727e-06,
+      "loss": 0.012,
+      "step": 7856
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6734907258317475e-06,
+      "loss": 0.1771,
+      "step": 7857
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6734489032015223e-06,
+      "loss": 0.2699,
+      "step": 7858
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673407080571297e-06,
+      "loss": 0.2529,
+      "step": 7859
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6733652579410718e-06,
+      "loss": 0.0815,
+      "step": 7860
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6733234353108466e-06,
+      "loss": 0.083,
+      "step": 7861
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6732816126806216e-06,
+      "loss": 1.7115,
+      "step": 7862
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6732397900503962e-06,
+      "loss": 0.0011,
+      "step": 7863
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673197967420171e-06,
+      "loss": 0.0015,
+      "step": 7864
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6731561447899458e-06,
+      "loss": 0.0013,
+      "step": 7865
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6731143221597205e-06,
+      "loss": 0.0024,
+      "step": 7866
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6730724995294953e-06,
+      "loss": 0.006,
+      "step": 7867
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6730306768992701e-06,
+      "loss": 1.4571,
+      "step": 7868
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672988854269045e-06,
+      "loss": 0.7948,
+      "step": 7869
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6729470316388197e-06,
+      "loss": 0.0012,
+      "step": 7870
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6729052090085945e-06,
+      "loss": 0.0016,
+      "step": 7871
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6728633863783693e-06,
+      "loss": 0.0032,
+      "step": 7872
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672821563748144e-06,
+      "loss": 0.0071,
+      "step": 7873
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6727797411179188e-06,
+      "loss": 0.0037,
+      "step": 7874
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6727379184876936e-06,
+      "loss": 1.4391,
+      "step": 7875
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6726960958574682e-06,
+      "loss": 0.1544,
+      "step": 7876
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6726542732272432e-06,
+      "loss": 0.0326,
+      "step": 7877
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672612450597018e-06,
+      "loss": 2.7223,
+      "step": 7878
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6725706279667928e-06,
+      "loss": 0.0071,
+      "step": 7879
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6725288053365676e-06,
+      "loss": 0.0037,
+      "step": 7880
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6724869827063423e-06,
+      "loss": 0.0008,
+      "step": 7881
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6724451600761171e-06,
+      "loss": 1.8391,
+      "step": 7882
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672403337445892e-06,
+      "loss": 0.0042,
+      "step": 7883
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6723615148156667e-06,
+      "loss": 0.0294,
+      "step": 7884
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6723196921854415e-06,
+      "loss": 0.022,
+      "step": 7885
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6722778695552163e-06,
+      "loss": 0.6093,
+      "step": 7886
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672236046924991e-06,
+      "loss": 0.1793,
+      "step": 7887
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6721942242947658e-06,
+      "loss": 0.0008,
+      "step": 7888
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6721524016645406e-06,
+      "loss": 0.0061,
+      "step": 7889
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6721105790343152e-06,
+      "loss": 0.0054,
+      "step": 7890
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6720687564040902e-06,
+      "loss": 2.6626,
+      "step": 7891
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672026933773865e-06,
+      "loss": 0.0024,
+      "step": 7892
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6719851111436398e-06,
+      "loss": 0.0358,
+      "step": 7893
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6719432885134146e-06,
+      "loss": 1.9874,
+      "step": 7894
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6719014658831893e-06,
+      "loss": 0.0023,
+      "step": 7895
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6718596432529641e-06,
+      "loss": 0.0128,
+      "step": 7896
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671817820622739e-06,
+      "loss": 0.0058,
+      "step": 7897
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6717759979925137e-06,
+      "loss": 1.6308,
+      "step": 7898
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6717341753622885e-06,
+      "loss": 0.0329,
+      "step": 7899
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6716923527320633e-06,
+      "loss": 3.224,
+      "step": 7900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.671650530101838e-06,
+      "loss": 0.0011,
+      "step": 7901
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6716087074716128e-06,
+      "loss": 0.2663,
+      "step": 7902
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6715668848413874e-06,
+      "loss": 0.0024,
+      "step": 7903
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6715250622111622e-06,
+      "loss": 0.7675,
+      "step": 7904
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.671483239580937e-06,
+      "loss": 0.0061,
+      "step": 7905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.671441416950712e-06,
+      "loss": 0.0022,
+      "step": 7906
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6713995943204868e-06,
+      "loss": 1.955,
+      "step": 7907
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6713577716902616e-06,
+      "loss": 0.0026,
+      "step": 7908
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6713159490600364e-06,
+      "loss": 0.2061,
+      "step": 7909
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6712741264298111e-06,
+      "loss": 0.0207,
+      "step": 7910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.671232303799586e-06,
+      "loss": 0.0047,
+      "step": 7911
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6711904811693607e-06,
+      "loss": 0.0048,
+      "step": 7912
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6711486585391355e-06,
+      "loss": 0.01,
+      "step": 7913
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6711068359089103e-06,
+      "loss": 0.0023,
+      "step": 7914
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.671065013278685e-06,
+      "loss": 2.7478,
+      "step": 7915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6710231906484599e-06,
+      "loss": 0.0016,
+      "step": 7916
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6709813680182344e-06,
+      "loss": 0.0095,
+      "step": 7917
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6709395453880092e-06,
+      "loss": 0.001,
+      "step": 7918
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.670897722757784e-06,
+      "loss": 0.0013,
+      "step": 7919
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6708559001275588e-06,
+      "loss": 1.3903,
+      "step": 7920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6708140774973338e-06,
+      "loss": 0.0138,
+      "step": 7921
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6707722548671086e-06,
+      "loss": 0.0009,
+      "step": 7922
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6707304322368834e-06,
+      "loss": 0.003,
+      "step": 7923
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6706886096066581e-06,
+      "loss": 2.7242,
+      "step": 7924
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.670646786976433e-06,
+      "loss": 0.0016,
+      "step": 7925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6706049643462077e-06,
+      "loss": 0.0008,
+      "step": 7926
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6705631417159825e-06,
+      "loss": 0.0012,
+      "step": 7927
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6705213190857573e-06,
+      "loss": 0.0013,
+      "step": 7928
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.670479496455532e-06,
+      "loss": 0.1736,
+      "step": 7929
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6704376738253067e-06,
+      "loss": 0.0044,
+      "step": 7930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6703958511950814e-06,
+      "loss": 0.0034,
+      "step": 7931
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6703540285648562e-06,
+      "loss": 0.001,
+      "step": 7932
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.670312205934631e-06,
+      "loss": 0.001,
+      "step": 7933
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6702703833044058e-06,
+      "loss": 0.0007,
+      "step": 7934
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6702285606741808e-06,
+      "loss": 0.7254,
+      "step": 7935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6701867380439556e-06,
+      "loss": 0.022,
+      "step": 7936
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6701449154137304e-06,
+      "loss": 0.0012,
+      "step": 7937
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6701030927835052e-06,
+      "loss": 0.1988,
+      "step": 7938
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.67006127015328e-06,
+      "loss": 0.001,
+      "step": 7939
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6700194475230547e-06,
+      "loss": 0.0011,
+      "step": 7940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6699776248928295e-06,
+      "loss": 0.001,
+      "step": 7941
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6699358022626043e-06,
+      "loss": 0.0152,
+      "step": 7942
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669893979632379e-06,
+      "loss": 0.0021,
+      "step": 7943
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6698521570021537e-06,
+      "loss": 0.0021,
+      "step": 7944
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6698103343719285e-06,
+      "loss": 0.322,
+      "step": 7945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6697685117417032e-06,
+      "loss": 0.0025,
+      "step": 7946
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669726689111478e-06,
+      "loss": 0.0021,
+      "step": 7947
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6696848664812528e-06,
+      "loss": 0.5922,
+      "step": 7948
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6696430438510276e-06,
+      "loss": 0.0019,
+      "step": 7949
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6696012212208026e-06,
+      "loss": 0.0026,
+      "step": 7950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6695593985905774e-06,
+      "loss": 0.07,
+      "step": 7951
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6695175759603522e-06,
+      "loss": 0.003,
+      "step": 7952
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669475753330127e-06,
+      "loss": 0.3785,
+      "step": 7953
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6694339306999017e-06,
+      "loss": 0.001,
+      "step": 7954
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6693921080696765e-06,
+      "loss": 0.0011,
+      "step": 7955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6693502854394513e-06,
+      "loss": 0.0013,
+      "step": 7956
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669308462809226e-06,
+      "loss": 0.0024,
+      "step": 7957
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6692666401790007e-06,
+      "loss": 0.0027,
+      "step": 7958
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6692248175487755e-06,
+      "loss": 2.1348,
+      "step": 7959
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6691829949185502e-06,
+      "loss": 1.4256,
+      "step": 7960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669141172288325e-06,
+      "loss": 0.5957,
+      "step": 7961
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6690993496580998e-06,
+      "loss": 2.5622,
+      "step": 7962
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6690575270278746e-06,
+      "loss": 0.0304,
+      "step": 7963
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6690157043976494e-06,
+      "loss": 0.0035,
+      "step": 7964
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6689738817674244e-06,
+      "loss": 0.0006,
+      "step": 7965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6689320591371992e-06,
+      "loss": 3.4287,
+      "step": 7966
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668890236506974e-06,
+      "loss": 0.0013,
+      "step": 7967
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6688484138767487e-06,
+      "loss": 0.7688,
+      "step": 7968
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6688065912465235e-06,
+      "loss": 0.0039,
+      "step": 7969
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6687647686162983e-06,
+      "loss": 1.0135,
+      "step": 7970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668722945986073e-06,
+      "loss": 0.0007,
+      "step": 7971
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6686811233558477e-06,
+      "loss": 0.3249,
+      "step": 7972
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6686393007256225e-06,
+      "loss": 0.9201,
+      "step": 7973
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6685974780953973e-06,
+      "loss": 0.006,
+      "step": 7974
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668555655465172e-06,
+      "loss": 0.0012,
+      "step": 7975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6685138328349468e-06,
+      "loss": 0.0008,
+      "step": 7976
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6684720102047216e-06,
+      "loss": 0.0013,
+      "step": 7977
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6684301875744964e-06,
+      "loss": 0.0015,
+      "step": 7978
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6683883649442714e-06,
+      "loss": 0.4828,
+      "step": 7979
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6683465423140462e-06,
+      "loss": 0.0018,
+      "step": 7980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668304719683821e-06,
+      "loss": 0.0026,
+      "step": 7981
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6682628970535958e-06,
+      "loss": 0.0057,
+      "step": 7982
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6682210744233705e-06,
+      "loss": 0.2956,
+      "step": 7983
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6681792517931453e-06,
+      "loss": 0.621,
+      "step": 7984
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66813742916292e-06,
+      "loss": 0.013,
+      "step": 7985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6680956065326947e-06,
+      "loss": 0.0008,
+      "step": 7986
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6680537839024695e-06,
+      "loss": 0.001,
+      "step": 7987
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6680119612722443e-06,
+      "loss": 0.0069,
+      "step": 7988
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667970138642019e-06,
+      "loss": 0.1229,
+      "step": 7989
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6679283160117938e-06,
+      "loss": 0.0036,
+      "step": 7990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6678864933815686e-06,
+      "loss": 0.0008,
+      "step": 7991
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6678446707513434e-06,
+      "loss": 0.0695,
+      "step": 7992
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6678028481211182e-06,
+      "loss": 0.0102,
+      "step": 7993
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6677610254908932e-06,
+      "loss": 0.3207,
+      "step": 7994
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667719202860668e-06,
+      "loss": 0.0011,
+      "step": 7995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6676773802304428e-06,
+      "loss": 1.6786,
+      "step": 7996
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6676355576002176e-06,
+      "loss": 0.001,
+      "step": 7997
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6675937349699921e-06,
+      "loss": 0.0019,
+      "step": 7998
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667551912339767e-06,
+      "loss": 1.3029,
+      "step": 7999
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6675100897095417e-06,
+      "loss": 0.1583,
+      "step": 8000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6674682670793165e-06,
+      "loss": 0.0024,
+      "step": 8001
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6674264444490913e-06,
+      "loss": 0.0026,
+      "step": 8002
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667384621818866e-06,
+      "loss": 0.0012,
+      "step": 8003
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6673427991886408e-06,
+      "loss": 1.9007,
+      "step": 8004
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6673009765584156e-06,
+      "loss": 0.0032,
+      "step": 8005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6672591539281904e-06,
+      "loss": 0.0008,
+      "step": 8006
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6672173312979652e-06,
+      "loss": 0.0014,
+      "step": 8007
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6671755086677402e-06,
+      "loss": 0.0029,
+      "step": 8008
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667133686037515e-06,
+      "loss": 0.0083,
+      "step": 8009
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6670918634072898e-06,
+      "loss": 0.1778,
+      "step": 8010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6670500407770646e-06,
+      "loss": 0.0012,
+      "step": 8011
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6670082181468391e-06,
+      "loss": 0.0019,
+      "step": 8012
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666966395516614e-06,
+      "loss": 0.1411,
+      "step": 8013
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6669245728863887e-06,
+      "loss": 0.1044,
+      "step": 8014
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6668827502561635e-06,
+      "loss": 0.0006,
+      "step": 8015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6668409276259383e-06,
+      "loss": 0.3125,
+      "step": 8016
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666799104995713e-06,
+      "loss": 0.0027,
+      "step": 8017
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6667572823654879e-06,
+      "loss": 0.0009,
+      "step": 8018
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6667154597352626e-06,
+      "loss": 0.3631,
+      "step": 8019
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6666736371050374e-06,
+      "loss": 0.0007,
+      "step": 8020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6666318144748122e-06,
+      "loss": 0.0031,
+      "step": 8021
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666589991844587e-06,
+      "loss": 2.1813,
+      "step": 8022
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666548169214362e-06,
+      "loss": 0.0016,
+      "step": 8023
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6665063465841368e-06,
+      "loss": 0.0008,
+      "step": 8024
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6664645239539114e-06,
+      "loss": 0.0008,
+      "step": 8025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6664227013236861e-06,
+      "loss": 0.0016,
+      "step": 8026
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666380878693461e-06,
+      "loss": 0.0038,
+      "step": 8027
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6663390560632357e-06,
+      "loss": 0.0025,
+      "step": 8028
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6662972334330105e-06,
+      "loss": 0.0012,
+      "step": 8029
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6662554108027853e-06,
+      "loss": 0.6328,
+      "step": 8030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66621358817256e-06,
+      "loss": 2.4613,
+      "step": 8031
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6661717655423349e-06,
+      "loss": 2.7168,
+      "step": 8032
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6661299429121096e-06,
+      "loss": 3.2174,
+      "step": 8033
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6660881202818844e-06,
+      "loss": 0.0031,
+      "step": 8034
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6660462976516592e-06,
+      "loss": 0.252,
+      "step": 8035
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666004475021434e-06,
+      "loss": 0.0111,
+      "step": 8036
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6659626523912088e-06,
+      "loss": 0.0999,
+      "step": 8037
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6659208297609838e-06,
+      "loss": 0.0007,
+      "step": 8038
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6658790071307584e-06,
+      "loss": 0.1192,
+      "step": 8039
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6658371845005332e-06,
+      "loss": 0.0014,
+      "step": 8040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665795361870308e-06,
+      "loss": 2.1563,
+      "step": 8041
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6657535392400827e-06,
+      "loss": 0.0006,
+      "step": 8042
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6657117166098575e-06,
+      "loss": 2.6136,
+      "step": 8043
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6656698939796323e-06,
+      "loss": 0.0467,
+      "step": 8044
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665628071349407e-06,
+      "loss": 0.0011,
+      "step": 8045
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6655862487191819e-06,
+      "loss": 0.0016,
+      "step": 8046
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6655444260889567e-06,
+      "loss": 0.0025,
+      "step": 8047
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6655026034587314e-06,
+      "loss": 0.0013,
+      "step": 8048
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6654607808285062e-06,
+      "loss": 2.6033,
+      "step": 8049
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665418958198281e-06,
+      "loss": 0.0008,
+      "step": 8050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6653771355680558e-06,
+      "loss": 0.0018,
+      "step": 8051
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6653353129378308e-06,
+      "loss": 0.0022,
+      "step": 8052
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6652934903076054e-06,
+      "loss": 0.0026,
+      "step": 8053
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6652516676773802e-06,
+      "loss": 0.0012,
+      "step": 8054
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665209845047155e-06,
+      "loss": 0.0165,
+      "step": 8055
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6651680224169297e-06,
+      "loss": 1.6846,
+      "step": 8056
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6651261997867045e-06,
+      "loss": 1.5324,
+      "step": 8057
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6650843771564793e-06,
+      "loss": 0.0022,
+      "step": 8058
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.665042554526254e-06,
+      "loss": 0.0113,
+      "step": 8059
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6650007318960289e-06,
+      "loss": 0.0035,
+      "step": 8060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6649589092658037e-06,
+      "loss": 1.9628,
+      "step": 8061
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6649170866355785e-06,
+      "loss": 0.0014,
+      "step": 8062
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6648752640053532e-06,
+      "loss": 1.2987,
+      "step": 8063
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.664833441375128e-06,
+      "loss": 0.0021,
+      "step": 8064
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6647916187449026e-06,
+      "loss": 0.0012,
+      "step": 8065
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6647497961146774e-06,
+      "loss": 0.0055,
+      "step": 8066
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6647079734844524e-06,
+      "loss": 0.0027,
+      "step": 8067
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6646661508542272e-06,
+      "loss": 0.0024,
+      "step": 8068
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.664624328224002e-06,
+      "loss": 0.0013,
+      "step": 8069
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6645825055937767e-06,
+      "loss": 0.0017,
+      "step": 8070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6645406829635515e-06,
+      "loss": 0.0029,
+      "step": 8071
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6644988603333263e-06,
+      "loss": 0.0017,
+      "step": 8072
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.664457037703101e-06,
+      "loss": 0.0091,
+      "step": 8073
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6644152150728759e-06,
+      "loss": 0.0237,
+      "step": 8074
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6643733924426507e-06,
+      "loss": 0.0077,
+      "step": 8075
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6643315698124255e-06,
+      "loss": 0.0035,
+      "step": 8076
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6642897471822002e-06,
+      "loss": 0.0127,
+      "step": 8077
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.664247924551975e-06,
+      "loss": 0.2719,
+      "step": 8078
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6642061019217496e-06,
+      "loss": 0.0028,
+      "step": 8079
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6641642792915244e-06,
+      "loss": 0.0052,
+      "step": 8080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6641224566612992e-06,
+      "loss": 0.1361,
+      "step": 8081
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6640806340310742e-06,
+      "loss": 1.5188,
+      "step": 8082
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.664038811400849e-06,
+      "loss": 0.1589,
+      "step": 8083
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6639969887706237e-06,
+      "loss": 0.2114,
+      "step": 8084
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6639551661403985e-06,
+      "loss": 0.2594,
+      "step": 8085
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6639133435101733e-06,
+      "loss": 0.0038,
+      "step": 8086
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6638715208799481e-06,
+      "loss": 0.0017,
+      "step": 8087
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.663829698249723e-06,
+      "loss": 0.5848,
+      "step": 8088
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6637878756194977e-06,
+      "loss": 0.0019,
+      "step": 8089
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6637460529892725e-06,
+      "loss": 0.5894,
+      "step": 8090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6637042303590473e-06,
+      "loss": 0.6503,
+      "step": 8091
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.663662407728822e-06,
+      "loss": 0.3469,
+      "step": 8092
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6636205850985966e-06,
+      "loss": 0.0033,
+      "step": 8093
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6635787624683714e-06,
+      "loss": 0.002,
+      "step": 8094
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6635369398381462e-06,
+      "loss": 0.1426,
+      "step": 8095
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6634951172079212e-06,
+      "loss": 0.0011,
+      "step": 8096
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.663453294577696e-06,
+      "loss": 0.0024,
+      "step": 8097
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6634114719474708e-06,
+      "loss": 0.0013,
+      "step": 8098
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6633696493172455e-06,
+      "loss": 0.0034,
+      "step": 8099
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6633278266870203e-06,
+      "loss": 0.021,
+      "step": 8100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6632860040567951e-06,
+      "loss": 0.0067,
+      "step": 8101
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.66324418142657e-06,
+      "loss": 0.0017,
+      "step": 8102
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6632023587963447e-06,
+      "loss": 4.2313,
+      "step": 8103
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6631605361661195e-06,
+      "loss": 0.001,
+      "step": 8104
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6631187135358943e-06,
+      "loss": 0.002,
+      "step": 8105
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6630768909056688e-06,
+      "loss": 0.0012,
+      "step": 8106
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6630350682754436e-06,
+      "loss": 1.5855,
+      "step": 8107
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6629932456452184e-06,
+      "loss": 1.6958,
+      "step": 8108
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6629514230149932e-06,
+      "loss": 0.0026,
+      "step": 8109
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662909600384768e-06,
+      "loss": 0.0031,
+      "step": 8110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662867777754543e-06,
+      "loss": 2.1431,
+      "step": 8111
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6628259551243178e-06,
+      "loss": 0.0017,
+      "step": 8112
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6627841324940926e-06,
+      "loss": 0.0022,
+      "step": 8113
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6627423098638673e-06,
+      "loss": 1.4857,
+      "step": 8114
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6627004872336421e-06,
+      "loss": 0.0053,
+      "step": 8115
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662658664603417e-06,
+      "loss": 0.0819,
+      "step": 8116
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6626168419731917e-06,
+      "loss": 0.0009,
+      "step": 8117
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6625750193429665e-06,
+      "loss": 0.0258,
+      "step": 8118
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6625331967127413e-06,
+      "loss": 0.003,
+      "step": 8119
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6624913740825158e-06,
+      "loss": 0.7241,
+      "step": 8120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6624495514522906e-06,
+      "loss": 2.3398,
+      "step": 8121
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6624077288220654e-06,
+      "loss": 0.0019,
+      "step": 8122
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6623659061918402e-06,
+      "loss": 0.0068,
+      "step": 8123
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662324083561615e-06,
+      "loss": 0.9055,
+      "step": 8124
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6622822609313898e-06,
+      "loss": 2.8388,
+      "step": 8125
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6622404383011648e-06,
+      "loss": 0.0012,
+      "step": 8126
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6621986156709396e-06,
+      "loss": 0.0025,
+      "step": 8127
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6621567930407143e-06,
+      "loss": 0.0129,
+      "step": 8128
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6621149704104891e-06,
+      "loss": 0.0056,
+      "step": 8129
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662073147780264e-06,
+      "loss": 0.0032,
+      "step": 8130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6620313251500387e-06,
+      "loss": 0.0015,
+      "step": 8131
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6619895025198135e-06,
+      "loss": 0.0045,
+      "step": 8132
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661947679889588e-06,
+      "loss": 0.0052,
+      "step": 8133
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6619058572593629e-06,
+      "loss": 0.0143,
+      "step": 8134
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6618640346291376e-06,
+      "loss": 2.8625,
+      "step": 8135
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6618222119989124e-06,
+      "loss": 0.0129,
+      "step": 8136
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6617803893686872e-06,
+      "loss": 0.0021,
+      "step": 8137
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661738566738462e-06,
+      "loss": 0.0805,
+      "step": 8138
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6616967441082368e-06,
+      "loss": 0.1147,
+      "step": 8139
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6616549214780118e-06,
+      "loss": 0.0011,
+      "step": 8140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6616130988477866e-06,
+      "loss": 0.0807,
+      "step": 8141
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6615712762175614e-06,
+      "loss": 0.0044,
+      "step": 8142
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6615294535873361e-06,
+      "loss": 0.0027,
+      "step": 8143
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661487630957111e-06,
+      "loss": 0.4565,
+      "step": 8144
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6614458083268857e-06,
+      "loss": 0.0068,
+      "step": 8145
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6614039856966605e-06,
+      "loss": 0.0019,
+      "step": 8146
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661362163066435e-06,
+      "loss": 0.1059,
+      "step": 8147
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6613203404362099e-06,
+      "loss": 0.0227,
+      "step": 8148
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6612785178059846e-06,
+      "loss": 0.2887,
+      "step": 8149
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6612366951757594e-06,
+      "loss": 0.0026,
+      "step": 8150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6611948725455342e-06,
+      "loss": 1.4322,
+      "step": 8151
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661153049915309e-06,
+      "loss": 0.0015,
+      "step": 8152
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6611112272850838e-06,
+      "loss": 0.0011,
+      "step": 8153
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6610694046548586e-06,
+      "loss": 0.0029,
+      "step": 8154
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6610275820246336e-06,
+      "loss": 0.5336,
+      "step": 8155
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6609857593944084e-06,
+      "loss": 0.1063,
+      "step": 8156
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6609439367641832e-06,
+      "loss": 0.0012,
+      "step": 8157
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660902114133958e-06,
+      "loss": 0.2262,
+      "step": 8158
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6608602915037327e-06,
+      "loss": 0.5843,
+      "step": 8159
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6608184688735075e-06,
+      "loss": 2.1133,
+      "step": 8160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660776646243282e-06,
+      "loss": 0.2082,
+      "step": 8161
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6607348236130569e-06,
+      "loss": 0.0282,
+      "step": 8162
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6606930009828317e-06,
+      "loss": 0.0006,
+      "step": 8163
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6606511783526064e-06,
+      "loss": 0.0037,
+      "step": 8164
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6606093557223812e-06,
+      "loss": 0.0022,
+      "step": 8165
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660567533092156e-06,
+      "loss": 0.004,
+      "step": 8166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6605257104619308e-06,
+      "loss": 0.0008,
+      "step": 8167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6604838878317056e-06,
+      "loss": 0.0059,
+      "step": 8168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6604420652014804e-06,
+      "loss": 0.0012,
+      "step": 8169
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6604002425712554e-06,
+      "loss": 0.0106,
+      "step": 8170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6603584199410302e-06,
+      "loss": 0.0028,
+      "step": 8171
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660316597310805e-06,
+      "loss": 0.0134,
+      "step": 8172
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6602747746805797e-06,
+      "loss": 0.0026,
+      "step": 8173
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6602329520503543e-06,
+      "loss": 0.0014,
+      "step": 8174
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660191129420129e-06,
+      "loss": 0.0015,
+      "step": 8175
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6601493067899039e-06,
+      "loss": 0.0009,
+      "step": 8176
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6601074841596787e-06,
+      "loss": 0.0013,
+      "step": 8177
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6600656615294535e-06,
+      "loss": 0.0064,
+      "step": 8178
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6600238388992282e-06,
+      "loss": 0.0075,
+      "step": 8179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659982016269003e-06,
+      "loss": 0.0022,
+      "step": 8180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6599401936387778e-06,
+      "loss": 0.0008,
+      "step": 8181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6598983710085526e-06,
+      "loss": 0.0017,
+      "step": 8182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6598565483783274e-06,
+      "loss": 0.001,
+      "step": 8183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6598147257481024e-06,
+      "loss": 0.0011,
+      "step": 8184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6597729031178772e-06,
+      "loss": 0.0014,
+      "step": 8185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659731080487652e-06,
+      "loss": 0.0022,
+      "step": 8186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6596892578574267e-06,
+      "loss": 0.0021,
+      "step": 8187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6596474352272013e-06,
+      "loss": 0.0005,
+      "step": 8188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659605612596976e-06,
+      "loss": 0.0066,
+      "step": 8189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6595637899667509e-06,
+      "loss": 0.8505,
+      "step": 8190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6595219673365257e-06,
+      "loss": 0.0014,
+      "step": 8191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6594801447063005e-06,
+      "loss": 0.0351,
+      "step": 8192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6594383220760752e-06,
+      "loss": 0.0072,
+      "step": 8193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.65939649944585e-06,
+      "loss": 0.0024,
+      "step": 8194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6593546768156248e-06,
+      "loss": 0.0019,
+      "step": 8195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6593128541853996e-06,
+      "loss": 0.0006,
+      "step": 8196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6592710315551744e-06,
+      "loss": 0.9448,
+      "step": 8197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6592292089249492e-06,
+      "loss": 1.3479,
+      "step": 8198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6591873862947242e-06,
+      "loss": 0.0022,
+      "step": 8199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659145563664499e-06,
+      "loss": 0.001,
+      "step": 8200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6591037410342735e-06,
+      "loss": 1.4501,
+      "step": 8201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6590619184040483e-06,
+      "loss": 0.0073,
+      "step": 8202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6590200957738231e-06,
+      "loss": 0.6534,
+      "step": 8203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658978273143598e-06,
+      "loss": 0.0062,
+      "step": 8204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6589364505133727e-06,
+      "loss": 1.4296,
+      "step": 8205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6588946278831475e-06,
+      "loss": 0.0051,
+      "step": 8206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6588528052529223e-06,
+      "loss": 0.1019,
+      "step": 8207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658810982622697e-06,
+      "loss": 0.1219,
+      "step": 8208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6587691599924718e-06,
+      "loss": 0.0007,
+      "step": 8209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6587273373622466e-06,
+      "loss": 0.1147,
+      "step": 8210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6586855147320214e-06,
+      "loss": 0.0012,
+      "step": 8211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6586436921017962e-06,
+      "loss": 0.0008,
+      "step": 8212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6586018694715712e-06,
+      "loss": 0.001,
+      "step": 8213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658560046841346e-06,
+      "loss": 0.0085,
+      "step": 8214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6585182242111205e-06,
+      "loss": 0.0052,
+      "step": 8215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6584764015808953e-06,
+      "loss": 0.0012,
+      "step": 8216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6584345789506701e-06,
+      "loss": 2.1136,
+      "step": 8217
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.658392756320445e-06,
+      "loss": 0.0033,
+      "step": 8218
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6583509336902197e-06,
+      "loss": 0.0018,
+      "step": 8219
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6583091110599945e-06,
+      "loss": 0.0018,
+      "step": 8220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6582672884297693e-06,
+      "loss": 0.0009,
+      "step": 8221
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.658225465799544e-06,
+      "loss": 0.0012,
+      "step": 8222
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6581836431693188e-06,
+      "loss": 0.0009,
+      "step": 8223
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6581418205390936e-06,
+      "loss": 0.0033,
+      "step": 8224
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6580999979088684e-06,
+      "loss": 0.0017,
+      "step": 8225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6580581752786432e-06,
+      "loss": 0.0005,
+      "step": 8226
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.658016352648418e-06,
+      "loss": 0.0007,
+      "step": 8227
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6579745300181928e-06,
+      "loss": 1.1467,
+      "step": 8228
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6579327073879676e-06,
+      "loss": 0.4177,
+      "step": 8229
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6578908847577423e-06,
+      "loss": 0.1651,
+      "step": 8230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6578490621275171e-06,
+      "loss": 0.7051,
+      "step": 8231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.657807239497292e-06,
+      "loss": 0.0041,
+      "step": 8232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6577654168670667e-06,
+      "loss": 0.003,
+      "step": 8233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6577235942368415e-06,
+      "loss": 0.0008,
+      "step": 8234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6576817716066163e-06,
+      "loss": 0.0007,
+      "step": 8235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.657639948976391e-06,
+      "loss": 0.0052,
+      "step": 8236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6575981263461658e-06,
+      "loss": 0.0019,
+      "step": 8237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6575563037159406e-06,
+      "loss": 3.2034,
+      "step": 8238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6575144810857154e-06,
+      "loss": 0.0016,
+      "step": 8239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6574726584554902e-06,
+      "loss": 0.0011,
+      "step": 8240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6574308358252648e-06,
+      "loss": 0.0027,
+      "step": 8241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6573890131950396e-06,
+      "loss": 0.6988,
+      "step": 8242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6573471905648146e-06,
+      "loss": 0.001,
+      "step": 8243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6573053679345894e-06,
+      "loss": 1.263,
+      "step": 8244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6572635453043641e-06,
+      "loss": 0.0009,
+      "step": 8245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.657221722674139e-06,
+      "loss": 0.0007,
+      "step": 8246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6571799000439137e-06,
+      "loss": 0.0012,
+      "step": 8247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6571380774136885e-06,
+      "loss": 0.0012,
+      "step": 8248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6570962547834633e-06,
+      "loss": 0.0006,
+      "step": 8249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.657054432153238e-06,
+      "loss": 0.0046,
+      "step": 8250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6570126095230129e-06,
+      "loss": 0.0077,
+      "step": 8251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6569707868927876e-06,
+      "loss": 2.2299,
+      "step": 8252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6569289642625624e-06,
+      "loss": 0.001,
+      "step": 8253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6568871416323372e-06,
+      "loss": 0.0013,
+      "step": 8254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6568453190021118e-06,
+      "loss": 0.0006,
+      "step": 8255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6568034963718866e-06,
+      "loss": 0.0006,
+      "step": 8256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6567616737416616e-06,
+      "loss": 0.1382,
+      "step": 8257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6567198511114364e-06,
+      "loss": 0.0012,
+      "step": 8258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6566780284812111e-06,
+      "loss": 0.0023,
+      "step": 8259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656636205850986e-06,
+      "loss": 0.0015,
+      "step": 8260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6565943832207607e-06,
+      "loss": 0.0013,
+      "step": 8261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6565525605905355e-06,
+      "loss": 0.0011,
+      "step": 8262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6565107379603103e-06,
+      "loss": 0.0011,
+      "step": 8263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656468915330085e-06,
+      "loss": 0.0005,
+      "step": 8264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6564270926998599e-06,
+      "loss": 0.001,
+      "step": 8265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6563852700696347e-06,
+      "loss": 0.0504,
+      "step": 8266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6563434474394094e-06,
+      "loss": 0.0005,
+      "step": 8267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6563016248091842e-06,
+      "loss": 0.0246,
+      "step": 8268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6562598021789588e-06,
+      "loss": 0.0013,
+      "step": 8269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6562179795487336e-06,
+      "loss": 0.0015,
+      "step": 8270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6561761569185084e-06,
+      "loss": 0.0024,
+      "step": 8271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6561343342882834e-06,
+      "loss": 0.0032,
+      "step": 8272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6560925116580582e-06,
+      "loss": 0.0026,
+      "step": 8273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656050689027833e-06,
+      "loss": 0.0007,
+      "step": 8274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6560088663976077e-06,
+      "loss": 0.0013,
+      "step": 8275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6559670437673825e-06,
+      "loss": 0.0026,
+      "step": 8276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6559252211371573e-06,
+      "loss": 0.0018,
+      "step": 8277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655883398506932e-06,
+      "loss": 0.0006,
+      "step": 8278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6558415758767069e-06,
+      "loss": 0.0011,
+      "step": 8279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6557997532464817e-06,
+      "loss": 0.0031,
+      "step": 8280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6557579306162564e-06,
+      "loss": 0.0009,
+      "step": 8281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655716107986031e-06,
+      "loss": 0.0013,
+      "step": 8282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6556742853558058e-06,
+      "loss": 0.0116,
+      "step": 8283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6556324627255806e-06,
+      "loss": 0.0011,
+      "step": 8284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6555906400953554e-06,
+      "loss": 0.0016,
+      "step": 8285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6555488174651302e-06,
+      "loss": 0.0089,
+      "step": 8286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6555069948349052e-06,
+      "loss": 0.234,
+      "step": 8287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65546517220468e-06,
+      "loss": 2.2489,
+      "step": 8288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6554233495744547e-06,
+      "loss": 0.0007,
+      "step": 8289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6553815269442295e-06,
+      "loss": 0.0011,
+      "step": 8290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6553397043140043e-06,
+      "loss": 0.0022,
+      "step": 8291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655297881683779e-06,
+      "loss": 0.0005,
+      "step": 8292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6552560590535539e-06,
+      "loss": 2.6751,
+      "step": 8293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6552142364233287e-06,
+      "loss": 3.3373,
+      "step": 8294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6551724137931035e-06,
+      "loss": 0.0013,
+      "step": 8295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655130591162878e-06,
+      "loss": 0.3369,
+      "step": 8296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6550887685326528e-06,
+      "loss": 0.0008,
+      "step": 8297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6550469459024276e-06,
+      "loss": 0.0107,
+      "step": 8298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6550051232722024e-06,
+      "loss": 1.5958,
+      "step": 8299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6549633006419772e-06,
+      "loss": 0.0008,
+      "step": 8300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6549214780117522e-06,
+      "loss": 0.002,
+      "step": 8301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654879655381527e-06,
+      "loss": 0.0006,
+      "step": 8302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6548378327513017e-06,
+      "loss": 0.0018,
+      "step": 8303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6547960101210765e-06,
+      "loss": 0.4925,
+      "step": 8304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6547541874908513e-06,
+      "loss": 0.0009,
+      "step": 8305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654712364860626e-06,
+      "loss": 0.0011,
+      "step": 8306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6546705422304009e-06,
+      "loss": 0.0023,
+      "step": 8307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6546287196001757e-06,
+      "loss": 0.003,
+      "step": 8308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6545868969699503e-06,
+      "loss": 0.0218,
+      "step": 8309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654545074339725e-06,
+      "loss": 1.0344,
+      "step": 8310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6545032517094998e-06,
+      "loss": 0.0011,
+      "step": 8311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6544614290792746e-06,
+      "loss": 0.0004,
+      "step": 8312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6544196064490494e-06,
+      "loss": 2.9945,
+      "step": 8313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6543777838188242e-06,
+      "loss": 0.0036,
+      "step": 8314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654335961188599e-06,
+      "loss": 0.8649,
+      "step": 8315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654294138558374e-06,
+      "loss": 0.0008,
+      "step": 8316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6542523159281488e-06,
+      "loss": 0.8839,
+      "step": 8317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6542104932979235e-06,
+      "loss": 0.0017,
+      "step": 8318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6541686706676983e-06,
+      "loss": 1.0116,
+      "step": 8319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6541268480374731e-06,
+      "loss": 0.0006,
+      "step": 8320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654085025407248e-06,
+      "loss": 0.0016,
+      "step": 8321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6540432027770227e-06,
+      "loss": 0.1495,
+      "step": 8322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6540013801467973e-06,
+      "loss": 2.4965,
+      "step": 8323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653959557516572e-06,
+      "loss": 0.0031,
+      "step": 8324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6539177348863468e-06,
+      "loss": 1.3432,
+      "step": 8325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6538759122561216e-06,
+      "loss": 0.0131,
+      "step": 8326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6538340896258964e-06,
+      "loss": 0.0007,
+      "step": 8327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6537922669956712e-06,
+      "loss": 0.6718,
+      "step": 8328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653750444365446e-06,
+      "loss": 2.2178,
+      "step": 8329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6537086217352208e-06,
+      "loss": 0.0013,
+      "step": 8330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6536667991049958e-06,
+      "loss": 0.0006,
+      "step": 8331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6536249764747705e-06,
+      "loss": 3.4613,
+      "step": 8332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6535831538445453e-06,
+      "loss": 1.6826,
+      "step": 8333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6535413312143201e-06,
+      "loss": 0.0024,
+      "step": 8334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653499508584095e-06,
+      "loss": 2.4428,
+      "step": 8335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6534576859538695e-06,
+      "loss": 0.0009,
+      "step": 8336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6534158633236443e-06,
+      "loss": 0.002,
+      "step": 8337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653374040693419e-06,
+      "loss": 0.0139,
+      "step": 8338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6533322180631938e-06,
+      "loss": 0.005,
+      "step": 8339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6532903954329686e-06,
+      "loss": 0.0016,
+      "step": 8340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6532485728027434e-06,
+      "loss": 0.189,
+      "step": 8341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6532067501725182e-06,
+      "loss": 0.0044,
+      "step": 8342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653164927542293e-06,
+      "loss": 0.2977,
+      "step": 8343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6531231049120678e-06,
+      "loss": 0.0056,
+      "step": 8344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6530812822818428e-06,
+      "loss": 0.9235,
+      "step": 8345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6530394596516176e-06,
+      "loss": 0.0095,
+      "step": 8346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6529976370213923e-06,
+      "loss": 0.0413,
+      "step": 8347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6529558143911671e-06,
+      "loss": 0.002,
+      "step": 8348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652913991760942e-06,
+      "loss": 0.0013,
+      "step": 8349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6528721691307165e-06,
+      "loss": 0.005,
+      "step": 8350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6528303465004913e-06,
+      "loss": 0.0023,
+      "step": 8351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652788523870266e-06,
+      "loss": 0.0011,
+      "step": 8352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6527467012400408e-06,
+      "loss": 0.0099,
+      "step": 8353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6527048786098156e-06,
+      "loss": 0.281,
+      "step": 8354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6526630559795904e-06,
+      "loss": 0.0414,
+      "step": 8355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6526212333493652e-06,
+      "loss": 0.0057,
+      "step": 8356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65257941071914e-06,
+      "loss": 2.6692,
+      "step": 8357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6525375880889148e-06,
+      "loss": 0.0019,
+      "step": 8358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6524957654586896e-06,
+      "loss": 1.2108,
+      "step": 8359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6524539428284646e-06,
+      "loss": 0.0009,
+      "step": 8360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6524121201982394e-06,
+      "loss": 0.2564,
+      "step": 8361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6523702975680141e-06,
+      "loss": 0.2743,
+      "step": 8362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652328474937789e-06,
+      "loss": 0.0007,
+      "step": 8363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6522866523075635e-06,
+      "loss": 0.003,
+      "step": 8364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6522448296773383e-06,
+      "loss": 3.7084,
+      "step": 8365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652203007047113e-06,
+      "loss": 0.0043,
+      "step": 8366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6521611844168879e-06,
+      "loss": 0.001,
+      "step": 8367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6521193617866626e-06,
+      "loss": 0.0008,
+      "step": 8368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6520775391564374e-06,
+      "loss": 0.0037,
+      "step": 8369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6520357165262122e-06,
+      "loss": 0.0011,
+      "step": 8370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651993893895987e-06,
+      "loss": 1.1636,
+      "step": 8371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6519520712657618e-06,
+      "loss": 0.0012,
+      "step": 8372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6519102486355366e-06,
+      "loss": 0.0027,
+      "step": 8373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6518684260053114e-06,
+      "loss": 0.003,
+      "step": 8374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6518266033750864e-06,
+      "loss": 0.0013,
+      "step": 8375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6517847807448611e-06,
+      "loss": 1.3134,
+      "step": 8376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6517429581146357e-06,
+      "loss": 0.7096,
+      "step": 8377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6517011354844105e-06,
+      "loss": 0.2432,
+      "step": 8378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6516593128541853e-06,
+      "loss": 0.0273,
+      "step": 8379
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.65161749022396e-06,
+      "loss": 0.0024,
+      "step": 8380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6515756675937349e-06,
+      "loss": 0.2232,
+      "step": 8381
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6515338449635097e-06,
+      "loss": 0.0021,
+      "step": 8382
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6514920223332844e-06,
+      "loss": 0.0022,
+      "step": 8383
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6514501997030592e-06,
+      "loss": 0.2702,
+      "step": 8384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.651408377072834e-06,
+      "loss": 0.0018,
+      "step": 8385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6513665544426088e-06,
+      "loss": 0.5942,
+      "step": 8386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6513247318123836e-06,
+      "loss": 0.0178,
+      "step": 8387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6512829091821584e-06,
+      "loss": 0.0021,
+      "step": 8388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6512410865519334e-06,
+      "loss": 0.0254,
+      "step": 8389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6511992639217082e-06,
+      "loss": 0.005,
+      "step": 8390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6511574412914827e-06,
+      "loss": 2.2393,
+      "step": 8391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6511156186612575e-06,
+      "loss": 2.3107,
+      "step": 8392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6510737960310323e-06,
+      "loss": 0.6159,
+      "step": 8393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.651031973400807e-06,
+      "loss": 0.0262,
+      "step": 8394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6509901507705819e-06,
+      "loss": 0.0016,
+      "step": 8395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6509483281403567e-06,
+      "loss": 0.0014,
+      "step": 8396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6509065055101314e-06,
+      "loss": 0.0221,
+      "step": 8397
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6508646828799062e-06,
+      "loss": 1.7227,
+      "step": 8398
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.650822860249681e-06,
+      "loss": 0.0036,
+      "step": 8399
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6507810376194558e-06,
+      "loss": 0.7044,
+      "step": 8400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6507392149892306e-06,
+      "loss": 0.0023,
+      "step": 8401
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6506973923590054e-06,
+      "loss": 0.6553,
+      "step": 8402
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6506555697287802e-06,
+      "loss": 0.0132,
+      "step": 8403
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.650613747098555e-06,
+      "loss": 0.009,
+      "step": 8404
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6505719244683297e-06,
+      "loss": 0.0017,
+      "step": 8405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6505301018381045e-06,
+      "loss": 0.0018,
+      "step": 8406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6504882792078793e-06,
+      "loss": 0.0024,
+      "step": 8407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.650446456577654e-06,
+      "loss": 0.0028,
+      "step": 8408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6504046339474289e-06,
+      "loss": 0.0017,
+      "step": 8409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6503628113172037e-06,
+      "loss": 0.0299,
+      "step": 8410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6503209886869785e-06,
+      "loss": 0.0024,
+      "step": 8411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6502791660567532e-06,
+      "loss": 0.0053,
+      "step": 8412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.650237343426528e-06,
+      "loss": 0.0123,
+      "step": 8413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6501955207963028e-06,
+      "loss": 0.0006,
+      "step": 8414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6501536981660776e-06,
+      "loss": 0.001,
+      "step": 8415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6501118755358524e-06,
+      "loss": 0.1762,
+      "step": 8416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.650070052905627e-06,
+      "loss": 0.008,
+      "step": 8417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.650028230275402e-06,
+      "loss": 0.5775,
+      "step": 8418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6499864076451767e-06,
+      "loss": 0.0093,
+      "step": 8419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6499445850149515e-06,
+      "loss": 2.2516,
+      "step": 8420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6499027623847263e-06,
+      "loss": 0.0011,
+      "step": 8421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649860939754501e-06,
+      "loss": 0.0032,
+      "step": 8422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6498191171242759e-06,
+      "loss": 0.0066,
+      "step": 8423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6497772944940507e-06,
+      "loss": 0.194,
+      "step": 8424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6497354718638255e-06,
+      "loss": 1.1657,
+      "step": 8425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6496936492336003e-06,
+      "loss": 0.0006,
+      "step": 8426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649651826603375e-06,
+      "loss": 0.007,
+      "step": 8427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6496100039731498e-06,
+      "loss": 0.001,
+      "step": 8428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6495681813429246e-06,
+      "loss": 0.0013,
+      "step": 8429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6495263587126994e-06,
+      "loss": 0.0007,
+      "step": 8430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649484536082474e-06,
+      "loss": 0.0004,
+      "step": 8431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6494427134522488e-06,
+      "loss": 1.6462,
+      "step": 8432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6494008908220238e-06,
+      "loss": 0.0018,
+      "step": 8433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6493590681917985e-06,
+      "loss": 0.0043,
+      "step": 8434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6493172455615733e-06,
+      "loss": 0.0012,
+      "step": 8435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6492754229313481e-06,
+      "loss": 0.0021,
+      "step": 8436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649233600301123e-06,
+      "loss": 0.0015,
+      "step": 8437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6491917776708977e-06,
+      "loss": 0.0031,
+      "step": 8438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6491499550406725e-06,
+      "loss": 0.0006,
+      "step": 8439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6491081324104473e-06,
+      "loss": 0.3973,
+      "step": 8440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649066309780222e-06,
+      "loss": 0.0013,
+      "step": 8441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6490244871499968e-06,
+      "loss": 0.0036,
+      "step": 8442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6489826645197716e-06,
+      "loss": 0.0009,
+      "step": 8443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6489408418895462e-06,
+      "loss": 0.0035,
+      "step": 8444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648899019259321e-06,
+      "loss": 0.0015,
+      "step": 8445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6488571966290958e-06,
+      "loss": 0.0011,
+      "step": 8446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6488153739988706e-06,
+      "loss": 0.0024,
+      "step": 8447
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6487735513686456e-06,
+      "loss": 0.0007,
+      "step": 8448
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6487317287384203e-06,
+      "loss": 0.0014,
+      "step": 8449
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6486899061081951e-06,
+      "loss": 0.0012,
+      "step": 8450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64864808347797e-06,
+      "loss": 0.0009,
+      "step": 8451
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6486062608477447e-06,
+      "loss": 0.0267,
+      "step": 8452
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6485644382175195e-06,
+      "loss": 0.3018,
+      "step": 8453
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6485226155872943e-06,
+      "loss": 1.8466,
+      "step": 8454
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648480792957069e-06,
+      "loss": 0.0019,
+      "step": 8455
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6484389703268438e-06,
+      "loss": 0.0006,
+      "step": 8456
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6483971476966186e-06,
+      "loss": 0.0016,
+      "step": 8457
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6483553250663932e-06,
+      "loss": 0.0012,
+      "step": 8458
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648313502436168e-06,
+      "loss": 0.0008,
+      "step": 8459
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6482716798059428e-06,
+      "loss": 2.555,
+      "step": 8460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6482298571757176e-06,
+      "loss": 0.0006,
+      "step": 8461
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6481880345454926e-06,
+      "loss": 1.1552,
+      "step": 8462
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6481462119152673e-06,
+      "loss": 0.0013,
+      "step": 8463
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6481043892850421e-06,
+      "loss": 0.0234,
+      "step": 8464
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648062566654817e-06,
+      "loss": 0.0008,
+      "step": 8465
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6480207440245917e-06,
+      "loss": 0.0032,
+      "step": 8466
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6479789213943665e-06,
+      "loss": 0.2818,
+      "step": 8467
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6479370987641413e-06,
+      "loss": 0.0013,
+      "step": 8468
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647895276133916e-06,
+      "loss": 0.0021,
+      "step": 8469
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6478534535036908e-06,
+      "loss": 0.009,
+      "step": 8470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6478116308734656e-06,
+      "loss": 0.3845,
+      "step": 8471
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6477698082432402e-06,
+      "loss": 0.1808,
+      "step": 8472
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647727985613015e-06,
+      "loss": 0.001,
+      "step": 8473
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6476861629827898e-06,
+      "loss": 0.0042,
+      "step": 8474
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6476443403525646e-06,
+      "loss": 1.0508,
+      "step": 8475
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6476025177223394e-06,
+      "loss": 0.0017,
+      "step": 8476
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6475606950921144e-06,
+      "loss": 1.4964,
+      "step": 8477
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6475188724618891e-06,
+      "loss": 0.001,
+      "step": 8478
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647477049831664e-06,
+      "loss": 0.0034,
+      "step": 8479
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6474352272014387e-06,
+      "loss": 0.0005,
+      "step": 8480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6473934045712135e-06,
+      "loss": 0.0008,
+      "step": 8481
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6473515819409883e-06,
+      "loss": 2.9633,
+      "step": 8482
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647309759310763e-06,
+      "loss": 0.0007,
+      "step": 8483
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6472679366805379e-06,
+      "loss": 0.0005,
+      "step": 8484
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6472261140503124e-06,
+      "loss": 0.0024,
+      "step": 8485
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6471842914200872e-06,
+      "loss": 0.0012,
+      "step": 8486
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647142468789862e-06,
+      "loss": 0.0012,
+      "step": 8487
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6471006461596368e-06,
+      "loss": 0.0009,
+      "step": 8488
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6470588235294116e-06,
+      "loss": 0.007,
+      "step": 8489
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6470170008991864e-06,
+      "loss": 0.001,
+      "step": 8490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6469751782689612e-06,
+      "loss": 0.1198,
+      "step": 8491
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6469333556387361e-06,
+      "loss": 0.0006,
+      "step": 8492
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646891533008511e-06,
+      "loss": 0.0006,
+      "step": 8493
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6468497103782857e-06,
+      "loss": 0.0004,
+      "step": 8494
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6468078877480605e-06,
+      "loss": 0.0015,
+      "step": 8495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6467660651178353e-06,
+      "loss": 0.0012,
+      "step": 8496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64672424248761e-06,
+      "loss": 0.0012,
+      "step": 8497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6466824198573849e-06,
+      "loss": 0.0022,
+      "step": 8498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6466405972271594e-06,
+      "loss": 0.001,
+      "step": 8499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6465987745969342e-06,
+      "loss": 0.0014,
+      "step": 8500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646556951966709e-06,
+      "loss": 0.0014,
+      "step": 8501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6465151293364838e-06,
+      "loss": 0.0008,
+      "step": 8502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6464733067062586e-06,
+      "loss": 0.0014,
+      "step": 8503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6464314840760334e-06,
+      "loss": 0.4806,
+      "step": 8504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6463896614458082e-06,
+      "loss": 0.0384,
+      "step": 8505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6463478388155832e-06,
+      "loss": 0.0044,
+      "step": 8506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646306016185358e-06,
+      "loss": 0.483,
+      "step": 8507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6462641935551327e-06,
+      "loss": 0.0015,
+      "step": 8508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6462223709249075e-06,
+      "loss": 0.0012,
+      "step": 8509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6461805482946823e-06,
+      "loss": 0.0022,
+      "step": 8510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646138725664457e-06,
+      "loss": 0.0079,
+      "step": 8511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6460969030342317e-06,
+      "loss": 0.191,
+      "step": 8512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6460550804040064e-06,
+      "loss": 0.001,
+      "step": 8513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6460132577737812e-06,
+      "loss": 0.6257,
+      "step": 8514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645971435143556e-06,
+      "loss": 0.0053,
+      "step": 8515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6459296125133308e-06,
+      "loss": 0.0071,
+      "step": 8516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6458877898831056e-06,
+      "loss": 0.0008,
+      "step": 8517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6458459672528804e-06,
+      "loss": 0.0011,
+      "step": 8518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6458041446226552e-06,
+      "loss": 0.0004,
+      "step": 8519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64576232199243e-06,
+      "loss": 0.0016,
+      "step": 8520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645720499362205e-06,
+      "loss": 0.0006,
+      "step": 8521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6456786767319797e-06,
+      "loss": 0.0013,
+      "step": 8522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6456368541017545e-06,
+      "loss": 0.765,
+      "step": 8523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6455950314715293e-06,
+      "loss": 0.0007,
+      "step": 8524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645553208841304e-06,
+      "loss": 0.0008,
+      "step": 8525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6455113862110787e-06,
+      "loss": 0.3101,
+      "step": 8526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6454695635808535e-06,
+      "loss": 0.001,
+      "step": 8527
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6454277409506282e-06,
+      "loss": 0.0005,
+      "step": 8528
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645385918320403e-06,
+      "loss": 0.0012,
+      "step": 8529
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6453440956901778e-06,
+      "loss": 0.0008,
+      "step": 8530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6453022730599526e-06,
+      "loss": 0.0013,
+      "step": 8531
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6452604504297274e-06,
+      "loss": 0.0005,
+      "step": 8532
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6452186277995022e-06,
+      "loss": 0.0009,
+      "step": 8533
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645176805169277e-06,
+      "loss": 3.1907,
+      "step": 8534
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6451349825390517e-06,
+      "loss": 0.0015,
+      "step": 8535
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6450931599088267e-06,
+      "loss": 0.0005,
+      "step": 8536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6450513372786015e-06,
+      "loss": 0.0016,
+      "step": 8537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6450095146483763e-06,
+      "loss": 3.0026,
+      "step": 8538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.644967692018151e-06,
+      "loss": 0.0005,
+      "step": 8539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6449258693879257e-06,
+      "loss": 0.1755,
+      "step": 8540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6448840467577005e-06,
+      "loss": 0.2982,
+      "step": 8541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6448422241274753e-06,
+      "loss": 0.0007,
+      "step": 8542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.64480040149725e-06,
+      "loss": 0.0027,
+      "step": 8543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6447585788670248e-06,
+      "loss": 0.001,
+      "step": 8544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6447167562367996e-06,
+      "loss": 0.0943,
+      "step": 8545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6446749336065744e-06,
+      "loss": 0.0021,
+      "step": 8546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6446331109763492e-06,
+      "loss": 0.0005,
+      "step": 8547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.644591288346124e-06,
+      "loss": 0.0011,
+      "step": 8548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6445494657158988e-06,
+      "loss": 1.6088,
+      "step": 8549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6445076430856738e-06,
+      "loss": 1.645,
+      "step": 8550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6444658204554485e-06,
+      "loss": 0.0011,
+      "step": 8551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6444239978252233e-06,
+      "loss": 0.0064,
+      "step": 8552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.644382175194998e-06,
+      "loss": 3.309,
+      "step": 8553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6443403525647727e-06,
+      "loss": 2.2792,
+      "step": 8554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6442985299345475e-06,
+      "loss": 0.0022,
+      "step": 8555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6442567073043223e-06,
+      "loss": 0.001,
+      "step": 8556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.644214884674097e-06,
+      "loss": 2.087,
+      "step": 8557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6441730620438718e-06,
+      "loss": 0.0119,
+      "step": 8558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6441312394136466e-06,
+      "loss": 2.3063,
+      "step": 8559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6440894167834214e-06,
+      "loss": 0.0071,
+      "step": 8560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6440475941531962e-06,
+      "loss": 0.0027,
+      "step": 8561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.644005771522971e-06,
+      "loss": 0.0379,
+      "step": 8562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6439639488927458e-06,
+      "loss": 1.1471,
+      "step": 8563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6439221262625206e-06,
+      "loss": 0.118,
+      "step": 8564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6438803036322956e-06,
+      "loss": 0.0029,
+      "step": 8565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6438384810020703e-06,
+      "loss": 0.0037,
+      "step": 8566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.643796658371845e-06,
+      "loss": 0.2045,
+      "step": 8567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6437548357416197e-06,
+      "loss": 0.0009,
+      "step": 8568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6437130131113945e-06,
+      "loss": 0.0208,
+      "step": 8569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6436711904811693e-06,
+      "loss": 0.0013,
+      "step": 8570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.643629367850944e-06,
+      "loss": 0.0019,
+      "step": 8571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6435875452207188e-06,
+      "loss": 0.0011,
+      "step": 8572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6435457225904936e-06,
+      "loss": 0.0013,
+      "step": 8573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6435038999602684e-06,
+      "loss": 0.0479,
+      "step": 8574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6434620773300432e-06,
+      "loss": 0.0014,
+      "step": 8575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.643420254699818e-06,
+      "loss": 0.0082,
+      "step": 8576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6433784320695928e-06,
+      "loss": 1.4274,
+      "step": 8577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6433366094393676e-06,
+      "loss": 0.0066,
+      "step": 8578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6432947868091421e-06,
+      "loss": 0.0343,
+      "step": 8579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6432529641789171e-06,
+      "loss": 0.0033,
+      "step": 8580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.643211141548692e-06,
+      "loss": 1.266,
+      "step": 8581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6431693189184667e-06,
+      "loss": 0.0007,
+      "step": 8582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6431274962882415e-06,
+      "loss": 0.0007,
+      "step": 8583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6430856736580163e-06,
+      "loss": 0.1675,
+      "step": 8584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.643043851027791e-06,
+      "loss": 0.0006,
+      "step": 8585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6430020283975659e-06,
+      "loss": 0.0025,
+      "step": 8586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6429602057673406e-06,
+      "loss": 0.0012,
+      "step": 8587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6429183831371154e-06,
+      "loss": 0.4015,
+      "step": 8588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6428765605068902e-06,
+      "loss": 0.0621,
+      "step": 8589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642834737876665e-06,
+      "loss": 0.0012,
+      "step": 8590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6427929152464398e-06,
+      "loss": 0.0015,
+      "step": 8591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6427510926162146e-06,
+      "loss": 0.001,
+      "step": 8592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6427092699859891e-06,
+      "loss": 0.6799,
+      "step": 8593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6426674473557641e-06,
+      "loss": 0.0005,
+      "step": 8594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642625624725539e-06,
+      "loss": 0.0023,
+      "step": 8595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6425838020953137e-06,
+      "loss": 0.1088,
+      "step": 8596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6425419794650885e-06,
+      "loss": 0.0011,
+      "step": 8597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6425001568348633e-06,
+      "loss": 0.7935,
+      "step": 8598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642458334204638e-06,
+      "loss": 0.0028,
+      "step": 8599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6424165115744129e-06,
+      "loss": 0.0006,
+      "step": 8600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6423746889441876e-06,
+      "loss": 0.0022,
+      "step": 8601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6423328663139624e-06,
+      "loss": 0.001,
+      "step": 8602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6422910436837372e-06,
+      "loss": 0.003,
+      "step": 8603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642249221053512e-06,
+      "loss": 0.0034,
+      "step": 8604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6422073984232868e-06,
+      "loss": 0.0055,
+      "step": 8605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6421655757930616e-06,
+      "loss": 1.7974,
+      "step": 8606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6421237531628362e-06,
+      "loss": 0.0005,
+      "step": 8607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642081930532611e-06,
+      "loss": 0.0033,
+      "step": 8608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642040107902386e-06,
+      "loss": 0.0005,
+      "step": 8609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6419982852721607e-06,
+      "loss": 0.0041,
+      "step": 8610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6419564626419355e-06,
+      "loss": 2.4893,
+      "step": 8611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6419146400117103e-06,
+      "loss": 0.0005,
+      "step": 8612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641872817381485e-06,
+      "loss": 0.0009,
+      "step": 8613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6418309947512599e-06,
+      "loss": 0.0027,
+      "step": 8614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6417891721210347e-06,
+      "loss": 0.3799,
+      "step": 8615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6417473494908094e-06,
+      "loss": 0.0008,
+      "step": 8616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6417055268605842e-06,
+      "loss": 0.0005,
+      "step": 8617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641663704230359e-06,
+      "loss": 0.0024,
+      "step": 8618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6416218816001338e-06,
+      "loss": 0.0074,
+      "step": 8619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6415800589699084e-06,
+      "loss": 3.9928,
+      "step": 8620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6415382363396832e-06,
+      "loss": 0.0112,
+      "step": 8621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641496413709458e-06,
+      "loss": 0.0007,
+      "step": 8622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641454591079233e-06,
+      "loss": 0.1696,
+      "step": 8623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6414127684490077e-06,
+      "loss": 0.2538,
+      "step": 8624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6413709458187825e-06,
+      "loss": 0.2493,
+      "step": 8625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6413291231885573e-06,
+      "loss": 0.0099,
+      "step": 8626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641287300558332e-06,
+      "loss": 0.0067,
+      "step": 8627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6412454779281069e-06,
+      "loss": 0.4804,
+      "step": 8628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6412036552978817e-06,
+      "loss": 0.0014,
+      "step": 8629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6411618326676565e-06,
+      "loss": 0.0012,
+      "step": 8630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6411200100374312e-06,
+      "loss": 0.3109,
+      "step": 8631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641078187407206e-06,
+      "loss": 0.0015,
+      "step": 8632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6410363647769808e-06,
+      "loss": 0.0038,
+      "step": 8633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6409945421467554e-06,
+      "loss": 0.0017,
+      "step": 8634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6409527195165302e-06,
+      "loss": 0.0005,
+      "step": 8635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640910896886305e-06,
+      "loss": 1.6549,
+      "step": 8636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6408690742560797e-06,
+      "loss": 2.1053,
+      "step": 8637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6408272516258547e-06,
+      "loss": 0.0008,
+      "step": 8638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6407854289956295e-06,
+      "loss": 0.0007,
+      "step": 8639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6407436063654043e-06,
+      "loss": 0.4876,
+      "step": 8640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640701783735179e-06,
+      "loss": 0.0036,
+      "step": 8641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6406599611049539e-06,
+      "loss": 0.0008,
+      "step": 8642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6406181384747287e-06,
+      "loss": 0.072,
+      "step": 8643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6405763158445035e-06,
+      "loss": 0.0015,
+      "step": 8644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6405344932142782e-06,
+      "loss": 0.1925,
+      "step": 8645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640492670584053e-06,
+      "loss": 0.0031,
+      "step": 8646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6404508479538276e-06,
+      "loss": 0.0009,
+      "step": 8647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6404090253236024e-06,
+      "loss": 0.0046,
+      "step": 8648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6403672026933772e-06,
+      "loss": 0.0009,
+      "step": 8649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640325380063152e-06,
+      "loss": 0.7585,
+      "step": 8650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6402835574329268e-06,
+      "loss": 0.0013,
+      "step": 8651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6402417348027015e-06,
+      "loss": 0.0014,
+      "step": 8652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6401999121724765e-06,
+      "loss": 0.0007,
+      "step": 8653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6401580895422513e-06,
+      "loss": 0.003,
+      "step": 8654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6401162669120261e-06,
+      "loss": 0.0069,
+      "step": 8655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640074444281801e-06,
+      "loss": 0.0008,
+      "step": 8656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6400326216515757e-06,
+      "loss": 1.3168,
+      "step": 8657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6399907990213505e-06,
+      "loss": 0.0044,
+      "step": 8658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6399489763911253e-06,
+      "loss": 0.0028,
+      "step": 8659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6399071537609e-06,
+      "loss": 0.0007,
+      "step": 8660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6398653311306746e-06,
+      "loss": 0.0009,
+      "step": 8661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6398235085004494e-06,
+      "loss": 2.3685,
+      "step": 8662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6397816858702242e-06,
+      "loss": 0.0021,
+      "step": 8663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639739863239999e-06,
+      "loss": 0.0013,
+      "step": 8664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6396980406097738e-06,
+      "loss": 0.0098,
+      "step": 8665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6396562179795485e-06,
+      "loss": 0.0006,
+      "step": 8666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6396143953493235e-06,
+      "loss": 0.0008,
+      "step": 8667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6395725727190983e-06,
+      "loss": 0.007,
+      "step": 8668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6395307500888731e-06,
+      "loss": 0.0007,
+      "step": 8669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639488927458648e-06,
+      "loss": 0.0007,
+      "step": 8670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6394471048284227e-06,
+      "loss": 0.019,
+      "step": 8671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6394052821981975e-06,
+      "loss": 0.0068,
+      "step": 8672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6393634595679723e-06,
+      "loss": 2.076,
+      "step": 8673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639321636937747e-06,
+      "loss": 0.0059,
+      "step": 8674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6392798143075216e-06,
+      "loss": 0.0011,
+      "step": 8675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6392379916772964e-06,
+      "loss": 0.001,
+      "step": 8676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6391961690470712e-06,
+      "loss": 0.0042,
+      "step": 8677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639154346416846e-06,
+      "loss": 0.8411,
+      "step": 8678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6391125237866208e-06,
+      "loss": 0.1264,
+      "step": 8679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6390707011563956e-06,
+      "loss": 0.0043,
+      "step": 8680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6390288785261703e-06,
+      "loss": 0.0007,
+      "step": 8681
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6389870558959453e-06,
+      "loss": 0.0005,
+      "step": 8682
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6389452332657201e-06,
+      "loss": 0.0263,
+      "step": 8683
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638903410635495e-06,
+      "loss": 0.0527,
+      "step": 8684
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6388615880052697e-06,
+      "loss": 0.0005,
+      "step": 8685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6388197653750445e-06,
+      "loss": 0.0006,
+      "step": 8686
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6387779427448193e-06,
+      "loss": 2.3825,
+      "step": 8687
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6387361201145938e-06,
+      "loss": 0.0024,
+      "step": 8688
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6386942974843686e-06,
+      "loss": 0.6587,
+      "step": 8689
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6386524748541434e-06,
+      "loss": 0.0032,
+      "step": 8690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6386106522239182e-06,
+      "loss": 0.0573,
+      "step": 8691
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638568829593693e-06,
+      "loss": 0.0009,
+      "step": 8692
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6385270069634678e-06,
+      "loss": 0.0022,
+      "step": 8693
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6384851843332426e-06,
+      "loss": 0.0016,
+      "step": 8694
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6384433617030174e-06,
+      "loss": 0.0194,
+      "step": 8695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6384015390727921e-06,
+      "loss": 0.2163,
+      "step": 8696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6383597164425671e-06,
+      "loss": 3.2387,
+      "step": 8697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.638317893812342e-06,
+      "loss": 0.0006,
+      "step": 8698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6382760711821167e-06,
+      "loss": 0.0011,
+      "step": 8699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6382342485518915e-06,
+      "loss": 0.0007,
+      "step": 8700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6381924259216663e-06,
+      "loss": 0.0011,
+      "step": 8701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6381506032914409e-06,
+      "loss": 0.0065,
+      "step": 8702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6381087806612156e-06,
+      "loss": 0.0035,
+      "step": 8703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6380669580309904e-06,
+      "loss": 0.0009,
+      "step": 8704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6380251354007652e-06,
+      "loss": 0.4656,
+      "step": 8705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63798331277054e-06,
+      "loss": 2.5285,
+      "step": 8706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6379414901403148e-06,
+      "loss": 0.5605,
+      "step": 8707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6378996675100896e-06,
+      "loss": 0.0007,
+      "step": 8708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6378578448798644e-06,
+      "loss": 0.001,
+      "step": 8709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6378160222496391e-06,
+      "loss": 0.0016,
+      "step": 8710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6377741996194141e-06,
+      "loss": 0.001,
+      "step": 8711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.637732376989189e-06,
+      "loss": 0.0095,
+      "step": 8712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6376905543589637e-06,
+      "loss": 0.0045,
+      "step": 8713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6376487317287385e-06,
+      "loss": 0.0029,
+      "step": 8714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.637606909098513e-06,
+      "loss": 0.0008,
+      "step": 8715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6375650864682879e-06,
+      "loss": 0.0009,
+      "step": 8716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6375232638380626e-06,
+      "loss": 0.1189,
+      "step": 8717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6374814412078374e-06,
+      "loss": 0.0065,
+      "step": 8718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6374396185776122e-06,
+      "loss": 0.0788,
+      "step": 8719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.637397795947387e-06,
+      "loss": 2.2311,
+      "step": 8720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6373559733171618e-06,
+      "loss": 0.0798,
+      "step": 8721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6373141506869366e-06,
+      "loss": 0.0006,
+      "step": 8722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6372723280567114e-06,
+      "loss": 0.0007,
+      "step": 8723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6372305054264862e-06,
+      "loss": 0.0284,
+      "step": 8724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.637188682796261e-06,
+      "loss": 0.8487,
+      "step": 8725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.637146860166036e-06,
+      "loss": 0.0026,
+      "step": 8726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6371050375358107e-06,
+      "loss": 0.02,
+      "step": 8727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6370632149055855e-06,
+      "loss": 0.0006,
+      "step": 8728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63702139227536e-06,
+      "loss": 0.2701,
+      "step": 8729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6369795696451349e-06,
+      "loss": 2.2172,
+      "step": 8730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6369377470149097e-06,
+      "loss": 0.0004,
+      "step": 8731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6368959243846844e-06,
+      "loss": 0.0212,
+      "step": 8732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6368541017544592e-06,
+      "loss": 0.003,
+      "step": 8733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636812279124234e-06,
+      "loss": 0.0023,
+      "step": 8734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6367704564940088e-06,
+      "loss": 0.0021,
+      "step": 8735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6367286338637836e-06,
+      "loss": 0.0024,
+      "step": 8736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6366868112335584e-06,
+      "loss": 0.001,
+      "step": 8737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6366449886033332e-06,
+      "loss": 0.4213,
+      "step": 8738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636603165973108e-06,
+      "loss": 0.0006,
+      "step": 8739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6365613433428827e-06,
+      "loss": 0.0026,
+      "step": 8740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6365195207126577e-06,
+      "loss": 0.0021,
+      "step": 8741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6364776980824323e-06,
+      "loss": 0.0012,
+      "step": 8742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636435875452207e-06,
+      "loss": 0.0023,
+      "step": 8743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6363940528219819e-06,
+      "loss": 0.0009,
+      "step": 8744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6363522301917567e-06,
+      "loss": 0.0029,
+      "step": 8745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6363104075615315e-06,
+      "loss": 1.6003,
+      "step": 8746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6362685849313062e-06,
+      "loss": 0.0012,
+      "step": 8747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636226762301081e-06,
+      "loss": 0.0016,
+      "step": 8748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6361849396708558e-06,
+      "loss": 0.0006,
+      "step": 8749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6361431170406306e-06,
+      "loss": 0.3876,
+      "step": 8750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6361012944104054e-06,
+      "loss": 0.0008,
+      "step": 8751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6360594717801802e-06,
+      "loss": 0.1062,
+      "step": 8752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636017649149955e-06,
+      "loss": 0.0007,
+      "step": 8753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6359758265197297e-06,
+      "loss": 0.759,
+      "step": 8754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6359340038895047e-06,
+      "loss": 0.0014,
+      "step": 8755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6358921812592793e-06,
+      "loss": 0.4988,
+      "step": 8756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635850358629054e-06,
+      "loss": 0.7827,
+      "step": 8757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6358085359988289e-06,
+      "loss": 0.0006,
+      "step": 8758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6357667133686037e-06,
+      "loss": 0.001,
+      "step": 8759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6357248907383785e-06,
+      "loss": 0.0005,
+      "step": 8760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6356830681081532e-06,
+      "loss": 0.5429,
+      "step": 8761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635641245477928e-06,
+      "loss": 0.8572,
+      "step": 8762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6355994228477028e-06,
+      "loss": 0.0016,
+      "step": 8763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6355576002174776e-06,
+      "loss": 0.0042,
+      "step": 8764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6355157775872524e-06,
+      "loss": 1.6085,
+      "step": 8765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6354739549570272e-06,
+      "loss": 0.0892,
+      "step": 8766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635432132326802e-06,
+      "loss": 0.4673,
+      "step": 8767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6353903096965768e-06,
+      "loss": 3.3937,
+      "step": 8768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6353484870663513e-06,
+      "loss": 0.0006,
+      "step": 8769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6353066644361263e-06,
+      "loss": 0.0013,
+      "step": 8770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6352648418059011e-06,
+      "loss": 0.0377,
+      "step": 8771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635223019175676e-06,
+      "loss": 0.0012,
+      "step": 8772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6351811965454507e-06,
+      "loss": 1.5451,
+      "step": 8773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6351393739152255e-06,
+      "loss": 0.0041,
+      "step": 8774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6350975512850003e-06,
+      "loss": 0.0016,
+      "step": 8775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635055728654775e-06,
+      "loss": 2.7553,
+      "step": 8776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6350139060245498e-06,
+      "loss": 1.3553,
+      "step": 8777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6349720833943246e-06,
+      "loss": 2.3258,
+      "step": 8778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6349302607640994e-06,
+      "loss": 0.0034,
+      "step": 8779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6348884381338742e-06,
+      "loss": 0.0007,
+      "step": 8780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634846615503649e-06,
+      "loss": 0.0012,
+      "step": 8781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6348047928734235e-06,
+      "loss": 0.0017,
+      "step": 8782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6347629702431983e-06,
+      "loss": 0.0017,
+      "step": 8783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6347211476129733e-06,
+      "loss": 0.0008,
+      "step": 8784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6346793249827481e-06,
+      "loss": 0.0788,
+      "step": 8785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634637502352523e-06,
+      "loss": 0.0014,
+      "step": 8786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6345956797222977e-06,
+      "loss": 0.0005,
+      "step": 8787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6345538570920725e-06,
+      "loss": 0.002,
+      "step": 8788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6345120344618473e-06,
+      "loss": 0.0008,
+      "step": 8789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634470211831622e-06,
+      "loss": 0.0118,
+      "step": 8790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6344283892013968e-06,
+      "loss": 2.6048,
+      "step": 8791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6343865665711716e-06,
+      "loss": 0.0015,
+      "step": 8792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6343447439409464e-06,
+      "loss": 0.001,
+      "step": 8793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6343029213107212e-06,
+      "loss": 0.3834,
+      "step": 8794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634261098680496e-06,
+      "loss": 0.0022,
+      "step": 8795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6342192760502706e-06,
+      "loss": 3.3173,
+      "step": 8796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6341774534200453e-06,
+      "loss": 0.0026,
+      "step": 8797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6341356307898201e-06,
+      "loss": 0.0014,
+      "step": 8798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6340938081595951e-06,
+      "loss": 0.3139,
+      "step": 8799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63405198552937e-06,
+      "loss": 2.2847,
+      "step": 8800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6340101628991447e-06,
+      "loss": 0.8982,
+      "step": 8801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6339683402689195e-06,
+      "loss": 0.0005,
+      "step": 8802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6339265176386943e-06,
+      "loss": 0.0039,
+      "step": 8803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633884695008469e-06,
+      "loss": 2.5813,
+      "step": 8804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6338428723782438e-06,
+      "loss": 0.0008,
+      "step": 8805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6338010497480186e-06,
+      "loss": 0.0007,
+      "step": 8806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6337592271177934e-06,
+      "loss": 2.8296,
+      "step": 8807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6337174044875682e-06,
+      "loss": 0.0024,
+      "step": 8808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633675581857343e-06,
+      "loss": 0.0061,
+      "step": 8809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6336337592271176e-06,
+      "loss": 0.0044,
+      "step": 8810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6335919365968924e-06,
+      "loss": 0.0015,
+      "step": 8811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6335501139666671e-06,
+      "loss": 0.0009,
+      "step": 8812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633508291336442e-06,
+      "loss": 0.0013,
+      "step": 8813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633466468706217e-06,
+      "loss": 1.3243,
+      "step": 8814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6334246460759917e-06,
+      "loss": 0.0457,
+      "step": 8815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6333828234457665e-06,
+      "loss": 0.4949,
+      "step": 8816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6333410008155413e-06,
+      "loss": 0.0007,
+      "step": 8817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633299178185316e-06,
+      "loss": 0.0057,
+      "step": 8818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6332573555550909e-06,
+      "loss": 0.564,
+      "step": 8819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6332155329248656e-06,
+      "loss": 0.0748,
+      "step": 8820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6331737102946404e-06,
+      "loss": 0.0034,
+      "step": 8821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6331318876644152e-06,
+      "loss": 0.7779,
+      "step": 8822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6330900650341898e-06,
+      "loss": 0.007,
+      "step": 8823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6330482424039646e-06,
+      "loss": 0.0026,
+      "step": 8824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6330064197737394e-06,
+      "loss": 0.0055,
+      "step": 8825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6329645971435141e-06,
+      "loss": 0.0098,
+      "step": 8826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632922774513289e-06,
+      "loss": 0.0012,
+      "step": 8827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632880951883064e-06,
+      "loss": 0.0029,
+      "step": 8828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6328391292528387e-06,
+      "loss": 0.8127,
+      "step": 8829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6327973066226135e-06,
+      "loss": 0.0016,
+      "step": 8830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6327554839923883e-06,
+      "loss": 0.4656,
+      "step": 8831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632713661362163e-06,
+      "loss": 0.0009,
+      "step": 8832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6326718387319379e-06,
+      "loss": 0.0209,
+      "step": 8833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6326300161017126e-06,
+      "loss": 0.3257,
+      "step": 8834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6325881934714874e-06,
+      "loss": 0.002,
+      "step": 8835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6325463708412622e-06,
+      "loss": 0.0022,
+      "step": 8836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6325045482110368e-06,
+      "loss": 0.0038,
+      "step": 8837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6324627255808116e-06,
+      "loss": 0.2861,
+      "step": 8838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6324209029505864e-06,
+      "loss": 0.0029,
+      "step": 8839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6323790803203612e-06,
+      "loss": 0.0012,
+      "step": 8840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632337257690136e-06,
+      "loss": 0.0023,
+      "step": 8841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6322954350599107e-06,
+      "loss": 0.0051,
+      "step": 8842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6322536124296857e-06,
+      "loss": 0.0016,
+      "step": 8843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6322117897994605e-06,
+      "loss": 0.0006,
+      "step": 8844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6321699671692353e-06,
+      "loss": 0.6391,
+      "step": 8845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63212814453901e-06,
+      "loss": 0.0015,
+      "step": 8846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6320863219087849e-06,
+      "loss": 0.0006,
+      "step": 8847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6320444992785597e-06,
+      "loss": 0.0015,
+      "step": 8848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6320026766483344e-06,
+      "loss": 0.0119,
+      "step": 8849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631960854018109e-06,
+      "loss": 2.9582,
+      "step": 8850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6319190313878838e-06,
+      "loss": 0.0867,
+      "step": 8851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6318772087576586e-06,
+      "loss": 0.0015,
+      "step": 8852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6318353861274334e-06,
+      "loss": 0.0007,
+      "step": 8853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6317935634972082e-06,
+      "loss": 0.0026,
+      "step": 8854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631751740866983e-06,
+      "loss": 0.0068,
+      "step": 8855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6317099182367577e-06,
+      "loss": 0.0017,
+      "step": 8856
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6316680956065325e-06,
+      "loss": 0.0014,
+      "step": 8857
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6316262729763075e-06,
+      "loss": 0.003,
+      "step": 8858
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6315844503460823e-06,
+      "loss": 0.0013,
+      "step": 8859
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.631542627715857e-06,
+      "loss": 0.0008,
+      "step": 8860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6315008050856319e-06,
+      "loss": 3.4782,
+      "step": 8861
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6314589824554067e-06,
+      "loss": 0.0008,
+      "step": 8862
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6314171598251815e-06,
+      "loss": 0.0151,
+      "step": 8863
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.631375337194956e-06,
+      "loss": 1.1712,
+      "step": 8864
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6313335145647308e-06,
+      "loss": 1.0984,
+      "step": 8865
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6312916919345056e-06,
+      "loss": 0.7689,
+      "step": 8866
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6312498693042804e-06,
+      "loss": 0.0006,
+      "step": 8867
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6312080466740552e-06,
+      "loss": 0.012,
+      "step": 8868
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.63116622404383e-06,
+      "loss": 0.2198,
+      "step": 8869
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6311244014136047e-06,
+      "loss": 0.0397,
+      "step": 8870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6310825787833795e-06,
+      "loss": 0.0069,
+      "step": 8871
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6310407561531545e-06,
+      "loss": 0.0081,
+      "step": 8872
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6309989335229293e-06,
+      "loss": 0.0026,
+      "step": 8873
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.630957110892704e-06,
+      "loss": 0.0023,
+      "step": 8874
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6309152882624789e-06,
+      "loss": 0.0014,
+      "step": 8875
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6308734656322537e-06,
+      "loss": 0.0033,
+      "step": 8876
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6308316430020285e-06,
+      "loss": 0.2673,
+      "step": 8877
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.630789820371803e-06,
+      "loss": 0.0006,
+      "step": 8878
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6307479977415778e-06,
+      "loss": 0.0044,
+      "step": 8879
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6307061751113526e-06,
+      "loss": 0.0019,
+      "step": 8880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6306643524811274e-06,
+      "loss": 0.0006,
+      "step": 8881
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6306225298509022e-06,
+      "loss": 1.7602,
+      "step": 8882
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.630580707220677e-06,
+      "loss": 0.0056,
+      "step": 8883
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6305388845904518e-06,
+      "loss": 0.0013,
+      "step": 8884
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6304970619602265e-06,
+      "loss": 2.3638,
+      "step": 8885
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6304552393300013e-06,
+      "loss": 0.0034,
+      "step": 8886
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6304134166997763e-06,
+      "loss": 0.0006,
+      "step": 8887
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6303715940695511e-06,
+      "loss": 0.0039,
+      "step": 8888
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.630329771439326e-06,
+      "loss": 0.0048,
+      "step": 8889
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6302879488091007e-06,
+      "loss": 0.0224,
+      "step": 8890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6302461261788753e-06,
+      "loss": 0.0007,
+      "step": 8891
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.63020430354865e-06,
+      "loss": 0.0178,
+      "step": 8892
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6301624809184248e-06,
+      "loss": 0.0015,
+      "step": 8893
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6301206582881996e-06,
+      "loss": 0.0009,
+      "step": 8894
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6300788356579744e-06,
+      "loss": 0.0004,
+      "step": 8895
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6300370130277492e-06,
+      "loss": 0.0009,
+      "step": 8896
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629995190397524e-06,
+      "loss": 0.1622,
+      "step": 8897
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6299533677672988e-06,
+      "loss": 0.0007,
+      "step": 8898
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6299115451370735e-06,
+      "loss": 1.9171,
+      "step": 8899
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6298697225068483e-06,
+      "loss": 0.0365,
+      "step": 8900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6298278998766231e-06,
+      "loss": 1.0812,
+      "step": 8901
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6297860772463981e-06,
+      "loss": 0.006,
+      "step": 8902
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629744254616173e-06,
+      "loss": 0.0009,
+      "step": 8903
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6297024319859477e-06,
+      "loss": 0.0275,
+      "step": 8904
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6296606093557223e-06,
+      "loss": 0.0009,
+      "step": 8905
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629618786725497e-06,
+      "loss": 0.0014,
+      "step": 8906
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6295769640952718e-06,
+      "loss": 0.097,
+      "step": 8907
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6295351414650466e-06,
+      "loss": 0.1534,
+      "step": 8908
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6294933188348214e-06,
+      "loss": 0.0005,
+      "step": 8909
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6294514962045962e-06,
+      "loss": 2.877,
+      "step": 8910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629409673574371e-06,
+      "loss": 2.7175,
+      "step": 8911
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6293678509441458e-06,
+      "loss": 0.0007,
+      "step": 8912
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6293260283139206e-06,
+      "loss": 0.0006,
+      "step": 8913
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6292842056836953e-06,
+      "loss": 0.005,
+      "step": 8914
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6292423830534701e-06,
+      "loss": 0.0112,
+      "step": 8915
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6292005604232451e-06,
+      "loss": 0.8813,
+      "step": 8916
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.62915873779302e-06,
+      "loss": 0.0062,
+      "step": 8917
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6291169151627945e-06,
+      "loss": 0.0035,
+      "step": 8918
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6290750925325693e-06,
+      "loss": 0.0036,
+      "step": 8919
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629033269902344e-06,
+      "loss": 0.0005,
+      "step": 8920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6289914472721188e-06,
+      "loss": 0.0016,
+      "step": 8921
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6289496246418936e-06,
+      "loss": 0.0061,
+      "step": 8922
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6289078020116684e-06,
+      "loss": 0.0004,
+      "step": 8923
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6288659793814432e-06,
+      "loss": 0.0031,
+      "step": 8924
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628824156751218e-06,
+      "loss": 0.0043,
+      "step": 8925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6287823341209928e-06,
+      "loss": 0.0095,
+      "step": 8926
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6287405114907676e-06,
+      "loss": 0.0015,
+      "step": 8927
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6286986888605424e-06,
+      "loss": 0.0007,
+      "step": 8928
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6286568662303171e-06,
+      "loss": 0.0009,
+      "step": 8929
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628615043600092e-06,
+      "loss": 0.295,
+      "step": 8930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628573220969867e-06,
+      "loss": 0.8273,
+      "step": 8931
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6285313983396415e-06,
+      "loss": 1.3692,
+      "step": 8932
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6284895757094163e-06,
+      "loss": 0.0034,
+      "step": 8933
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628447753079191e-06,
+      "loss": 0.0016,
+      "step": 8934
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6284059304489659e-06,
+      "loss": 0.0009,
+      "step": 8935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6283641078187406e-06,
+      "loss": 0.0248,
+      "step": 8936
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6283222851885154e-06,
+      "loss": 0.0563,
+      "step": 8937
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6282804625582902e-06,
+      "loss": 2.1776,
+      "step": 8938
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628238639928065e-06,
+      "loss": 0.0005,
+      "step": 8939
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6281968172978398e-06,
+      "loss": 0.0587,
+      "step": 8940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6281549946676146e-06,
+      "loss": 0.0005,
+      "step": 8941
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6281131720373894e-06,
+      "loss": 0.7523,
+      "step": 8942
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6280713494071641e-06,
+      "loss": 0.0014,
+      "step": 8943
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628029526776939e-06,
+      "loss": 0.007,
+      "step": 8944
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6279877041467135e-06,
+      "loss": 0.003,
+      "step": 8945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6279458815164885e-06,
+      "loss": 0.02,
+      "step": 8946
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6279040588862633e-06,
+      "loss": 0.0011,
+      "step": 8947
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627862236256038e-06,
+      "loss": 0.0016,
+      "step": 8948
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6278204136258129e-06,
+      "loss": 0.0083,
+      "step": 8949
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6277785909955877e-06,
+      "loss": 0.0005,
+      "step": 8950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6277367683653624e-06,
+      "loss": 0.0026,
+      "step": 8951
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6276949457351372e-06,
+      "loss": 0.0006,
+      "step": 8952
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627653123104912e-06,
+      "loss": 0.0006,
+      "step": 8953
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6276113004746868e-06,
+      "loss": 0.0083,
+      "step": 8954
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6275694778444616e-06,
+      "loss": 0.0024,
+      "step": 8955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6275276552142364e-06,
+      "loss": 0.0025,
+      "step": 8956
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6274858325840112e-06,
+      "loss": 0.0403,
+      "step": 8957
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6274440099537857e-06,
+      "loss": 0.002,
+      "step": 8958
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6274021873235605e-06,
+      "loss": 0.0006,
+      "step": 8959
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6273603646933355e-06,
+      "loss": 0.0945,
+      "step": 8960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6273185420631103e-06,
+      "loss": 0.0008,
+      "step": 8961
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627276719432885e-06,
+      "loss": 0.0012,
+      "step": 8962
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6272348968026599e-06,
+      "loss": 0.011,
+      "step": 8963
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6271930741724347e-06,
+      "loss": 0.0011,
+      "step": 8964
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6271512515422094e-06,
+      "loss": 0.0027,
+      "step": 8965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6271094289119842e-06,
+      "loss": 0.0802,
+      "step": 8966
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627067606281759e-06,
+      "loss": 0.003,
+      "step": 8967
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6270257836515338e-06,
+      "loss": 0.0069,
+      "step": 8968
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6269839610213086e-06,
+      "loss": 0.0017,
+      "step": 8969
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6269421383910834e-06,
+      "loss": 0.494,
+      "step": 8970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6269003157608582e-06,
+      "loss": 1.5083,
+      "step": 8971
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6268584931306327e-06,
+      "loss": 0.001,
+      "step": 8972
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6268166705004075e-06,
+      "loss": 0.0102,
+      "step": 8973
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6267748478701823e-06,
+      "loss": 0.0188,
+      "step": 8974
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6267330252399573e-06,
+      "loss": 0.0023,
+      "step": 8975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626691202609732e-06,
+      "loss": 0.0026,
+      "step": 8976
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6266493799795069e-06,
+      "loss": 0.0006,
+      "step": 8977
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6266075573492817e-06,
+      "loss": 0.0035,
+      "step": 8978
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6265657347190565e-06,
+      "loss": 0.4345,
+      "step": 8979
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6265239120888312e-06,
+      "loss": 0.0021,
+      "step": 8980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626482089458606e-06,
+      "loss": 1.1236,
+      "step": 8981
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6264402668283808e-06,
+      "loss": 0.0287,
+      "step": 8982
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6263984441981556e-06,
+      "loss": 0.1453,
+      "step": 8983
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6263566215679304e-06,
+      "loss": 0.5127,
+      "step": 8984
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6263147989377052e-06,
+      "loss": 0.0026,
+      "step": 8985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6262729763074797e-06,
+      "loss": 0.5856,
+      "step": 8986
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6262311536772545e-06,
+      "loss": 0.0008,
+      "step": 8987
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6261893310470293e-06,
+      "loss": 0.5794,
+      "step": 8988
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6261475084168043e-06,
+      "loss": 0.0028,
+      "step": 8989
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626105685786579e-06,
+      "loss": 1.688,
+      "step": 8990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6260638631563539e-06,
+      "loss": 0.0009,
+      "step": 8991
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6260220405261287e-06,
+      "loss": 0.0013,
+      "step": 8992
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6259802178959035e-06,
+      "loss": 0.0021,
+      "step": 8993
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6259383952656783e-06,
+      "loss": 0.0005,
+      "step": 8994
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625896572635453e-06,
+      "loss": 3.749,
+      "step": 8995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6258547500052278e-06,
+      "loss": 0.005,
+      "step": 8996
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6258129273750026e-06,
+      "loss": 0.0056,
+      "step": 8997
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6257711047447774e-06,
+      "loss": 0.0096,
+      "step": 8998
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625729282114552e-06,
+      "loss": 0.001,
+      "step": 8999
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6256874594843268e-06,
+      "loss": 0.1231,
+      "step": 9000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6256456368541015e-06,
+      "loss": 0.0008,
+      "step": 9001
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6256038142238763e-06,
+      "loss": 0.0008,
+      "step": 9002
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6255619915936511e-06,
+      "loss": 0.0582,
+      "step": 9003
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6255201689634261e-06,
+      "loss": 4.2075,
+      "step": 9004
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625478346333201e-06,
+      "loss": 1.7467,
+      "step": 9005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6254365237029757e-06,
+      "loss": 0.0025,
+      "step": 9006
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6253947010727505e-06,
+      "loss": 1.3272,
+      "step": 9007
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6253528784425253e-06,
+      "loss": 0.4004,
+      "step": 9008
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6253110558123e-06,
+      "loss": 0.8063,
+      "step": 9009
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6252692331820748e-06,
+      "loss": 0.012,
+      "step": 9010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6252274105518496e-06,
+      "loss": 0.1027,
+      "step": 9011
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6251855879216244e-06,
+      "loss": 0.0327,
+      "step": 9012
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625143765291399e-06,
+      "loss": 0.0006,
+      "step": 9013
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6251019426611738e-06,
+      "loss": 0.6715,
+      "step": 9014
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6250601200309486e-06,
+      "loss": 0.0017,
+      "step": 9015
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6250182974007233e-06,
+      "loss": 0.0115,
+      "step": 9016
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6249764747704981e-06,
+      "loss": 0.0009,
+      "step": 9017
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.624934652140273e-06,
+      "loss": 0.0031,
+      "step": 9018
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.624892829510048e-06,
+      "loss": 0.0003,
+      "step": 9019
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6248510068798227e-06,
+      "loss": 0.002,
+      "step": 9020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6248091842495975e-06,
+      "loss": 0.4494,
+      "step": 9021
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6247673616193723e-06,
+      "loss": 0.0018,
+      "step": 9022
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.624725538989147e-06,
+      "loss": 0.0013,
+      "step": 9023
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6246837163589218e-06,
+      "loss": 0.7119,
+      "step": 9024
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6246418937286966e-06,
+      "loss": 0.0073,
+      "step": 9025
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6246000710984712e-06,
+      "loss": 0.0012,
+      "step": 9026
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.624558248468246e-06,
+      "loss": 0.0031,
+      "step": 9027
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6245164258380208e-06,
+      "loss": 1.3722,
+      "step": 9028
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6244746032077956e-06,
+      "loss": 0.001,
+      "step": 9029
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6244327805775703e-06,
+      "loss": 0.0213,
+      "step": 9030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6243909579473451e-06,
+      "loss": 0.0019,
+      "step": 9031
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.62434913531712e-06,
+      "loss": 2.1461,
+      "step": 9032
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.624307312686895e-06,
+      "loss": 0.0039,
+      "step": 9033
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6242654900566697e-06,
+      "loss": 0.9932,
+      "step": 9034
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6242236674264445e-06,
+      "loss": 0.0092,
+      "step": 9035
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6241818447962193e-06,
+      "loss": 0.0012,
+      "step": 9036
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.624140022165994e-06,
+      "loss": 0.0005,
+      "step": 9037
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6240981995357688e-06,
+      "loss": 0.0005,
+      "step": 9038
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6240563769055436e-06,
+      "loss": 0.0006,
+      "step": 9039
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6240145542753182e-06,
+      "loss": 0.0006,
+      "step": 9040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623972731645093e-06,
+      "loss": 0.001,
+      "step": 9041
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6239309090148678e-06,
+      "loss": 2.0212,
+      "step": 9042
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6238890863846426e-06,
+      "loss": 0.0005,
+      "step": 9043
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6238472637544174e-06,
+      "loss": 0.0009,
+      "step": 9044
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6238054411241921e-06,
+      "loss": 0.2174,
+      "step": 9045
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623763618493967e-06,
+      "loss": 0.0007,
+      "step": 9046
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6237217958637417e-06,
+      "loss": 0.0009,
+      "step": 9047
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6236799732335167e-06,
+      "loss": 0.0033,
+      "step": 9048
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6236381506032915e-06,
+      "loss": 0.0054,
+      "step": 9049
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6235963279730663e-06,
+      "loss": 0.0026,
+      "step": 9050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623554505342841e-06,
+      "loss": 0.0029,
+      "step": 9051
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6235126827126159e-06,
+      "loss": 0.0012,
+      "step": 9052
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6234708600823904e-06,
+      "loss": 0.2028,
+      "step": 9053
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6234290374521652e-06,
+      "loss": 0.0009,
+      "step": 9054
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.62338721482194e-06,
+      "loss": 0.001,
+      "step": 9055
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6233453921917148e-06,
+      "loss": 0.003,
+      "step": 9056
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6233035695614896e-06,
+      "loss": 1.6496,
+      "step": 9057
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6232617469312644e-06,
+      "loss": 0.0005,
+      "step": 9058
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6232199243010392e-06,
+      "loss": 0.0013,
+      "step": 9059
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623178101670814e-06,
+      "loss": 0.7947,
+      "step": 9060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6231362790405887e-06,
+      "loss": 0.0005,
+      "step": 9061
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6230944564103635e-06,
+      "loss": 0.0642,
+      "step": 9062
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6230526337801385e-06,
+      "loss": 0.0007,
+      "step": 9063
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6230108111499133e-06,
+      "loss": 0.0014,
+      "step": 9064
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622968988519688e-06,
+      "loss": 0.0396,
+      "step": 9065
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6229271658894629e-06,
+      "loss": 0.0005,
+      "step": 9066
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6228853432592374e-06,
+      "loss": 0.0019,
+      "step": 9067
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6228435206290122e-06,
+      "loss": 3.0823,
+      "step": 9068
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622801697998787e-06,
+      "loss": 0.0026,
+      "step": 9069
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6227598753685618e-06,
+      "loss": 0.0013,
+      "step": 9070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6227180527383366e-06,
+      "loss": 0.3495,
+      "step": 9071
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6226762301081114e-06,
+      "loss": 0.0028,
+      "step": 9072
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6226344074778862e-06,
+      "loss": 0.0068,
+      "step": 9073
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622592584847661e-06,
+      "loss": 0.0009,
+      "step": 9074
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6225507622174357e-06,
+      "loss": 3.2237,
+      "step": 9075
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6225089395872105e-06,
+      "loss": 0.0004,
+      "step": 9076
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6224671169569855e-06,
+      "loss": 2.2712,
+      "step": 9077
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6224252943267603e-06,
+      "loss": 0.0004,
+      "step": 9078
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622383471696535e-06,
+      "loss": 0.0021,
+      "step": 9079
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6223416490663099e-06,
+      "loss": 0.0005,
+      "step": 9080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6222998264360844e-06,
+      "loss": 0.0438,
+      "step": 9081
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6222580038058592e-06,
+      "loss": 0.0005,
+      "step": 9082
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622216181175634e-06,
+      "loss": 0.0005,
+      "step": 9083
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6221743585454088e-06,
+      "loss": 0.0013,
+      "step": 9084
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6221325359151836e-06,
+      "loss": 0.1232,
+      "step": 9085
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6220907132849584e-06,
+      "loss": 2.2731,
+      "step": 9086
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6220488906547332e-06,
+      "loss": 0.0022,
+      "step": 9087
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622007068024508e-06,
+      "loss": 0.2095,
+      "step": 9088
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6219652453942827e-06,
+      "loss": 0.0008,
+      "step": 9089
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6219234227640575e-06,
+      "loss": 0.0007,
+      "step": 9090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6218816001338323e-06,
+      "loss": 2.0181,
+      "step": 9091
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6218397775036073e-06,
+      "loss": 0.0006,
+      "step": 9092
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621797954873382e-06,
+      "loss": 0.0432,
+      "step": 9093
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6217561322431567e-06,
+      "loss": 0.2576,
+      "step": 9094
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6217143096129315e-06,
+      "loss": 0.0022,
+      "step": 9095
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6216724869827062e-06,
+      "loss": 2.1003,
+      "step": 9096
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621630664352481e-06,
+      "loss": 1.3,
+      "step": 9097
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6215888417222558e-06,
+      "loss": 0.0023,
+      "step": 9098
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6215470190920306e-06,
+      "loss": 0.4979,
+      "step": 9099
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6215051964618054e-06,
+      "loss": 0.1203,
+      "step": 9100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6214633738315802e-06,
+      "loss": 1.2697,
+      "step": 9101
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621421551201355e-06,
+      "loss": 0.0073,
+      "step": 9102
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6213797285711297e-06,
+      "loss": 0.2854,
+      "step": 9103
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6213379059409045e-06,
+      "loss": 0.0007,
+      "step": 9104
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6212960833106793e-06,
+      "loss": 0.0007,
+      "step": 9105
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6212542606804541e-06,
+      "loss": 2.0176,
+      "step": 9106
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6212124380502291e-06,
+      "loss": 0.0014,
+      "step": 9107
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6211706154200037e-06,
+      "loss": 0.0022,
+      "step": 9108
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6211287927897785e-06,
+      "loss": 0.0006,
+      "step": 9109
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6210869701595533e-06,
+      "loss": 0.0015,
+      "step": 9110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621045147529328e-06,
+      "loss": 0.0005,
+      "step": 9111
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6210033248991028e-06,
+      "loss": 0.0805,
+      "step": 9112
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6209615022688776e-06,
+      "loss": 0.8556,
+      "step": 9113
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6209196796386524e-06,
+      "loss": 0.0011,
+      "step": 9114
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6208778570084272e-06,
+      "loss": 0.0028,
+      "step": 9115
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620836034378202e-06,
+      "loss": 0.0156,
+      "step": 9116
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6207942117479768e-06,
+      "loss": 0.0017,
+      "step": 9117
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6207523891177515e-06,
+      "loss": 0.3756,
+      "step": 9118
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6207105664875263e-06,
+      "loss": 0.0009,
+      "step": 9119
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6206687438573011e-06,
+      "loss": 0.5461,
+      "step": 9120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620626921227076e-06,
+      "loss": 2.0785,
+      "step": 9121
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6205850985968507e-06,
+      "loss": 0.0115,
+      "step": 9122
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6205432759666255e-06,
+      "loss": 0.0023,
+      "step": 9123
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6205014533364003e-06,
+      "loss": 0.0008,
+      "step": 9124
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620459630706175e-06,
+      "loss": 0.0077,
+      "step": 9125
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6204178080759498e-06,
+      "loss": 0.0007,
+      "step": 9126
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6203759854457246e-06,
+      "loss": 0.2693,
+      "step": 9127
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6203341628154994e-06,
+      "loss": 0.1093,
+      "step": 9128
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6202923401852742e-06,
+      "loss": 0.006,
+      "step": 9129
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620250517555049e-06,
+      "loss": 0.3129,
+      "step": 9130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6202086949248238e-06,
+      "loss": 0.0008,
+      "step": 9131
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6201668722945986e-06,
+      "loss": 0.1496,
+      "step": 9132
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6201250496643733e-06,
+      "loss": 0.0027,
+      "step": 9133
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620083227034148e-06,
+      "loss": 0.0015,
+      "step": 9134
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6200414044039227e-06,
+      "loss": 0.0006,
+      "step": 9135
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6199995817736977e-06,
+      "loss": 0.0073,
+      "step": 9136
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6199577591434725e-06,
+      "loss": 0.0009,
+      "step": 9137
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6199159365132473e-06,
+      "loss": 0.0026,
+      "step": 9138
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619874113883022e-06,
+      "loss": 0.0873,
+      "step": 9139
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6198322912527968e-06,
+      "loss": 0.0598,
+      "step": 9140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6197904686225716e-06,
+      "loss": 0.0004,
+      "step": 9141
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6197486459923464e-06,
+      "loss": 0.0011,
+      "step": 9142
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6197068233621212e-06,
+      "loss": 0.8146,
+      "step": 9143
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619665000731896e-06,
+      "loss": 1.3509,
+      "step": 9144
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6196231781016708e-06,
+      "loss": 0.0006,
+      "step": 9145
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6195813554714456e-06,
+      "loss": 0.0017,
+      "step": 9146
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6195395328412203e-06,
+      "loss": 0.001,
+      "step": 9147
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619497710210995e-06,
+      "loss": 0.0008,
+      "step": 9148
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6194558875807697e-06,
+      "loss": 0.0017,
+      "step": 9149
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6194140649505445e-06,
+      "loss": 0.0025,
+      "step": 9150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6193722423203195e-06,
+      "loss": 0.3309,
+      "step": 9151
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6193304196900943e-06,
+      "loss": 1.4266,
+      "step": 9152
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619288597059869e-06,
+      "loss": 0.0011,
+      "step": 9153
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6192467744296439e-06,
+      "loss": 1.6545,
+      "step": 9154
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6192049517994186e-06,
+      "loss": 0.0057,
+      "step": 9155
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6191631291691934e-06,
+      "loss": 0.1461,
+      "step": 9156
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6191213065389682e-06,
+      "loss": 0.307,
+      "step": 9157
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619079483908743e-06,
+      "loss": 0.0007,
+      "step": 9158
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6190376612785178e-06,
+      "loss": 0.3534,
+      "step": 9159
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6189958386482926e-06,
+      "loss": 0.0019,
+      "step": 9160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6189540160180671e-06,
+      "loss": 0.0014,
+      "step": 9161
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618912193387842e-06,
+      "loss": 0.0008,
+      "step": 9162
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6188703707576167e-06,
+      "loss": 0.0005,
+      "step": 9163
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6188285481273915e-06,
+      "loss": 0.0004,
+      "step": 9164
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6187867254971665e-06,
+      "loss": 0.0015,
+      "step": 9165
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6187449028669413e-06,
+      "loss": 0.0026,
+      "step": 9166
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618703080236716e-06,
+      "loss": 2.0218,
+      "step": 9167
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6186612576064909e-06,
+      "loss": 0.001,
+      "step": 9168
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6186194349762656e-06,
+      "loss": 0.0042,
+      "step": 9169
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6185776123460404e-06,
+      "loss": 0.7639,
+      "step": 9170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6185357897158152e-06,
+      "loss": 0.0034,
+      "step": 9171
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.61849396708559e-06,
+      "loss": 0.0025,
+      "step": 9172
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6184521444553648e-06,
+      "loss": 0.9693,
+      "step": 9173
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6184103218251396e-06,
+      "loss": 0.0017,
+      "step": 9174
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6183684991949142e-06,
+      "loss": 0.0007,
+      "step": 9175
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.618326676564689e-06,
+      "loss": 0.1806,
+      "step": 9176
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6182848539344637e-06,
+      "loss": 0.0152,
+      "step": 9177
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6182430313042385e-06,
+      "loss": 0.0015,
+      "step": 9178
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6182012086740133e-06,
+      "loss": 0.002,
+      "step": 9179
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6181593860437883e-06,
+      "loss": 0.8198,
+      "step": 9180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.618117563413563e-06,
+      "loss": 0.0027,
+      "step": 9181
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6180757407833379e-06,
+      "loss": 0.0028,
+      "step": 9182
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6180339181531127e-06,
+      "loss": 0.0006,
+      "step": 9183
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6179920955228874e-06,
+      "loss": 0.001,
+      "step": 9184
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6179502728926622e-06,
+      "loss": 2.4489,
+      "step": 9185
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.617908450262437e-06,
+      "loss": 0.0009,
+      "step": 9186
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6178666276322118e-06,
+      "loss": 0.1436,
+      "step": 9187
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6178248050019866e-06,
+      "loss": 0.0059,
+      "step": 9188
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6177829823717612e-06,
+      "loss": 0.0015,
+      "step": 9189
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.617741159741536e-06,
+      "loss": 0.8331,
+      "step": 9190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6176993371113107e-06,
+      "loss": 0.0042,
+      "step": 9191
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6176575144810855e-06,
+      "loss": 0.0007,
+      "step": 9192
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6176156918508603e-06,
+      "loss": 0.0007,
+      "step": 9193
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6175738692206353e-06,
+      "loss": 3.2915,
+      "step": 9194
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61753204659041e-06,
+      "loss": 0.0039,
+      "step": 9195
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6174902239601849e-06,
+      "loss": 0.0027,
+      "step": 9196
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6174484013299597e-06,
+      "loss": 0.0022,
+      "step": 9197
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6174065786997345e-06,
+      "loss": 0.0052,
+      "step": 9198
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6173647560695092e-06,
+      "loss": 0.0016,
+      "step": 9199
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.617322933439284e-06,
+      "loss": 0.0018,
+      "step": 9200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6172811108090588e-06,
+      "loss": 0.2744,
+      "step": 9201
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6172392881788334e-06,
+      "loss": 0.0038,
+      "step": 9202
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6171974655486082e-06,
+      "loss": 0.0077,
+      "step": 9203
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.617155642918383e-06,
+      "loss": 0.0087,
+      "step": 9204
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6171138202881577e-06,
+      "loss": 0.0009,
+      "step": 9205
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6170719976579325e-06,
+      "loss": 0.0057,
+      "step": 9206
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6170301750277073e-06,
+      "loss": 0.0014,
+      "step": 9207
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616988352397482e-06,
+      "loss": 0.0012,
+      "step": 9208
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616946529767257e-06,
+      "loss": 0.0004,
+      "step": 9209
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6169047071370319e-06,
+      "loss": 0.0014,
+      "step": 9210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6168628845068067e-06,
+      "loss": 0.0005,
+      "step": 9211
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6168210618765815e-06,
+      "loss": 0.0006,
+      "step": 9212
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6167792392463562e-06,
+      "loss": 0.0011,
+      "step": 9213
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616737416616131e-06,
+      "loss": 0.0004,
+      "step": 9214
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6166955939859058e-06,
+      "loss": 0.001,
+      "step": 9215
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6166537713556804e-06,
+      "loss": 0.0005,
+      "step": 9216
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6166119487254552e-06,
+      "loss": 0.0008,
+      "step": 9217
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61657012609523e-06,
+      "loss": 0.0079,
+      "step": 9218
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6165283034650048e-06,
+      "loss": 0.009,
+      "step": 9219
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6164864808347795e-06,
+      "loss": 0.0004,
+      "step": 9220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6164446582045543e-06,
+      "loss": 0.0011,
+      "step": 9221
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6164028355743291e-06,
+      "loss": 0.0021,
+      "step": 9222
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616361012944104e-06,
+      "loss": 0.0012,
+      "step": 9223
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616319190313879e-06,
+      "loss": 0.0038,
+      "step": 9224
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6162773676836537e-06,
+      "loss": 0.0006,
+      "step": 9225
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6162355450534285e-06,
+      "loss": 0.0015,
+      "step": 9226
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6161937224232033e-06,
+      "loss": 0.0016,
+      "step": 9227
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616151899792978e-06,
+      "loss": 0.0009,
+      "step": 9228
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6161100771627526e-06,
+      "loss": 0.0094,
+      "step": 9229
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6160682545325274e-06,
+      "loss": 0.0007,
+      "step": 9230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6160264319023022e-06,
+      "loss": 0.0006,
+      "step": 9231
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615984609272077e-06,
+      "loss": 0.0011,
+      "step": 9232
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6159427866418518e-06,
+      "loss": 0.0024,
+      "step": 9233
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6159009640116265e-06,
+      "loss": 0.0077,
+      "step": 9234
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6158591413814013e-06,
+      "loss": 0.0013,
+      "step": 9235
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6158173187511761e-06,
+      "loss": 0.0028,
+      "step": 9236
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615775496120951e-06,
+      "loss": 0.0004,
+      "step": 9237
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615733673490726e-06,
+      "loss": 0.0054,
+      "step": 9238
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6156918508605007e-06,
+      "loss": 0.0081,
+      "step": 9239
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6156500282302755e-06,
+      "loss": 0.0005,
+      "step": 9240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6156082056000503e-06,
+      "loss": 0.1322,
+      "step": 9241
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615566382969825e-06,
+      "loss": 0.0008,
+      "step": 9242
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6155245603395996e-06,
+      "loss": 0.001,
+      "step": 9243
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6154827377093744e-06,
+      "loss": 0.0007,
+      "step": 9244
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6154409150791492e-06,
+      "loss": 1.5173,
+      "step": 9245
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615399092448924e-06,
+      "loss": 0.0006,
+      "step": 9246
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6153572698186988e-06,
+      "loss": 0.0004,
+      "step": 9247
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6153154471884736e-06,
+      "loss": 0.0014,
+      "step": 9248
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6152736245582483e-06,
+      "loss": 0.0061,
+      "step": 9249
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6152318019280231e-06,
+      "loss": 0.0005,
+      "step": 9250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615189979297798e-06,
+      "loss": 2.3455,
+      "step": 9251
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6151481566675727e-06,
+      "loss": 0.851,
+      "step": 9252
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6151063340373477e-06,
+      "loss": 0.0072,
+      "step": 9253
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6150645114071225e-06,
+      "loss": 0.0015,
+      "step": 9254
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6150226887768973e-06,
+      "loss": 0.0045,
+      "step": 9255
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6149808661466718e-06,
+      "loss": 2.7348,
+      "step": 9256
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6149390435164466e-06,
+      "loss": 2.2405,
+      "step": 9257
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6148972208862214e-06,
+      "loss": 0.009,
+      "step": 9258
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6148553982559962e-06,
+      "loss": 0.0004,
+      "step": 9259
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614813575625771e-06,
+      "loss": 0.0213,
+      "step": 9260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6147717529955458e-06,
+      "loss": 2.8913,
+      "step": 9261
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6147299303653206e-06,
+      "loss": 0.0066,
+      "step": 9262
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6146881077350953e-06,
+      "loss": 0.9092,
+      "step": 9263
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6146462851048701e-06,
+      "loss": 3.0851,
+      "step": 9264
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614604462474645e-06,
+      "loss": 0.5177,
+      "step": 9265
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6145626398444197e-06,
+      "loss": 0.001,
+      "step": 9266
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6145208172141945e-06,
+      "loss": 0.0044,
+      "step": 9267
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6144789945839695e-06,
+      "loss": 0.0004,
+      "step": 9268
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6144371719537443e-06,
+      "loss": 0.0018,
+      "step": 9269
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6143953493235189e-06,
+      "loss": 4.1771,
+      "step": 9270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6143535266932936e-06,
+      "loss": 0.0009,
+      "step": 9271
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6143117040630684e-06,
+      "loss": 1.5403,
+      "step": 9272
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6142698814328432e-06,
+      "loss": 0.0007,
+      "step": 9273
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614228058802618e-06,
+      "loss": 0.0014,
+      "step": 9274
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6141862361723928e-06,
+      "loss": 0.0041,
+      "step": 9275
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6141444135421676e-06,
+      "loss": 0.4822,
+      "step": 9276
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6141025909119424e-06,
+      "loss": 0.0014,
+      "step": 9277
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6140607682817171e-06,
+      "loss": 1.2623,
+      "step": 9278
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614018945651492e-06,
+      "loss": 0.0004,
+      "step": 9279
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6139771230212667e-06,
+      "loss": 0.0044,
+      "step": 9280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6139353003910415e-06,
+      "loss": 0.0009,
+      "step": 9281
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6138934777608165e-06,
+      "loss": 0.0036,
+      "step": 9282
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6138516551305913e-06,
+      "loss": 0.0034,
+      "step": 9283
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6138098325003659e-06,
+      "loss": 0.0007,
+      "step": 9284
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6137680098701406e-06,
+      "loss": 0.0356,
+      "step": 9285
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6137261872399154e-06,
+      "loss": 0.0027,
+      "step": 9286
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6136843646096902e-06,
+      "loss": 0.004,
+      "step": 9287
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613642541979465e-06,
+      "loss": 0.0808,
+      "step": 9288
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6136007193492398e-06,
+      "loss": 0.003,
+      "step": 9289
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6135588967190146e-06,
+      "loss": 1.1347,
+      "step": 9290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6135170740887894e-06,
+      "loss": 0.0063,
+      "step": 9291
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6134752514585642e-06,
+      "loss": 0.0062,
+      "step": 9292
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613433428828339e-06,
+      "loss": 4.0563,
+      "step": 9293
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6133916061981137e-06,
+      "loss": 0.5664,
+      "step": 9294
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6133497835678885e-06,
+      "loss": 0.4817,
+      "step": 9295
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613307960937663e-06,
+      "loss": 3.589,
+      "step": 9296
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613266138307438e-06,
+      "loss": 0.005,
+      "step": 9297
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6132243156772129e-06,
+      "loss": 0.0011,
+      "step": 9298
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6131824930469877e-06,
+      "loss": 0.072,
+      "step": 9299
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6131406704167624e-06,
+      "loss": 0.0013,
+      "step": 9300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6130988477865372e-06,
+      "loss": 1.6913,
+      "step": 9301
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613057025156312e-06,
+      "loss": 0.0227,
+      "step": 9302
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6130152025260868e-06,
+      "loss": 2.1562,
+      "step": 9303
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6129733798958616e-06,
+      "loss": 0.001,
+      "step": 9304
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6129315572656364e-06,
+      "loss": 0.0016,
+      "step": 9305
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6128897346354112e-06,
+      "loss": 0.0007,
+      "step": 9306
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612847912005186e-06,
+      "loss": 0.0023,
+      "step": 9307
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6128060893749607e-06,
+      "loss": 0.003,
+      "step": 9308
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6127642667447355e-06,
+      "loss": 0.0006,
+      "step": 9309
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61272244411451e-06,
+      "loss": 0.033,
+      "step": 9310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6126806214842849e-06,
+      "loss": 0.0008,
+      "step": 9311
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6126387988540599e-06,
+      "loss": 0.0008,
+      "step": 9312
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6125969762238347e-06,
+      "loss": 1.6049,
+      "step": 9313
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6125551535936095e-06,
+      "loss": 0.0041,
+      "step": 9314
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6125133309633842e-06,
+      "loss": 0.1794,
+      "step": 9315
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612471508333159e-06,
+      "loss": 1.465,
+      "step": 9316
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6124296857029338e-06,
+      "loss": 0.0023,
+      "step": 9317
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6123878630727086e-06,
+      "loss": 0.0125,
+      "step": 9318
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6123460404424834e-06,
+      "loss": 0.0107,
+      "step": 9319
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6123042178122582e-06,
+      "loss": 0.0009,
+      "step": 9320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612262395182033e-06,
+      "loss": 0.0012,
+      "step": 9321
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6122205725518077e-06,
+      "loss": 0.0022,
+      "step": 9322
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6121787499215825e-06,
+      "loss": 0.0027,
+      "step": 9323
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612136927291357e-06,
+      "loss": 0.0031,
+      "step": 9324
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6120951046611319e-06,
+      "loss": 0.0031,
+      "step": 9325
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6120532820309069e-06,
+      "loss": 0.0007,
+      "step": 9326
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6120114594006817e-06,
+      "loss": 0.0005,
+      "step": 9327
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6119696367704565e-06,
+      "loss": 0.002,
+      "step": 9328
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6119278141402312e-06,
+      "loss": 0.0256,
+      "step": 9329
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611885991510006e-06,
+      "loss": 0.9899,
+      "step": 9330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6118441688797808e-06,
+      "loss": 2.1754,
+      "step": 9331
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6118023462495556e-06,
+      "loss": 1.7547,
+      "step": 9332
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6117605236193304e-06,
+      "loss": 0.0011,
+      "step": 9333
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6117187009891052e-06,
+      "loss": 0.0011,
+      "step": 9334
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.61167687835888e-06,
+      "loss": 0.6973,
+      "step": 9335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6116350557286548e-06,
+      "loss": 0.0018,
+      "step": 9336
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6115932330984293e-06,
+      "loss": 0.0006,
+      "step": 9337
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6115514104682041e-06,
+      "loss": 0.1847,
+      "step": 9338
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.611509587837979e-06,
+      "loss": 0.0053,
+      "step": 9339
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6114677652077537e-06,
+      "loss": 0.0025,
+      "step": 9340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6114259425775287e-06,
+      "loss": 0.0029,
+      "step": 9341
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6113841199473035e-06,
+      "loss": 0.0054,
+      "step": 9342
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6113422973170783e-06,
+      "loss": 0.0005,
+      "step": 9343
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.611300474686853e-06,
+      "loss": 0.0019,
+      "step": 9344
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6112586520566278e-06,
+      "loss": 0.1301,
+      "step": 9345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6112168294264026e-06,
+      "loss": 0.1075,
+      "step": 9346
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6111750067961774e-06,
+      "loss": 0.0006,
+      "step": 9347
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6111331841659522e-06,
+      "loss": 0.0023,
+      "step": 9348
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.611091361535727e-06,
+      "loss": 0.0875,
+      "step": 9349
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6110495389055018e-06,
+      "loss": 1.4556,
+      "step": 9350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6110077162752763e-06,
+      "loss": 0.0013,
+      "step": 9351
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6109658936450511e-06,
+      "loss": 0.0011,
+      "step": 9352
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.610924071014826e-06,
+      "loss": 0.0005,
+      "step": 9353
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6108822483846007e-06,
+      "loss": 0.0016,
+      "step": 9354
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6108404257543755e-06,
+      "loss": 0.0011,
+      "step": 9355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6107986031241505e-06,
+      "loss": 0.0028,
+      "step": 9356
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6107567804939253e-06,
+      "loss": 2.0044,
+      "step": 9357
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6107149578637e-06,
+      "loss": 0.0058,
+      "step": 9358
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6106731352334748e-06,
+      "loss": 0.0144,
+      "step": 9359
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6106313126032496e-06,
+      "loss": 0.0066,
+      "step": 9360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6105894899730244e-06,
+      "loss": 0.0018,
+      "step": 9361
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6105476673427992e-06,
+      "loss": 0.0016,
+      "step": 9362
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.610505844712574e-06,
+      "loss": 0.2546,
+      "step": 9363
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6104640220823486e-06,
+      "loss": 0.3271,
+      "step": 9364
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6104221994521233e-06,
+      "loss": 0.0004,
+      "step": 9365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6103803768218981e-06,
+      "loss": 2.7148,
+      "step": 9366
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.610338554191673e-06,
+      "loss": 0.0004,
+      "step": 9367
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6102967315614477e-06,
+      "loss": 0.0005,
+      "step": 9368
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6102549089312225e-06,
+      "loss": 0.0015,
+      "step": 9369
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6102130863009975e-06,
+      "loss": 0.0009,
+      "step": 9370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6101712636707723e-06,
+      "loss": 0.0018,
+      "step": 9371
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.610129441040547e-06,
+      "loss": 0.0027,
+      "step": 9372
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6100876184103218e-06,
+      "loss": 0.0007,
+      "step": 9373
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6100457957800966e-06,
+      "loss": 0.0007,
+      "step": 9374
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6100039731498714e-06,
+      "loss": 0.0009,
+      "step": 9375
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6099621505196462e-06,
+      "loss": 0.6725,
+      "step": 9376
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609920327889421e-06,
+      "loss": 0.0023,
+      "step": 9377
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6098785052591956e-06,
+      "loss": 0.0005,
+      "step": 9378
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6098366826289704e-06,
+      "loss": 0.001,
+      "step": 9379
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6097948599987451e-06,
+      "loss": 0.0059,
+      "step": 9380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.60975303736852e-06,
+      "loss": 0.0009,
+      "step": 9381
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6097112147382947e-06,
+      "loss": 0.0019,
+      "step": 9382
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6096693921080695e-06,
+      "loss": 0.0004,
+      "step": 9383
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6096275694778443e-06,
+      "loss": 0.0047,
+      "step": 9384
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6095857468476193e-06,
+      "loss": 0.0018,
+      "step": 9385
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609543924217394e-06,
+      "loss": 0.0006,
+      "step": 9386
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6095021015871689e-06,
+      "loss": 0.1658,
+      "step": 9387
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6094602789569436e-06,
+      "loss": 0.1602,
+      "step": 9388
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6094184563267184e-06,
+      "loss": 0.0474,
+      "step": 9389
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6093766336964932e-06,
+      "loss": 0.0008,
+      "step": 9390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609334811066268e-06,
+      "loss": 0.0029,
+      "step": 9391
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6092929884360426e-06,
+      "loss": 0.0004,
+      "step": 9392
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6092511658058174e-06,
+      "loss": 0.0005,
+      "step": 9393
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6092093431755921e-06,
+      "loss": 0.0015,
+      "step": 9394
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609167520545367e-06,
+      "loss": 0.2394,
+      "step": 9395
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6091256979151417e-06,
+      "loss": 0.0073,
+      "step": 9396
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6090838752849165e-06,
+      "loss": 0.0017,
+      "step": 9397
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6090420526546913e-06,
+      "loss": 2.3648,
+      "step": 9398
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6090002300244663e-06,
+      "loss": 0.001,
+      "step": 9399
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608958407394241e-06,
+      "loss": 2.679,
+      "step": 9400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6089165847640159e-06,
+      "loss": 0.0005,
+      "step": 9401
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6088747621337906e-06,
+      "loss": 0.0006,
+      "step": 9402
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6088329395035654e-06,
+      "loss": 0.0034,
+      "step": 9403
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6087911168733402e-06,
+      "loss": 0.0006,
+      "step": 9404
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6087492942431148e-06,
+      "loss": 0.0019,
+      "step": 9405
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6087074716128896e-06,
+      "loss": 0.0047,
+      "step": 9406
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6086656489826644e-06,
+      "loss": 0.002,
+      "step": 9407
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6086238263524392e-06,
+      "loss": 0.0065,
+      "step": 9408
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608582003722214e-06,
+      "loss": 0.0006,
+      "step": 9409
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6085401810919887e-06,
+      "loss": 0.0007,
+      "step": 9410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6084983584617635e-06,
+      "loss": 0.0007,
+      "step": 9411
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6084565358315383e-06,
+      "loss": 0.0007,
+      "step": 9412
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608414713201313e-06,
+      "loss": 0.0008,
+      "step": 9413
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608372890571088e-06,
+      "loss": 0.0023,
+      "step": 9414
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6083310679408629e-06,
+      "loss": 0.0022,
+      "step": 9415
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6082892453106377e-06,
+      "loss": 0.0005,
+      "step": 9416
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6082474226804124e-06,
+      "loss": 0.0006,
+      "step": 9417
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6082056000501872e-06,
+      "loss": 3.257,
+      "step": 9418
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6081637774199618e-06,
+      "loss": 0.0011,
+      "step": 9419
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6081219547897366e-06,
+      "loss": 0.0006,
+      "step": 9420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6080801321595114e-06,
+      "loss": 0.0008,
+      "step": 9421
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6080383095292862e-06,
+      "loss": 0.002,
+      "step": 9422
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607996486899061e-06,
+      "loss": 1.4135,
+      "step": 9423
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6079546642688357e-06,
+      "loss": 0.0157,
+      "step": 9424
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6079128416386105e-06,
+      "loss": 0.0005,
+      "step": 9425
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6078710190083853e-06,
+      "loss": 0.0154,
+      "step": 9426
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.60782919637816e-06,
+      "loss": 0.0011,
+      "step": 9427
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6077873737479349e-06,
+      "loss": 0.0006,
+      "step": 9428
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6077455511177099e-06,
+      "loss": 2.8966,
+      "step": 9429
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6077037284874847e-06,
+      "loss": 1.1258,
+      "step": 9430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6076619058572595e-06,
+      "loss": 0.001,
+      "step": 9431
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607620083227034e-06,
+      "loss": 0.0178,
+      "step": 9432
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6075782605968088e-06,
+      "loss": 0.0009,
+      "step": 9433
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6075364379665836e-06,
+      "loss": 0.0018,
+      "step": 9434
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6074946153363584e-06,
+      "loss": 0.001,
+      "step": 9435
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6074527927061332e-06,
+      "loss": 0.0007,
+      "step": 9436
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607410970075908e-06,
+      "loss": 0.0054,
+      "step": 9437
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6073691474456827e-06,
+      "loss": 0.0715,
+      "step": 9438
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6073273248154575e-06,
+      "loss": 0.3803,
+      "step": 9439
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6072855021852323e-06,
+      "loss": 0.0018,
+      "step": 9440
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607243679555007e-06,
+      "loss": 0.0042,
+      "step": 9441
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6072018569247819e-06,
+      "loss": 0.001,
+      "step": 9442
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6071600342945569e-06,
+      "loss": 0.0215,
+      "step": 9443
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6071182116643317e-06,
+      "loss": 1.8176,
+      "step": 9444
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6070763890341065e-06,
+      "loss": 0.0022,
+      "step": 9445
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607034566403881e-06,
+      "loss": 0.0041,
+      "step": 9446
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6069927437736558e-06,
+      "loss": 0.2054,
+      "step": 9447
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6069509211434306e-06,
+      "loss": 0.006,
+      "step": 9448
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6069090985132054e-06,
+      "loss": 0.023,
+      "step": 9449
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6068672758829802e-06,
+      "loss": 0.0014,
+      "step": 9450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606825453252755e-06,
+      "loss": 0.0015,
+      "step": 9451
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6067836306225298e-06,
+      "loss": 0.6075,
+      "step": 9452
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6067418079923045e-06,
+      "loss": 0.0073,
+      "step": 9453
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6066999853620793e-06,
+      "loss": 0.005,
+      "step": 9454
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6066581627318541e-06,
+      "loss": 0.0018,
+      "step": 9455
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606616340101629e-06,
+      "loss": 0.0015,
+      "step": 9456
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6065745174714037e-06,
+      "loss": 0.001,
+      "step": 9457
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6065326948411787e-06,
+      "loss": 0.0007,
+      "step": 9458
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6064908722109533e-06,
+      "loss": 0.0005,
+      "step": 9459
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606449049580728e-06,
+      "loss": 0.0005,
+      "step": 9460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6064072269505028e-06,
+      "loss": 0.0004,
+      "step": 9461
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6063654043202776e-06,
+      "loss": 0.0009,
+      "step": 9462
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6063235816900524e-06,
+      "loss": 0.0005,
+      "step": 9463
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6062817590598272e-06,
+      "loss": 1.5512,
+      "step": 9464
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606239936429602e-06,
+      "loss": 0.7276,
+      "step": 9465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6061981137993768e-06,
+      "loss": 0.5707,
+      "step": 9466
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6061562911691515e-06,
+      "loss": 0.002,
+      "step": 9467
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6061144685389263e-06,
+      "loss": 0.002,
+      "step": 9468
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6060726459087011e-06,
+      "loss": 0.0006,
+      "step": 9469
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606030823278476e-06,
+      "loss": 0.047,
+      "step": 9470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6059890006482507e-06,
+      "loss": 0.0015,
+      "step": 9471
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6059471780180253e-06,
+      "loss": 1.8285,
+      "step": 9472
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6059053553878003e-06,
+      "loss": 2.3693,
+      "step": 9473
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605863532757575e-06,
+      "loss": 0.0005,
+      "step": 9474
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6058217101273498e-06,
+      "loss": 2.1235,
+      "step": 9475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6057798874971246e-06,
+      "loss": 0.3759,
+      "step": 9476
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6057380648668994e-06,
+      "loss": 0.0006,
+      "step": 9477
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6056962422366742e-06,
+      "loss": 0.0042,
+      "step": 9478
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605654419606449e-06,
+      "loss": 0.0004,
+      "step": 9479
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6056125969762238e-06,
+      "loss": 0.0018,
+      "step": 9480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6055707743459986e-06,
+      "loss": 0.0263,
+      "step": 9481
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6055289517157733e-06,
+      "loss": 0.0008,
+      "step": 9482
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6054871290855481e-06,
+      "loss": 0.0005,
+      "step": 9483
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605445306455323e-06,
+      "loss": 0.0018,
+      "step": 9484
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6054034838250977e-06,
+      "loss": 0.0226,
+      "step": 9485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6053616611948723e-06,
+      "loss": 0.7525,
+      "step": 9486
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6053198385646473e-06,
+      "loss": 0.0014,
+      "step": 9487
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605278015934422e-06,
+      "loss": 0.0022,
+      "step": 9488
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6052361933041968e-06,
+      "loss": 0.0078,
+      "step": 9489
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6051943706739716e-06,
+      "loss": 0.0006,
+      "step": 9490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6051525480437464e-06,
+      "loss": 0.0007,
+      "step": 9491
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6051107254135212e-06,
+      "loss": 0.004,
+      "step": 9492
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605068902783296e-06,
+      "loss": 0.0939,
+      "step": 9493
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6050270801530708e-06,
+      "loss": 0.0012,
+      "step": 9494
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6049852575228456e-06,
+      "loss": 0.0026,
+      "step": 9495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6049434348926204e-06,
+      "loss": 0.0005,
+      "step": 9496
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6049016122623951e-06,
+      "loss": 1.3428,
+      "step": 9497
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.60485978963217e-06,
+      "loss": 0.0087,
+      "step": 9498
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6048179670019445e-06,
+      "loss": 0.0007,
+      "step": 9499
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6047761443717193e-06,
+      "loss": 1.734,
+      "step": 9500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.604734321741494e-06,
+      "loss": 0.0005,
+      "step": 9501
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.604692499111269e-06,
+      "loss": 0.0004,
+      "step": 9502
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6046506764810439e-06,
+      "loss": 0.0006,
+      "step": 9503
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6046088538508186e-06,
+      "loss": 0.0009,
+      "step": 9504
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6045670312205934e-06,
+      "loss": 0.0118,
+      "step": 9505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6045252085903682e-06,
+      "loss": 0.0062,
+      "step": 9506
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.604483385960143e-06,
+      "loss": 0.0969,
+      "step": 9507
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6044415633299178e-06,
+      "loss": 0.0004,
+      "step": 9508
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6043997406996926e-06,
+      "loss": 0.0012,
+      "step": 9509
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6043579180694674e-06,
+      "loss": 0.0005,
+      "step": 9510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6043160954392421e-06,
+      "loss": 0.0031,
+      "step": 9511
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.604274272809017e-06,
+      "loss": 0.0018,
+      "step": 9512
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6042324501787915e-06,
+      "loss": 0.0005,
+      "step": 9513
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6041906275485663e-06,
+      "loss": 0.0006,
+      "step": 9514
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.604148804918341e-06,
+      "loss": 0.0123,
+      "step": 9515
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6041069822881159e-06,
+      "loss": 1.3317,
+      "step": 9516
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6040651596578909e-06,
+      "loss": 2.6765,
+      "step": 9517
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6040233370276657e-06,
+      "loss": 0.2171,
+      "step": 9518
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6039815143974404e-06,
+      "loss": 0.0018,
+      "step": 9519
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6039396917672152e-06,
+      "loss": 0.0027,
+      "step": 9520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.60389786913699e-06,
+      "loss": 0.0011,
+      "step": 9521
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6038560465067648e-06,
+      "loss": 0.0032,
+      "step": 9522
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6038142238765396e-06,
+      "loss": 0.0007,
+      "step": 9523
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6037724012463144e-06,
+      "loss": 0.1998,
+      "step": 9524
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6037305786160892e-06,
+      "loss": 0.1301,
+      "step": 9525
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.603688755985864e-06,
+      "loss": 0.6271,
+      "step": 9526
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6036469333556385e-06,
+      "loss": 0.004,
+      "step": 9527
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6036051107254133e-06,
+      "loss": 0.0014,
+      "step": 9528
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.603563288095188e-06,
+      "loss": 0.0037,
+      "step": 9529
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6035214654649629e-06,
+      "loss": 0.0007,
+      "step": 9530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6034796428347379e-06,
+      "loss": 0.0029,
+      "step": 9531
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6034378202045127e-06,
+      "loss": 0.105,
+      "step": 9532
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6033959975742874e-06,
+      "loss": 0.0066,
+      "step": 9533
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6033541749440622e-06,
+      "loss": 0.0011,
+      "step": 9534
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.603312352313837e-06,
+      "loss": 0.0093,
+      "step": 9535
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6032705296836118e-06,
+      "loss": 0.0005,
+      "step": 9536
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6032287070533866e-06,
+      "loss": 0.129,
+      "step": 9537
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6031868844231614e-06,
+      "loss": 0.0035,
+      "step": 9538
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6031450617929362e-06,
+      "loss": 0.0006,
+      "step": 9539
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6031032391627107e-06,
+      "loss": 0.0015,
+      "step": 9540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6030614165324855e-06,
+      "loss": 0.0033,
+      "step": 9541
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6030195939022603e-06,
+      "loss": 0.0006,
+      "step": 9542
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602977771272035e-06,
+      "loss": 0.0046,
+      "step": 9543
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6029359486418099e-06,
+      "loss": 0.0006,
+      "step": 9544
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6028941260115847e-06,
+      "loss": 0.0015,
+      "step": 9545
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6028523033813597e-06,
+      "loss": 0.0019,
+      "step": 9546
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6028104807511345e-06,
+      "loss": 0.0048,
+      "step": 9547
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6027686581209092e-06,
+      "loss": 0.0007,
+      "step": 9548
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602726835490684e-06,
+      "loss": 0.0055,
+      "step": 9549
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6026850128604588e-06,
+      "loss": 0.0019,
+      "step": 9550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6026431902302336e-06,
+      "loss": 0.0007,
+      "step": 9551
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6026013676000084e-06,
+      "loss": 0.2838,
+      "step": 9552
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6025595449697832e-06,
+      "loss": 0.0019,
+      "step": 9553
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6025177223395577e-06,
+      "loss": 0.002,
+      "step": 9554
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6024758997093325e-06,
+      "loss": 0.0381,
+      "step": 9555
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6024340770791073e-06,
+      "loss": 0.4039,
+      "step": 9556
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6023922544488821e-06,
+      "loss": 0.0007,
+      "step": 9557
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6023504318186569e-06,
+      "loss": 0.0008,
+      "step": 9558
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6023086091884317e-06,
+      "loss": 0.0006,
+      "step": 9559
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6022667865582065e-06,
+      "loss": 0.0004,
+      "step": 9560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6022249639279815e-06,
+      "loss": 0.0006,
+      "step": 9561
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6021831412977563e-06,
+      "loss": 0.0004,
+      "step": 9562
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602141318667531e-06,
+      "loss": 0.0005,
+      "step": 9563
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6020994960373058e-06,
+      "loss": 0.0031,
+      "step": 9564
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6020576734070806e-06,
+      "loss": 0.0027,
+      "step": 9565
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6020158507768554e-06,
+      "loss": 0.7338,
+      "step": 9566
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.60197402814663e-06,
+      "loss": 0.0004,
+      "step": 9567
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6019322055164048e-06,
+      "loss": 0.0007,
+      "step": 9568
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6018903828861795e-06,
+      "loss": 0.2305,
+      "step": 9569
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6018485602559543e-06,
+      "loss": 0.0425,
+      "step": 9570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6018067376257291e-06,
+      "loss": 1.6527,
+      "step": 9571
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601764914995504e-06,
+      "loss": 0.0017,
+      "step": 9572
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6017230923652787e-06,
+      "loss": 0.73,
+      "step": 9573
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6016812697350535e-06,
+      "loss": 0.0009,
+      "step": 9574
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6016394471048285e-06,
+      "loss": 0.0004,
+      "step": 9575
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6015976244746033e-06,
+      "loss": 0.0271,
+      "step": 9576
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601555801844378e-06,
+      "loss": 0.0032,
+      "step": 9577
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6015139792141528e-06,
+      "loss": 0.001,
+      "step": 9578
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6014721565839276e-06,
+      "loss": 0.0012,
+      "step": 9579
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6014303339537024e-06,
+      "loss": 0.0004,
+      "step": 9580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601388511323477e-06,
+      "loss": 0.0016,
+      "step": 9581
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6013466886932518e-06,
+      "loss": 0.0005,
+      "step": 9582
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6013048660630266e-06,
+      "loss": 0.0003,
+      "step": 9583
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6012630434328013e-06,
+      "loss": 0.6435,
+      "step": 9584
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6012212208025761e-06,
+      "loss": 0.0006,
+      "step": 9585
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601179398172351e-06,
+      "loss": 0.0018,
+      "step": 9586
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6011375755421257e-06,
+      "loss": 0.3304,
+      "step": 9587
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6010957529119005e-06,
+      "loss": 0.0023,
+      "step": 9588
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6010539302816753e-06,
+      "loss": 0.0578,
+      "step": 9589
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6010121076514503e-06,
+      "loss": 0.0004,
+      "step": 9590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600970285021225e-06,
+      "loss": 2.4848,
+      "step": 9591
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6009284623909998e-06,
+      "loss": 0.0005,
+      "step": 9592
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6008866397607746e-06,
+      "loss": 0.0006,
+      "step": 9593
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6008448171305494e-06,
+      "loss": 0.0022,
+      "step": 9594
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600802994500324e-06,
+      "loss": 0.001,
+      "step": 9595
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6007611718700988e-06,
+      "loss": 0.0051,
+      "step": 9596
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6007193492398736e-06,
+      "loss": 0.0005,
+      "step": 9597
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6006775266096483e-06,
+      "loss": 0.0004,
+      "step": 9598
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6006357039794231e-06,
+      "loss": 2.9536,
+      "step": 9599
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600593881349198e-06,
+      "loss": 2.9792,
+      "step": 9600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6005520587189727e-06,
+      "loss": 0.0007,
+      "step": 9601
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6005102360887475e-06,
+      "loss": 1.1869,
+      "step": 9602
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6004684134585223e-06,
+      "loss": 4.0664,
+      "step": 9603
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6004265908282973e-06,
+      "loss": 0.0005,
+      "step": 9604
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600384768198072e-06,
+      "loss": 0.0009,
+      "step": 9605
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6003429455678468e-06,
+      "loss": 0.0003,
+      "step": 9606
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6003011229376216e-06,
+      "loss": 0.0003,
+      "step": 9607
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6002593003073962e-06,
+      "loss": 3.0994,
+      "step": 9608
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600217477677171e-06,
+      "loss": 0.0041,
+      "step": 9609
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6001756550469458e-06,
+      "loss": 0.0097,
+      "step": 9610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6001338324167206e-06,
+      "loss": 0.001,
+      "step": 9611
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6000920097864954e-06,
+      "loss": 0.0046,
+      "step": 9612
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6000501871562701e-06,
+      "loss": 0.0012,
+      "step": 9613
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600008364526045e-06,
+      "loss": 0.0006,
+      "step": 9614
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5999665418958197e-06,
+      "loss": 0.0007,
+      "step": 9615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5999247192655945e-06,
+      "loss": 0.0046,
+      "step": 9616
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5998828966353693e-06,
+      "loss": 0.0044,
+      "step": 9617
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599841074005144e-06,
+      "loss": 0.0012,
+      "step": 9618
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599799251374919e-06,
+      "loss": 0.0086,
+      "step": 9619
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5997574287446939e-06,
+      "loss": 0.0006,
+      "step": 9620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5997156061144686e-06,
+      "loss": 0.0006,
+      "step": 9621
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5996737834842432e-06,
+      "loss": 0.5777,
+      "step": 9622
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599631960854018e-06,
+      "loss": 0.0016,
+      "step": 9623
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5995901382237928e-06,
+      "loss": 0.0005,
+      "step": 9624
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5995483155935676e-06,
+      "loss": 0.0024,
+      "step": 9625
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5995064929633424e-06,
+      "loss": 1.9725,
+      "step": 9626
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5994646703331172e-06,
+      "loss": 0.0005,
+      "step": 9627
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599422847702892e-06,
+      "loss": 2.6062,
+      "step": 9628
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5993810250726667e-06,
+      "loss": 0.019,
+      "step": 9629
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5993392024424415e-06,
+      "loss": 0.0027,
+      "step": 9630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5992973798122163e-06,
+      "loss": 0.9721,
+      "step": 9631
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599255557181991e-06,
+      "loss": 0.0005,
+      "step": 9632
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5992137345517659e-06,
+      "loss": 0.0008,
+      "step": 9633
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5991719119215409e-06,
+      "loss": 0.003,
+      "step": 9634
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5991300892913154e-06,
+      "loss": 0.0008,
+      "step": 9635
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5990882666610902e-06,
+      "loss": 0.0069,
+      "step": 9636
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599046444030865e-06,
+      "loss": 0.0008,
+      "step": 9637
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5990046214006398e-06,
+      "loss": 0.2728,
+      "step": 9638
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5989627987704146e-06,
+      "loss": 0.001,
+      "step": 9639
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5989209761401894e-06,
+      "loss": 1.9659,
+      "step": 9640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5988791535099642e-06,
+      "loss": 1.7003,
+      "step": 9641
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598837330879739e-06,
+      "loss": 1.1295,
+      "step": 9642
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5987955082495137e-06,
+      "loss": 0.0009,
+      "step": 9643
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5987536856192885e-06,
+      "loss": 0.0021,
+      "step": 9644
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5987118629890633e-06,
+      "loss": 0.0049,
+      "step": 9645
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598670040358838e-06,
+      "loss": 0.0009,
+      "step": 9646
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5986282177286129e-06,
+      "loss": 0.0048,
+      "step": 9647
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5985863950983879e-06,
+      "loss": 0.002,
+      "step": 9648
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5985445724681624e-06,
+      "loss": 0.0007,
+      "step": 9649
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5985027498379372e-06,
+      "loss": 0.0008,
+      "step": 9650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598460927207712e-06,
+      "loss": 0.0009,
+      "step": 9651
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5984191045774868e-06,
+      "loss": 0.0016,
+      "step": 9652
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5983772819472616e-06,
+      "loss": 0.0003,
+      "step": 9653
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5983354593170364e-06,
+      "loss": 0.0006,
+      "step": 9654
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5982936366868112e-06,
+      "loss": 0.0007,
+      "step": 9655
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.598251814056586e-06,
+      "loss": 0.0005,
+      "step": 9656
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5982099914263607e-06,
+      "loss": 1.1944,
+      "step": 9657
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5981681687961355e-06,
+      "loss": 0.0038,
+      "step": 9658
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5981263461659103e-06,
+      "loss": 0.0005,
+      "step": 9659
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.598084523535685e-06,
+      "loss": 0.9669,
+      "step": 9660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5980427009054599e-06,
+      "loss": 0.0016,
+      "step": 9661
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5980008782752345e-06,
+      "loss": 0.0011,
+      "step": 9662
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5979590556450095e-06,
+      "loss": 0.0023,
+      "step": 9663
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5979172330147842e-06,
+      "loss": 1.8874,
+      "step": 9664
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.597875410384559e-06,
+      "loss": 0.6813,
+      "step": 9665
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5978335877543338e-06,
+      "loss": 0.0005,
+      "step": 9666
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5977917651241086e-06,
+      "loss": 0.2654,
+      "step": 9667
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5977499424938834e-06,
+      "loss": 0.0038,
+      "step": 9668
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5977081198636582e-06,
+      "loss": 0.0178,
+      "step": 9669
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.597666297233433e-06,
+      "loss": 0.2285,
+      "step": 9670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5976244746032077e-06,
+      "loss": 0.006,
+      "step": 9671
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5975826519729825e-06,
+      "loss": 0.7416,
+      "step": 9672
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5975408293427573e-06,
+      "loss": 0.0011,
+      "step": 9673
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5974990067125321e-06,
+      "loss": 0.0022,
+      "step": 9674
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5974571840823067e-06,
+      "loss": 0.0009,
+      "step": 9675
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5974153614520815e-06,
+      "loss": 0.0014,
+      "step": 9676
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5973735388218563e-06,
+      "loss": 0.0006,
+      "step": 9677
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5973317161916313e-06,
+      "loss": 0.0109,
+      "step": 9678
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.597289893561406e-06,
+      "loss": 0.001,
+      "step": 9679
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5972480709311808e-06,
+      "loss": 0.0008,
+      "step": 9680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5972062483009556e-06,
+      "loss": 0.1668,
+      "step": 9681
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5971644256707304e-06,
+      "loss": 0.1062,
+      "step": 9682
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5971226030405052e-06,
+      "loss": 0.3473,
+      "step": 9683
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.59708078041028e-06,
+      "loss": 0.0785,
+      "step": 9684
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5970389577800548e-06,
+      "loss": 3.3303,
+      "step": 9685
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5969971351498295e-06,
+      "loss": 0.0018,
+      "step": 9686
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5969553125196043e-06,
+      "loss": 0.2705,
+      "step": 9687
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5969134898893791e-06,
+      "loss": 0.0035,
+      "step": 9688
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5968716672591537e-06,
+      "loss": 0.0219,
+      "step": 9689
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5968298446289285e-06,
+      "loss": 0.0067,
+      "step": 9690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5967880219987033e-06,
+      "loss": 0.0296,
+      "step": 9691
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5967461993684783e-06,
+      "loss": 0.2875,
+      "step": 9692
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596704376738253e-06,
+      "loss": 0.0025,
+      "step": 9693
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5966625541080278e-06,
+      "loss": 0.0006,
+      "step": 9694
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5966207314778026e-06,
+      "loss": 0.0013,
+      "step": 9695
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5965789088475774e-06,
+      "loss": 0.0006,
+      "step": 9696
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5965370862173522e-06,
+      "loss": 2.6912,
+      "step": 9697
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596495263587127e-06,
+      "loss": 0.0038,
+      "step": 9698
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5964534409569018e-06,
+      "loss": 0.0016,
+      "step": 9699
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5964116183266766e-06,
+      "loss": 3.4954,
+      "step": 9700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5963697956964513e-06,
+      "loss": 0.0014,
+      "step": 9701
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5963279730662261e-06,
+      "loss": 0.0007,
+      "step": 9702
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5962861504360007e-06,
+      "loss": 0.0007,
+      "step": 9703
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5962443278057755e-06,
+      "loss": 3.294,
+      "step": 9704
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5962025051755503e-06,
+      "loss": 2.1052,
+      "step": 9705
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596160682545325e-06,
+      "loss": 3.9191,
+      "step": 9706
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5961188599151e-06,
+      "loss": 0.001,
+      "step": 9707
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5960770372848748e-06,
+      "loss": 0.0407,
+      "step": 9708
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5960352146546496e-06,
+      "loss": 0.0006,
+      "step": 9709
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5959933920244244e-06,
+      "loss": 0.0008,
+      "step": 9710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5959515693941992e-06,
+      "loss": 0.0011,
+      "step": 9711
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595909746763974e-06,
+      "loss": 0.3246,
+      "step": 9712
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5958679241337488e-06,
+      "loss": 0.0207,
+      "step": 9713
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5958261015035236e-06,
+      "loss": 0.0029,
+      "step": 9714
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5957842788732983e-06,
+      "loss": 0.0007,
+      "step": 9715
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595742456243073e-06,
+      "loss": 0.0028,
+      "step": 9716
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5957006336128477e-06,
+      "loss": 0.0014,
+      "step": 9717
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5956588109826225e-06,
+      "loss": 0.7802,
+      "step": 9718
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5956169883523973e-06,
+      "loss": 0.0008,
+      "step": 9719
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595575165722172e-06,
+      "loss": 0.0014,
+      "step": 9720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5955333430919469e-06,
+      "loss": 0.0056,
+      "step": 9721
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5954915204617219e-06,
+      "loss": 0.7908,
+      "step": 9722
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5954496978314966e-06,
+      "loss": 0.3245,
+      "step": 9723
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5954078752012714e-06,
+      "loss": 0.001,
+      "step": 9724
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5953660525710462e-06,
+      "loss": 0.0017,
+      "step": 9725
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595324229940821e-06,
+      "loss": 0.001,
+      "step": 9726
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5952824073105958e-06,
+      "loss": 0.0005,
+      "step": 9727
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5952405846803706e-06,
+      "loss": 0.0009,
+      "step": 9728
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5951987620501454e-06,
+      "loss": 0.0045,
+      "step": 9729
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.59515693941992e-06,
+      "loss": 0.0016,
+      "step": 9730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5951151167896947e-06,
+      "loss": 2.1219,
+      "step": 9731
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5950732941594695e-06,
+      "loss": 0.0004,
+      "step": 9732
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5950314715292443e-06,
+      "loss": 0.0013,
+      "step": 9733
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594989648899019e-06,
+      "loss": 0.0024,
+      "step": 9734
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5949478262687939e-06,
+      "loss": 0.1528,
+      "step": 9735
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5949060036385689e-06,
+      "loss": 0.0025,
+      "step": 9736
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5948641810083436e-06,
+      "loss": 0.3165,
+      "step": 9737
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5948223583781184e-06,
+      "loss": 0.0017,
+      "step": 9738
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5947805357478932e-06,
+      "loss": 0.0005,
+      "step": 9739
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594738713117668e-06,
+      "loss": 0.0006,
+      "step": 9740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5946968904874428e-06,
+      "loss": 1.2732,
+      "step": 9741
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5946550678572176e-06,
+      "loss": 0.0004,
+      "step": 9742
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5946132452269922e-06,
+      "loss": 1.7864,
+      "step": 9743
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594571422596767e-06,
+      "loss": 0.0009,
+      "step": 9744
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5945295999665417e-06,
+      "loss": 3.648,
+      "step": 9745
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5944877773363165e-06,
+      "loss": 1.1437,
+      "step": 9746
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5944459547060913e-06,
+      "loss": 0.0028,
+      "step": 9747
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594404132075866e-06,
+      "loss": 0.006,
+      "step": 9748
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5943623094456409e-06,
+      "loss": 0.0007,
+      "step": 9749
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5943204868154157e-06,
+      "loss": 0.4818,
+      "step": 9750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5942786641851907e-06,
+      "loss": 0.0005,
+      "step": 9751
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5942368415549654e-06,
+      "loss": 0.0003,
+      "step": 9752
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5941950189247402e-06,
+      "loss": 0.0026,
+      "step": 9753
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594153196294515e-06,
+      "loss": 0.0012,
+      "step": 9754
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5941113736642898e-06,
+      "loss": 2.4469,
+      "step": 9755
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5940695510340646e-06,
+      "loss": 0.0016,
+      "step": 9756
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5940277284038392e-06,
+      "loss": 0.0004,
+      "step": 9757
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593985905773614e-06,
+      "loss": 0.0027,
+      "step": 9758
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5939440831433887e-06,
+      "loss": 0.002,
+      "step": 9759
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5939022605131635e-06,
+      "loss": 0.0605,
+      "step": 9760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5938604378829383e-06,
+      "loss": 0.0045,
+      "step": 9761
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593818615252713e-06,
+      "loss": 0.0028,
+      "step": 9762
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5937767926224879e-06,
+      "loss": 1.3471,
+      "step": 9763
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5937349699922627e-06,
+      "loss": 0.0016,
+      "step": 9764
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5936931473620375e-06,
+      "loss": 0.0026,
+      "step": 9765
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5936513247318125e-06,
+      "loss": 0.0114,
+      "step": 9766
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5936095021015872e-06,
+      "loss": 0.0036,
+      "step": 9767
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593567679471362e-06,
+      "loss": 1.9868,
+      "step": 9768
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5935258568411368e-06,
+      "loss": 2.8214,
+      "step": 9769
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5934840342109114e-06,
+      "loss": 0.0562,
+      "step": 9770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5934422115806862e-06,
+      "loss": 0.0042,
+      "step": 9771
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593400388950461e-06,
+      "loss": 0.0004,
+      "step": 9772
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5933585663202357e-06,
+      "loss": 0.6866,
+      "step": 9773
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5933167436900105e-06,
+      "loss": 0.0007,
+      "step": 9774
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5932749210597853e-06,
+      "loss": 0.0007,
+      "step": 9775
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.59323309842956e-06,
+      "loss": 1.9192,
+      "step": 9776
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5931912757993349e-06,
+      "loss": 0.0007,
+      "step": 9777
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5931494531691097e-06,
+      "loss": 0.0009,
+      "step": 9778
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5931076305388845e-06,
+      "loss": 0.0007,
+      "step": 9779
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5930658079086595e-06,
+      "loss": 0.0005,
+      "step": 9780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5930239852784342e-06,
+      "loss": 0.0052,
+      "step": 9781
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592982162648209e-06,
+      "loss": 1.5649,
+      "step": 9782
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5929403400179838e-06,
+      "loss": 3.317,
+      "step": 9783
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5928985173877584e-06,
+      "loss": 3.4015,
+      "step": 9784
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5928566947575332e-06,
+      "loss": 0.0457,
+      "step": 9785
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592814872127308e-06,
+      "loss": 2.2852,
+      "step": 9786
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5927730494970828e-06,
+      "loss": 0.0031,
+      "step": 9787
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5927312268668575e-06,
+      "loss": 1.5837,
+      "step": 9788
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5926894042366323e-06,
+      "loss": 0.2343,
+      "step": 9789
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5926475816064071e-06,
+      "loss": 2.3991,
+      "step": 9790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592605758976182e-06,
+      "loss": 0.4239,
+      "step": 9791
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5925639363459567e-06,
+      "loss": 3.4168,
+      "step": 9792
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5925221137157315e-06,
+      "loss": 0.0026,
+      "step": 9793
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5924802910855063e-06,
+      "loss": 2.9937,
+      "step": 9794
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5924384684552813e-06,
+      "loss": 0.6964,
+      "step": 9795
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592396645825056e-06,
+      "loss": 0.0023,
+      "step": 9796
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5923548231948308e-06,
+      "loss": 0.166,
+      "step": 9797
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5923130005646054e-06,
+      "loss": 0.0268,
+      "step": 9798
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5922711779343802e-06,
+      "loss": 0.0021,
+      "step": 9799
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592229355304155e-06,
+      "loss": 1.4706,
+      "step": 9800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5921875326739298e-06,
+      "loss": 0.2958,
+      "step": 9801
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5921457100437045e-06,
+      "loss": 1.4444,
+      "step": 9802
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5921038874134793e-06,
+      "loss": 0.0011,
+      "step": 9803
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5920620647832541e-06,
+      "loss": 0.0059,
+      "step": 9804
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592020242153029e-06,
+      "loss": 0.0182,
+      "step": 9805
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5919784195228037e-06,
+      "loss": 0.0034,
+      "step": 9806
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5919365968925785e-06,
+      "loss": 0.0022,
+      "step": 9807
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5918947742623533e-06,
+      "loss": 0.0069,
+      "step": 9808
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5918529516321283e-06,
+      "loss": 0.0026,
+      "step": 9809
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591811129001903e-06,
+      "loss": 0.0099,
+      "step": 9810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5917693063716776e-06,
+      "loss": 0.0014,
+      "step": 9811
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5917274837414524e-06,
+      "loss": 0.0037,
+      "step": 9812
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5916856611112272e-06,
+      "loss": 0.0414,
+      "step": 9813
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.591643838481002e-06,
+      "loss": 0.0017,
+      "step": 9814
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5916020158507768e-06,
+      "loss": 0.0021,
+      "step": 9815
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5915601932205516e-06,
+      "loss": 2.5161,
+      "step": 9816
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5915183705903263e-06,
+      "loss": 0.0026,
+      "step": 9817
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5914765479601011e-06,
+      "loss": 2.0582,
+      "step": 9818
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.591434725329876e-06,
+      "loss": 0.0027,
+      "step": 9819
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5913929026996507e-06,
+      "loss": 0.0047,
+      "step": 9820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5913510800694255e-06,
+      "loss": 0.3607,
+      "step": 9821
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5913092574392003e-06,
+      "loss": 0.001,
+      "step": 9822
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.591267434808975e-06,
+      "loss": 0.0013,
+      "step": 9823
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.59122561217875e-06,
+      "loss": 0.0124,
+      "step": 9824
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5911837895485246e-06,
+      "loss": 0.0021,
+      "step": 9825
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5911419669182994e-06,
+      "loss": 0.7481,
+      "step": 9826
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5911001442880742e-06,
+      "loss": 0.007,
+      "step": 9827
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.591058321657849e-06,
+      "loss": 2.2008,
+      "step": 9828
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5910164990276238e-06,
+      "loss": 0.0045,
+      "step": 9829
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5909746763973986e-06,
+      "loss": 0.0077,
+      "step": 9830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5909328537671733e-06,
+      "loss": 1.1141,
+      "step": 9831
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5908910311369481e-06,
+      "loss": 0.0026,
+      "step": 9832
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.590849208506723e-06,
+      "loss": 0.1176,
+      "step": 9833
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5908073858764977e-06,
+      "loss": 0.4092,
+      "step": 9834
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5907655632462725e-06,
+      "loss": 0.0113,
+      "step": 9835
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5907237406160473e-06,
+      "loss": 0.011,
+      "step": 9836
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.590681917985822e-06,
+      "loss": 0.0116,
+      "step": 9837
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5906400953555966e-06,
+      "loss": 0.003,
+      "step": 9838
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5905982727253716e-06,
+      "loss": 0.0036,
+      "step": 9839
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5905564500951464e-06,
+      "loss": 0.0082,
+      "step": 9840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5905146274649212e-06,
+      "loss": 0.0029,
+      "step": 9841
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.590472804834696e-06,
+      "loss": 0.0073,
+      "step": 9842
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5904309822044708e-06,
+      "loss": 0.002,
+      "step": 9843
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5903891595742456e-06,
+      "loss": 0.0093,
+      "step": 9844
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5903473369440204e-06,
+      "loss": 0.0038,
+      "step": 9845
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5903055143137951e-06,
+      "loss": 0.0035,
+      "step": 9846
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.59026369168357e-06,
+      "loss": 0.0038,
+      "step": 9847
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5902218690533447e-06,
+      "loss": 0.0028,
+      "step": 9848
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5901800464231195e-06,
+      "loss": 0.0022,
+      "step": 9849
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5901382237928943e-06,
+      "loss": 0.0012,
+      "step": 9850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5900964011626689e-06,
+      "loss": 0.0124,
+      "step": 9851
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5900545785324437e-06,
+      "loss": 0.0009,
+      "step": 9852
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5900127559022186e-06,
+      "loss": 0.0008,
+      "step": 9853
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5899709332719934e-06,
+      "loss": 0.0014,
+      "step": 9854
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5899291106417682e-06,
+      "loss": 0.0014,
+      "step": 9855
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589887288011543e-06,
+      "loss": 1.9125,
+      "step": 9856
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5898454653813178e-06,
+      "loss": 0.0007,
+      "step": 9857
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5898036427510926e-06,
+      "loss": 0.009,
+      "step": 9858
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5897618201208674e-06,
+      "loss": 0.0031,
+      "step": 9859
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5897199974906422e-06,
+      "loss": 0.0052,
+      "step": 9860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589678174860417e-06,
+      "loss": 0.0031,
+      "step": 9861
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5896363522301917e-06,
+      "loss": 0.0028,
+      "step": 9862
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5895945295999665e-06,
+      "loss": 0.005,
+      "step": 9863
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5895527069697413e-06,
+      "loss": 0.0006,
+      "step": 9864
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5895108843395159e-06,
+      "loss": 0.0117,
+      "step": 9865
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5894690617092907e-06,
+      "loss": 0.0019,
+      "step": 9866
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5894272390790654e-06,
+      "loss": 0.0056,
+      "step": 9867
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5893854164488404e-06,
+      "loss": 0.5658,
+      "step": 9868
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5893435938186152e-06,
+      "loss": 0.0058,
+      "step": 9869
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.58930177118839e-06,
+      "loss": 0.0008,
+      "step": 9870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5892599485581648e-06,
+      "loss": 0.0017,
+      "step": 9871
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5892181259279396e-06,
+      "loss": 0.0257,
+      "step": 9872
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5891763032977144e-06,
+      "loss": 1.8213,
+      "step": 9873
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5891344806674892e-06,
+      "loss": 0.0012,
+      "step": 9874
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589092658037264e-06,
+      "loss": 0.0017,
+      "step": 9875
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5890508354070387e-06,
+      "loss": 0.0021,
+      "step": 9876
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5890090127768135e-06,
+      "loss": 2.3946,
+      "step": 9877
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588967190146588e-06,
+      "loss": 0.0099,
+      "step": 9878
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5889253675163629e-06,
+      "loss": 0.0043,
+      "step": 9879
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5888835448861377e-06,
+      "loss": 0.0014,
+      "step": 9880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5888417222559125e-06,
+      "loss": 0.0063,
+      "step": 9881
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5887998996256872e-06,
+      "loss": 0.001,
+      "step": 9882
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5887580769954622e-06,
+      "loss": 0.0045,
+      "step": 9883
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588716254365237e-06,
+      "loss": 2.7455,
+      "step": 9884
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5886744317350118e-06,
+      "loss": 1.5676,
+      "step": 9885
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5886326091047866e-06,
+      "loss": 0.0005,
+      "step": 9886
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5885907864745614e-06,
+      "loss": 0.0006,
+      "step": 9887
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5885489638443362e-06,
+      "loss": 0.0011,
+      "step": 9888
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588507141214111e-06,
+      "loss": 1.1154,
+      "step": 9889
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5884653185838857e-06,
+      "loss": 0.0017,
+      "step": 9890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5884234959536605e-06,
+      "loss": 0.0007,
+      "step": 9891
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588381673323435e-06,
+      "loss": 0.0007,
+      "step": 9892
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5883398506932099e-06,
+      "loss": 0.0005,
+      "step": 9893
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5882980280629847e-06,
+      "loss": 0.0004,
+      "step": 9894
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5882562054327595e-06,
+      "loss": 0.0021,
+      "step": 9895
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5882143828025342e-06,
+      "loss": 1.7719,
+      "step": 9896
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5881725601723092e-06,
+      "loss": 2.1531,
+      "step": 9897
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588130737542084e-06,
+      "loss": 0.0053,
+      "step": 9898
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5880889149118588e-06,
+      "loss": 0.0013,
+      "step": 9899
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5880470922816336e-06,
+      "loss": 0.0008,
+      "step": 9900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5880052696514084e-06,
+      "loss": 0.0062,
+      "step": 9901
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5879634470211832e-06,
+      "loss": 0.2049,
+      "step": 9902
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587921624390958e-06,
+      "loss": 0.0009,
+      "step": 9903
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5878798017607328e-06,
+      "loss": 1.3342,
+      "step": 9904
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5878379791305075e-06,
+      "loss": 0.3979,
+      "step": 9905
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5877961565002821e-06,
+      "loss": 0.0012,
+      "step": 9906
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587754333870057e-06,
+      "loss": 0.0022,
+      "step": 9907
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5877125112398317e-06,
+      "loss": 0.0042,
+      "step": 9908
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5876706886096065e-06,
+      "loss": 0.0006,
+      "step": 9909
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5876288659793813e-06,
+      "loss": 0.0005,
+      "step": 9910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587587043349156e-06,
+      "loss": 1.9615,
+      "step": 9911
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587545220718931e-06,
+      "loss": 1.3242,
+      "step": 9912
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5875033980887058e-06,
+      "loss": 2.2235,
+      "step": 9913
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5874615754584806e-06,
+      "loss": 1.2807,
+      "step": 9914
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5874197528282554e-06,
+      "loss": 0.0022,
+      "step": 9915
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5873779301980302e-06,
+      "loss": 0.0009,
+      "step": 9916
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587336107567805e-06,
+      "loss": 0.0158,
+      "step": 9917
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5872942849375798e-06,
+      "loss": 0.0031,
+      "step": 9918
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5872524623073543e-06,
+      "loss": 3.2589,
+      "step": 9919
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5872106396771291e-06,
+      "loss": 0.8969,
+      "step": 9920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587168817046904e-06,
+      "loss": 3.9404,
+      "step": 9921
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5871269944166787e-06,
+      "loss": 0.0022,
+      "step": 9922
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5870851717864535e-06,
+      "loss": 0.0005,
+      "step": 9923
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5870433491562283e-06,
+      "loss": 0.0019,
+      "step": 9924
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587001526526003e-06,
+      "loss": 0.6712,
+      "step": 9925
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5869597038957778e-06,
+      "loss": 0.8337,
+      "step": 9926
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5869178812655528e-06,
+      "loss": 0.6104,
+      "step": 9927
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5868760586353276e-06,
+      "loss": 0.0054,
+      "step": 9928
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5868342360051024e-06,
+      "loss": 0.0014,
+      "step": 9929
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5867924133748772e-06,
+      "loss": 0.0028,
+      "step": 9930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.586750590744652e-06,
+      "loss": 0.0018,
+      "step": 9931
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5867087681144268e-06,
+      "loss": 0.0023,
+      "step": 9932
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5866669454842013e-06,
+      "loss": 0.0091,
+      "step": 9933
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5866251228539761e-06,
+      "loss": 0.0041,
+      "step": 9934
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.586583300223751e-06,
+      "loss": 0.0048,
+      "step": 9935
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5865414775935257e-06,
+      "loss": 0.6982,
+      "step": 9936
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5864996549633005e-06,
+      "loss": 0.0027,
+      "step": 9937
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5864578323330753e-06,
+      "loss": 0.3644,
+      "step": 9938
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.58641600970285e-06,
+      "loss": 0.0031,
+      "step": 9939
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5863741870726248e-06,
+      "loss": 0.0017,
+      "step": 9940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5863323644423998e-06,
+      "loss": 0.0037,
+      "step": 9941
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5862905418121746e-06,
+      "loss": 1.0684,
+      "step": 9942
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5862487191819494e-06,
+      "loss": 0.9201,
+      "step": 9943
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5862068965517242e-06,
+      "loss": 0.0041,
+      "step": 9944
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.586165073921499e-06,
+      "loss": 0.0007,
+      "step": 9945
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5861232512912736e-06,
+      "loss": 0.0006,
+      "step": 9946
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5860814286610484e-06,
+      "loss": 0.0006,
+      "step": 9947
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5860396060308231e-06,
+      "loss": 0.0019,
+      "step": 9948
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585997783400598e-06,
+      "loss": 0.0024,
+      "step": 9949
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5859559607703727e-06,
+      "loss": 0.0008,
+      "step": 9950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5859141381401475e-06,
+      "loss": 0.001,
+      "step": 9951
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5858723155099223e-06,
+      "loss": 0.0117,
+      "step": 9952
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585830492879697e-06,
+      "loss": 1.8152,
+      "step": 9953
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5857886702494719e-06,
+      "loss": 0.0021,
+      "step": 9954
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5857468476192466e-06,
+      "loss": 0.0015,
+      "step": 9955
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5857050249890216e-06,
+      "loss": 0.0019,
+      "step": 9956
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5856632023587964e-06,
+      "loss": 0.0013,
+      "step": 9957
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5856213797285712e-06,
+      "loss": 0.001,
+      "step": 9958
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585579557098346e-06,
+      "loss": 0.4158,
+      "step": 9959
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5855377344681206e-06,
+      "loss": 0.002,
+      "step": 9960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5854959118378954e-06,
+      "loss": 0.0016,
+      "step": 9961
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5854540892076701e-06,
+      "loss": 0.0037,
+      "step": 9962
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585412266577445e-06,
+      "loss": 0.0009,
+      "step": 9963
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5853704439472197e-06,
+      "loss": 2.1476,
+      "step": 9964
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5853286213169945e-06,
+      "loss": 0.9464,
+      "step": 9965
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5852867986867693e-06,
+      "loss": 2.0858,
+      "step": 9966
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585244976056544e-06,
+      "loss": 0.0067,
+      "step": 9967
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5852031534263189e-06,
+      "loss": 0.0013,
+      "step": 9968
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5851613307960937e-06,
+      "loss": 0.0013,
+      "step": 9969
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5851195081658684e-06,
+      "loss": 0.0055,
+      "step": 9970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5850776855356434e-06,
+      "loss": 0.0031,
+      "step": 9971
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5850358629054182e-06,
+      "loss": 0.0025,
+      "step": 9972
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5849940402751928e-06,
+      "loss": 0.0029,
+      "step": 9973
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5849522176449676e-06,
+      "loss": 0.0032,
+      "step": 9974
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5849103950147424e-06,
+      "loss": 1.8481,
+      "step": 9975
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5848685723845172e-06,
+      "loss": 1.0069,
+      "step": 9976
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.584826749754292e-06,
+      "loss": 0.003,
+      "step": 9977
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5847849271240667e-06,
+      "loss": 0.1691,
+      "step": 9978
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5847431044938415e-06,
+      "loss": 1.171,
+      "step": 9979
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5847012818636163e-06,
+      "loss": 0.0025,
+      "step": 9980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.584659459233391e-06,
+      "loss": 0.003,
+      "step": 9981
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5846176366031659e-06,
+      "loss": 0.0129,
+      "step": 9982
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5845758139729407e-06,
+      "loss": 0.1473,
+      "step": 9983
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5845339913427154e-06,
+      "loss": 1.0712,
+      "step": 9984
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5844921687124904e-06,
+      "loss": 0.0096,
+      "step": 9985
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5844503460822652e-06,
+      "loss": 0.7045,
+      "step": 9986
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5844085234520398e-06,
+      "loss": 0.2587,
+      "step": 9987
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5843667008218146e-06,
+      "loss": 0.001,
+      "step": 9988
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5843248781915894e-06,
+      "loss": 0.0014,
+      "step": 9989
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5842830555613642e-06,
+      "loss": 0.0156,
+      "step": 9990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.584241232931139e-06,
+      "loss": 0.0006,
+      "step": 9991
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5841994103009137e-06,
+      "loss": 0.0009,
+      "step": 9992
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5841575876706885e-06,
+      "loss": 0.0054,
+      "step": 9993
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5841157650404633e-06,
+      "loss": 0.0347,
+      "step": 9994
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.584073942410238e-06,
+      "loss": 0.0006,
+      "step": 9995
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5840321197800129e-06,
+      "loss": 0.0027,
+      "step": 9996
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5839902971497877e-06,
+      "loss": 0.005,
+      "step": 9997
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5839484745195625e-06,
+      "loss": 0.0049,
+      "step": 9998
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5839066518893372e-06,
+      "loss": 1.4688,
+      "step": 9999
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5838648292591122e-06,
+      "loss": 0.0036,
+      "step": 10000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5838230066288868e-06,
+      "loss": 0.0013,
+      "step": 10001
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5837811839986616e-06,
+      "loss": 0.3791,
+      "step": 10002
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5837393613684364e-06,
+      "loss": 0.6395,
+      "step": 10003
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5836975387382112e-06,
+      "loss": 0.001,
+      "step": 10004
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583655716107986e-06,
+      "loss": 0.0462,
+      "step": 10005
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5836138934777607e-06,
+      "loss": 0.002,
+      "step": 10006
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5835720708475355e-06,
+      "loss": 0.0302,
+      "step": 10007
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5835302482173103e-06,
+      "loss": 0.0026,
+      "step": 10008
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583488425587085e-06,
+      "loss": 0.0016,
+      "step": 10009
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5834466029568599e-06,
+      "loss": 0.0052,
+      "step": 10010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5834047803266347e-06,
+      "loss": 0.0775,
+      "step": 10011
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5833629576964095e-06,
+      "loss": 0.2689,
+      "step": 10012
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583321135066184e-06,
+      "loss": 0.0005,
+      "step": 10013
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583279312435959e-06,
+      "loss": 0.0008,
+      "step": 10014
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5832374898057338e-06,
+      "loss": 0.0009,
+      "step": 10015
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5831956671755086e-06,
+      "loss": 0.001,
+      "step": 10016
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5831538445452834e-06,
+      "loss": 0.0007,
+      "step": 10017
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5831120219150582e-06,
+      "loss": 0.0027,
+      "step": 10018
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583070199284833e-06,
+      "loss": 0.0011,
+      "step": 10019
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5830283766546078e-06,
+      "loss": 0.0014,
+      "step": 10020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5829865540243825e-06,
+      "loss": 0.0005,
+      "step": 10021
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5829447313941573e-06,
+      "loss": 0.0072,
+      "step": 10022
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5829029087639321e-06,
+      "loss": 0.0012,
+      "step": 10023
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582861086133707e-06,
+      "loss": 0.0007,
+      "step": 10024
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5828192635034817e-06,
+      "loss": 0.016,
+      "step": 10025
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5827774408732565e-06,
+      "loss": 0.0015,
+      "step": 10026
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582735618243031e-06,
+      "loss": 0.8608,
+      "step": 10027
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5826937956128058e-06,
+      "loss": 2.5225,
+      "step": 10028
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5826519729825808e-06,
+      "loss": 0.6919,
+      "step": 10029
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5826101503523556e-06,
+      "loss": 0.0005,
+      "step": 10030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5825683277221304e-06,
+      "loss": 0.0016,
+      "step": 10031
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5825265050919052e-06,
+      "loss": 2.8147,
+      "step": 10032
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.58248468246168e-06,
+      "loss": 0.0234,
+      "step": 10033
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5824428598314548e-06,
+      "loss": 0.0341,
+      "step": 10034
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5824010372012295e-06,
+      "loss": 0.0724,
+      "step": 10035
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5823592145710043e-06,
+      "loss": 0.001,
+      "step": 10036
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5823173919407791e-06,
+      "loss": 0.0037,
+      "step": 10037
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582275569310554e-06,
+      "loss": 0.4369,
+      "step": 10038
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5822337466803287e-06,
+      "loss": 0.0103,
+      "step": 10039
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5821919240501035e-06,
+      "loss": 0.002,
+      "step": 10040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582150101419878e-06,
+      "loss": 0.0009,
+      "step": 10041
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5821082787896528e-06,
+      "loss": 2.0249,
+      "step": 10042
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5820664561594276e-06,
+      "loss": 0.0008,
+      "step": 10043
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5820246335292026e-06,
+      "loss": 0.011,
+      "step": 10044
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5819828108989774e-06,
+      "loss": 0.0011,
+      "step": 10045
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5819409882687522e-06,
+      "loss": 0.0005,
+      "step": 10046
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581899165638527e-06,
+      "loss": 0.5092,
+      "step": 10047
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5818573430083018e-06,
+      "loss": 0.0017,
+      "step": 10048
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5818155203780766e-06,
+      "loss": 0.0132,
+      "step": 10049
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5817736977478513e-06,
+      "loss": 0.0046,
+      "step": 10050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5817318751176261e-06,
+      "loss": 3.0432,
+      "step": 10051
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581690052487401e-06,
+      "loss": 1.1731,
+      "step": 10052
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5816482298571757e-06,
+      "loss": 0.0005,
+      "step": 10053
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5816064072269503e-06,
+      "loss": 0.0009,
+      "step": 10054
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581564584596725e-06,
+      "loss": 0.0009,
+      "step": 10055
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5815227619664999e-06,
+      "loss": 0.0022,
+      "step": 10056
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5814809393362746e-06,
+      "loss": 0.0005,
+      "step": 10057
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5814391167060496e-06,
+      "loss": 0.0079,
+      "step": 10058
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5813972940758244e-06,
+      "loss": 0.0008,
+      "step": 10059
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5813554714455992e-06,
+      "loss": 0.006,
+      "step": 10060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581313648815374e-06,
+      "loss": 0.0009,
+      "step": 10061
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5812718261851488e-06,
+      "loss": 0.0014,
+      "step": 10062
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5812300035549236e-06,
+      "loss": 0.0026,
+      "step": 10063
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5811881809246984e-06,
+      "loss": 0.7573,
+      "step": 10064
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5811463582944731e-06,
+      "loss": 0.0016,
+      "step": 10065
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581104535664248e-06,
+      "loss": 0.0054,
+      "step": 10066
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5810627130340227e-06,
+      "loss": 0.0013,
+      "step": 10067
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5810208904037973e-06,
+      "loss": 0.0033,
+      "step": 10068
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580979067773572e-06,
+      "loss": 0.0013,
+      "step": 10069
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5809372451433469e-06,
+      "loss": 2.003,
+      "step": 10070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5808954225131216e-06,
+      "loss": 0.0007,
+      "step": 10071
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5808535998828964e-06,
+      "loss": 0.0062,
+      "step": 10072
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5808117772526714e-06,
+      "loss": 0.0004,
+      "step": 10073
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5807699546224462e-06,
+      "loss": 0.0008,
+      "step": 10074
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580728131992221e-06,
+      "loss": 0.0224,
+      "step": 10075
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5806863093619958e-06,
+      "loss": 1.0787,
+      "step": 10076
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5806444867317706e-06,
+      "loss": 0.5159,
+      "step": 10077
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5806026641015454e-06,
+      "loss": 2.3444,
+      "step": 10078
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5805608414713201e-06,
+      "loss": 0.0005,
+      "step": 10079
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580519018841095e-06,
+      "loss": 0.0011,
+      "step": 10080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5804771962108695e-06,
+      "loss": 0.2058,
+      "step": 10081
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5804353735806443e-06,
+      "loss": 0.1066,
+      "step": 10082
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580393550950419e-06,
+      "loss": 0.0009,
+      "step": 10083
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5803517283201939e-06,
+      "loss": 2.6984,
+      "step": 10084
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5803099056899687e-06,
+      "loss": 0.0086,
+      "step": 10085
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5802680830597434e-06,
+      "loss": 0.0076,
+      "step": 10086
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5802262604295182e-06,
+      "loss": 0.0007,
+      "step": 10087
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5801844377992932e-06,
+      "loss": 0.002,
+      "step": 10088
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580142615169068e-06,
+      "loss": 0.002,
+      "step": 10089
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5801007925388428e-06,
+      "loss": 0.0021,
+      "step": 10090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5800589699086176e-06,
+      "loss": 1.79,
+      "step": 10091
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5800171472783924e-06,
+      "loss": 0.0042,
+      "step": 10092
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5799753246481672e-06,
+      "loss": 0.001,
+      "step": 10093
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579933502017942e-06,
+      "loss": 0.0005,
+      "step": 10094
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5798916793877165e-06,
+      "loss": 0.05,
+      "step": 10095
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5798498567574913e-06,
+      "loss": 0.0142,
+      "step": 10096
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579808034127266e-06,
+      "loss": 0.5989,
+      "step": 10097
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5797662114970409e-06,
+      "loss": 0.0013,
+      "step": 10098
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5797243888668157e-06,
+      "loss": 0.0009,
+      "step": 10099
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5796825662365904e-06,
+      "loss": 0.111,
+      "step": 10100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5796407436063652e-06,
+      "loss": 0.0006,
+      "step": 10101
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5795989209761402e-06,
+      "loss": 2.7413,
+      "step": 10102
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579557098345915e-06,
+      "loss": 0.0007,
+      "step": 10103
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5795152757156898e-06,
+      "loss": 0.0012,
+      "step": 10104
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5794734530854646e-06,
+      "loss": 0.0131,
+      "step": 10105
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5794316304552394e-06,
+      "loss": 0.0014,
+      "step": 10106
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5793898078250142e-06,
+      "loss": 0.0007,
+      "step": 10107
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579347985194789e-06,
+      "loss": 1.3972,
+      "step": 10108
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5793061625645635e-06,
+      "loss": 1.9165,
+      "step": 10109
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5792643399343383e-06,
+      "loss": 1.1985,
+      "step": 10110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579222517304113e-06,
+      "loss": 0.0014,
+      "step": 10111
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5791806946738879e-06,
+      "loss": 0.0022,
+      "step": 10112
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5791388720436627e-06,
+      "loss": 0.0008,
+      "step": 10113
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5790970494134375e-06,
+      "loss": 0.0017,
+      "step": 10114
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5790552267832122e-06,
+      "loss": 0.0007,
+      "step": 10115
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579013404152987e-06,
+      "loss": 0.0013,
+      "step": 10116
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578971581522762e-06,
+      "loss": 0.0021,
+      "step": 10117
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5789297588925368e-06,
+      "loss": 0.9561,
+      "step": 10118
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5788879362623116e-06,
+      "loss": 0.0006,
+      "step": 10119
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5788461136320864e-06,
+      "loss": 0.0036,
+      "step": 10120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5788042910018612e-06,
+      "loss": 0.0005,
+      "step": 10121
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5787624683716357e-06,
+      "loss": 0.0009,
+      "step": 10122
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5787206457414105e-06,
+      "loss": 0.002,
+      "step": 10123
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5786788231111853e-06,
+      "loss": 0.0071,
+      "step": 10124
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5786370004809601e-06,
+      "loss": 1.8902,
+      "step": 10125
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5785951778507349e-06,
+      "loss": 0.0014,
+      "step": 10126
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5785533552205097e-06,
+      "loss": 0.0009,
+      "step": 10127
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5785115325902845e-06,
+      "loss": 0.0141,
+      "step": 10128
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5784697099600593e-06,
+      "loss": 0.0019,
+      "step": 10129
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578427887329834e-06,
+      "loss": 0.0015,
+      "step": 10130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5783860646996088e-06,
+      "loss": 0.2081,
+      "step": 10131
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5783442420693838e-06,
+      "loss": 0.0005,
+      "step": 10132
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5783024194391586e-06,
+      "loss": 0.0016,
+      "step": 10133
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5782605968089334e-06,
+      "loss": 0.0017,
+      "step": 10134
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5782187741787082e-06,
+      "loss": 0.0177,
+      "step": 10135
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5781769515484828e-06,
+      "loss": 0.0016,
+      "step": 10136
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5781351289182575e-06,
+      "loss": 0.0017,
+      "step": 10137
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5780933062880323e-06,
+      "loss": 0.0014,
+      "step": 10138
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5780514836578071e-06,
+      "loss": 0.8215,
+      "step": 10139
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.578009661027582e-06,
+      "loss": 0.0004,
+      "step": 10140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5779678383973567e-06,
+      "loss": 0.0015,
+      "step": 10141
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5779260157671315e-06,
+      "loss": 0.329,
+      "step": 10142
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5778841931369063e-06,
+      "loss": 0.0009,
+      "step": 10143
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.577842370506681e-06,
+      "loss": 0.5718,
+      "step": 10144
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5778005478764558e-06,
+      "loss": 0.0008,
+      "step": 10145
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5777587252462308e-06,
+      "loss": 0.002,
+      "step": 10146
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5777169026160056e-06,
+      "loss": 0.301,
+      "step": 10147
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5776750799857804e-06,
+      "loss": 0.781,
+      "step": 10148
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.577633257355555e-06,
+      "loss": 0.0022,
+      "step": 10149
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5775914347253298e-06,
+      "loss": 0.0005,
+      "step": 10150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5775496120951046e-06,
+      "loss": 0.0031,
+      "step": 10151
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5775077894648793e-06,
+      "loss": 0.0005,
+      "step": 10152
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5774659668346541e-06,
+      "loss": 0.0029,
+      "step": 10153
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.577424144204429e-06,
+      "loss": 0.0029,
+      "step": 10154
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5773823215742037e-06,
+      "loss": 0.482,
+      "step": 10155
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5773404989439785e-06,
+      "loss": 0.0006,
+      "step": 10156
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5772986763137533e-06,
+      "loss": 0.0174,
+      "step": 10157
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.577256853683528e-06,
+      "loss": 1.5132,
+      "step": 10158
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5772150310533028e-06,
+      "loss": 0.0017,
+      "step": 10159
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5771732084230776e-06,
+      "loss": 0.7013,
+      "step": 10160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5771313857928526e-06,
+      "loss": 0.0041,
+      "step": 10161
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5770895631626274e-06,
+      "loss": 0.0006,
+      "step": 10162
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.577047740532402e-06,
+      "loss": 0.0042,
+      "step": 10163
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5770059179021768e-06,
+      "loss": 1.7439,
+      "step": 10164
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5769640952719516e-06,
+      "loss": 0.0021,
+      "step": 10165
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5769222726417263e-06,
+      "loss": 0.0005,
+      "step": 10166
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5768804500115011e-06,
+      "loss": 0.0032,
+      "step": 10167
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576838627381276e-06,
+      "loss": 0.0022,
+      "step": 10168
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5767968047510507e-06,
+      "loss": 2.2044,
+      "step": 10169
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5767549821208255e-06,
+      "loss": 0.0006,
+      "step": 10170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5767131594906003e-06,
+      "loss": 0.0014,
+      "step": 10171
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576671336860375e-06,
+      "loss": 0.0006,
+      "step": 10172
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5766295142301499e-06,
+      "loss": 0.002,
+      "step": 10173
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5765876915999246e-06,
+      "loss": 0.0699,
+      "step": 10174
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5765458689696996e-06,
+      "loss": 0.0042,
+      "step": 10175
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5765040463394742e-06,
+      "loss": 0.0004,
+      "step": 10176
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576462223709249e-06,
+      "loss": 0.0005,
+      "step": 10177
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5764204010790238e-06,
+      "loss": 0.0021,
+      "step": 10178
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5763785784487986e-06,
+      "loss": 0.0006,
+      "step": 10179
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5763367558185734e-06,
+      "loss": 0.0007,
+      "step": 10180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5762949331883481e-06,
+      "loss": 1.5599,
+      "step": 10181
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576253110558123e-06,
+      "loss": 0.0007,
+      "step": 10182
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5762112879278977e-06,
+      "loss": 0.0126,
+      "step": 10183
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5761694652976725e-06,
+      "loss": 0.001,
+      "step": 10184
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5761276426674473e-06,
+      "loss": 0.2072,
+      "step": 10185
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576085820037222e-06,
+      "loss": 0.2237,
+      "step": 10186
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5760439974069969e-06,
+      "loss": 0.0015,
+      "step": 10187
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5760021747767716e-06,
+      "loss": 0.0008,
+      "step": 10188
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5759603521465462e-06,
+      "loss": 2.2739,
+      "step": 10189
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5759185295163212e-06,
+      "loss": 0.0007,
+      "step": 10190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575876706886096e-06,
+      "loss": 0.7078,
+      "step": 10191
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5758348842558708e-06,
+      "loss": 0.0005,
+      "step": 10192
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5757930616256456e-06,
+      "loss": 0.0025,
+      "step": 10193
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5757512389954204e-06,
+      "loss": 0.0007,
+      "step": 10194
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5757094163651952e-06,
+      "loss": 2.4231,
+      "step": 10195
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.57566759373497e-06,
+      "loss": 0.0005,
+      "step": 10196
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5756257711047447e-06,
+      "loss": 0.0011,
+      "step": 10197
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5755839484745195e-06,
+      "loss": 0.0007,
+      "step": 10198
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5755421258442943e-06,
+      "loss": 0.001,
+      "step": 10199
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575500303214069e-06,
+      "loss": 0.0025,
+      "step": 10200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5754584805838439e-06,
+      "loss": 0.009,
+      "step": 10201
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5754166579536187e-06,
+      "loss": 0.0018,
+      "step": 10202
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5753748353233932e-06,
+      "loss": 0.001,
+      "step": 10203
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575333012693168e-06,
+      "loss": 0.1445,
+      "step": 10204
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575291190062943e-06,
+      "loss": 0.0007,
+      "step": 10205
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5752493674327178e-06,
+      "loss": 3.3586,
+      "step": 10206
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5752075448024926e-06,
+      "loss": 0.0155,
+      "step": 10207
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5751657221722674e-06,
+      "loss": 0.0035,
+      "step": 10208
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5751238995420422e-06,
+      "loss": 0.4535,
+      "step": 10209
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575082076911817e-06,
+      "loss": 0.0014,
+      "step": 10210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5750402542815917e-06,
+      "loss": 0.0064,
+      "step": 10211
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5749984316513665e-06,
+      "loss": 0.0012,
+      "step": 10212
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5749566090211413e-06,
+      "loss": 0.0067,
+      "step": 10213
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574914786390916e-06,
+      "loss": 1.9168,
+      "step": 10214
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5748729637606909e-06,
+      "loss": 0.0281,
+      "step": 10215
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5748311411304655e-06,
+      "loss": 0.002,
+      "step": 10216
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5747893185002402e-06,
+      "loss": 0.0008,
+      "step": 10217
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574747495870015e-06,
+      "loss": 0.0011,
+      "step": 10218
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.57470567323979e-06,
+      "loss": 0.0026,
+      "step": 10219
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5746638506095648e-06,
+      "loss": 0.0017,
+      "step": 10220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5746220279793396e-06,
+      "loss": 1.4639,
+      "step": 10221
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5745802053491144e-06,
+      "loss": 0.0008,
+      "step": 10222
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5745383827188892e-06,
+      "loss": 0.0014,
+      "step": 10223
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574496560088664e-06,
+      "loss": 0.2163,
+      "step": 10224
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5744547374584387e-06,
+      "loss": 0.0021,
+      "step": 10225
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5744129148282135e-06,
+      "loss": 0.021,
+      "step": 10226
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5743710921979883e-06,
+      "loss": 0.0008,
+      "step": 10227
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574329269567763e-06,
+      "loss": 0.0019,
+      "step": 10228
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5742874469375379e-06,
+      "loss": 1.9377,
+      "step": 10229
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5742456243073125e-06,
+      "loss": 2.1974,
+      "step": 10230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5742038016770872e-06,
+      "loss": 0.0014,
+      "step": 10231
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574161979046862e-06,
+      "loss": 0.001,
+      "step": 10232
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5741201564166368e-06,
+      "loss": 0.0021,
+      "step": 10233
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5740783337864118e-06,
+      "loss": 0.0046,
+      "step": 10234
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5740365111561866e-06,
+      "loss": 0.1279,
+      "step": 10235
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5739946885259614e-06,
+      "loss": 0.0012,
+      "step": 10236
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5739528658957362e-06,
+      "loss": 0.0029,
+      "step": 10237
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.573911043265511e-06,
+      "loss": 0.0017,
+      "step": 10238
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5738692206352857e-06,
+      "loss": 0.0007,
+      "step": 10239
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5738273980050605e-06,
+      "loss": 0.0018,
+      "step": 10240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5737855753748353e-06,
+      "loss": 1.095,
+      "step": 10241
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5737437527446101e-06,
+      "loss": 3.2213,
+      "step": 10242
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5737019301143849e-06,
+      "loss": 0.001,
+      "step": 10243
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5736601074841595e-06,
+      "loss": 0.005,
+      "step": 10244
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5736182848539343e-06,
+      "loss": 0.001,
+      "step": 10245
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.573576462223709e-06,
+      "loss": 0.0009,
+      "step": 10246
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5735346395934838e-06,
+      "loss": 0.0019,
+      "step": 10247
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5734928169632586e-06,
+      "loss": 0.0028,
+      "step": 10248
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5734509943330336e-06,
+      "loss": 0.2571,
+      "step": 10249
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5734091717028084e-06,
+      "loss": 0.3109,
+      "step": 10250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5733673490725832e-06,
+      "loss": 0.002,
+      "step": 10251
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.573325526442358e-06,
+      "loss": 0.2675,
+      "step": 10252
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5732837038121328e-06,
+      "loss": 0.0052,
+      "step": 10253
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5732418811819075e-06,
+      "loss": 1.4336,
+      "step": 10254
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5732000585516823e-06,
+      "loss": 0.002,
+      "step": 10255
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5731582359214571e-06,
+      "loss": 0.026,
+      "step": 10256
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5731164132912317e-06,
+      "loss": 0.0015,
+      "step": 10257
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5730745906610065e-06,
+      "loss": 0.0009,
+      "step": 10258
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5730327680307813e-06,
+      "loss": 0.0043,
+      "step": 10259
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572990945400556e-06,
+      "loss": 0.0129,
+      "step": 10260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5729491227703308e-06,
+      "loss": 0.0007,
+      "step": 10261
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5729073001401056e-06,
+      "loss": 0.0008,
+      "step": 10262
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5728654775098806e-06,
+      "loss": 0.0011,
+      "step": 10263
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5728236548796554e-06,
+      "loss": 0.0039,
+      "step": 10264
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5727818322494302e-06,
+      "loss": 0.0675,
+      "step": 10265
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572740009619205e-06,
+      "loss": 0.0122,
+      "step": 10266
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5726981869889798e-06,
+      "loss": 0.0032,
+      "step": 10267
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5726563643587546e-06,
+      "loss": 0.0084,
+      "step": 10268
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5726145417285293e-06,
+      "loss": 0.0006,
+      "step": 10269
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5725727190983041e-06,
+      "loss": 0.0009,
+      "step": 10270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5725308964680787e-06,
+      "loss": 0.0008,
+      "step": 10271
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5724890738378535e-06,
+      "loss": 0.255,
+      "step": 10272
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5724472512076283e-06,
+      "loss": 0.0006,
+      "step": 10273
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572405428577403e-06,
+      "loss": 1.7427,
+      "step": 10274
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5723636059471778e-06,
+      "loss": 0.0006,
+      "step": 10275
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5723217833169526e-06,
+      "loss": 0.0018,
+      "step": 10276
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5722799606867274e-06,
+      "loss": 0.01,
+      "step": 10277
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5722381380565024e-06,
+      "loss": 0.0006,
+      "step": 10278
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5721963154262772e-06,
+      "loss": 0.0049,
+      "step": 10279
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572154492796052e-06,
+      "loss": 0.0005,
+      "step": 10280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5721126701658268e-06,
+      "loss": 3.001,
+      "step": 10281
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5720708475356016e-06,
+      "loss": 1.9625,
+      "step": 10282
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5720290249053763e-06,
+      "loss": 0.0015,
+      "step": 10283
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571987202275151e-06,
+      "loss": 2.1404,
+      "step": 10284
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5719453796449257e-06,
+      "loss": 0.0022,
+      "step": 10285
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5719035570147005e-06,
+      "loss": 0.0013,
+      "step": 10286
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5718617343844753e-06,
+      "loss": 0.0012,
+      "step": 10287
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.57181991175425e-06,
+      "loss": 0.0356,
+      "step": 10288
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5717780891240249e-06,
+      "loss": 0.0422,
+      "step": 10289
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5717362664937996e-06,
+      "loss": 0.0048,
+      "step": 10290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5716944438635744e-06,
+      "loss": 0.0009,
+      "step": 10291
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5716526212333492e-06,
+      "loss": 0.0005,
+      "step": 10292
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5716107986031242e-06,
+      "loss": 0.0808,
+      "step": 10293
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.571568975972899e-06,
+      "loss": 0.6379,
+      "step": 10294
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5715271533426738e-06,
+      "loss": 0.0012,
+      "step": 10295
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5714853307124486e-06,
+      "loss": 0.0096,
+      "step": 10296
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5714435080822234e-06,
+      "loss": 2.2622,
+      "step": 10297
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.571401685451998e-06,
+      "loss": 0.0008,
+      "step": 10298
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5713598628217727e-06,
+      "loss": 0.0007,
+      "step": 10299
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5713180401915475e-06,
+      "loss": 0.001,
+      "step": 10300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5712762175613223e-06,
+      "loss": 0.0005,
+      "step": 10301
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.571234394931097e-06,
+      "loss": 1.3225,
+      "step": 10302
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5711925723008719e-06,
+      "loss": 0.541,
+      "step": 10303
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5711507496706466e-06,
+      "loss": 0.3649,
+      "step": 10304
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5711089270404214e-06,
+      "loss": 0.8614,
+      "step": 10305
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5710671044101962e-06,
+      "loss": 0.0018,
+      "step": 10306
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5710252817799712e-06,
+      "loss": 0.994,
+      "step": 10307
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.570983459149746e-06,
+      "loss": 2.3454,
+      "step": 10308
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5709416365195208e-06,
+      "loss": 3.2684,
+      "step": 10309
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5708998138892956e-06,
+      "loss": 0.0007,
+      "step": 10310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5708579912590704e-06,
+      "loss": 1.6856,
+      "step": 10311
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.570816168628845e-06,
+      "loss": 0.8146,
+      "step": 10312
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5707743459986197e-06,
+      "loss": 0.0009,
+      "step": 10313
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5707325233683945e-06,
+      "loss": 0.0006,
+      "step": 10314
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5706907007381693e-06,
+      "loss": 0.0064,
+      "step": 10315
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.570648878107944e-06,
+      "loss": 0.0012,
+      "step": 10316
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5706070554777189e-06,
+      "loss": 1.6473,
+      "step": 10317
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5705652328474937e-06,
+      "loss": 1.3542,
+      "step": 10318
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5705234102172684e-06,
+      "loss": 0.011,
+      "step": 10319
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5704815875870432e-06,
+      "loss": 0.5898,
+      "step": 10320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.570439764956818e-06,
+      "loss": 0.0613,
+      "step": 10321
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.570397942326593e-06,
+      "loss": 1.5247,
+      "step": 10322
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5703561196963678e-06,
+      "loss": 0.0013,
+      "step": 10323
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5703142970661426e-06,
+      "loss": 0.193,
+      "step": 10324
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5702724744359172e-06,
+      "loss": 0.0014,
+      "step": 10325
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.570230651805692e-06,
+      "loss": 0.6157,
+      "step": 10326
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5701888291754667e-06,
+      "loss": 0.0023,
+      "step": 10327
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5701470065452415e-06,
+      "loss": 1.4915,
+      "step": 10328
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5701051839150163e-06,
+      "loss": 2.8768,
+      "step": 10329
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.570063361284791e-06,
+      "loss": 0.0011,
+      "step": 10330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5700215386545659e-06,
+      "loss": 0.0293,
+      "step": 10331
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5699797160243407e-06,
+      "loss": 0.0012,
+      "step": 10332
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5699378933941155e-06,
+      "loss": 0.0715,
+      "step": 10333
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5698960707638902e-06,
+      "loss": 0.003,
+      "step": 10334
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569854248133665e-06,
+      "loss": 0.0024,
+      "step": 10335
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5698124255034398e-06,
+      "loss": 0.0044,
+      "step": 10336
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5697706028732148e-06,
+      "loss": 0.0008,
+      "step": 10337
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5697287802429896e-06,
+      "loss": 0.0007,
+      "step": 10338
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5696869576127642e-06,
+      "loss": 2.4755,
+      "step": 10339
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569645134982539e-06,
+      "loss": 0.0012,
+      "step": 10340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5696033123523137e-06,
+      "loss": 0.017,
+      "step": 10341
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5695614897220885e-06,
+      "loss": 0.0011,
+      "step": 10342
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5695196670918633e-06,
+      "loss": 0.0009,
+      "step": 10343
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569477844461638e-06,
+      "loss": 0.0069,
+      "step": 10344
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5694360218314129e-06,
+      "loss": 0.019,
+      "step": 10345
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5693941992011877e-06,
+      "loss": 0.0007,
+      "step": 10346
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5693523765709625e-06,
+      "loss": 0.011,
+      "step": 10347
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5693105539407372e-06,
+      "loss": 1.8353,
+      "step": 10348
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569268731310512e-06,
+      "loss": 0.004,
+      "step": 10349
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5692269086802868e-06,
+      "loss": 0.002,
+      "step": 10350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5691850860500618e-06,
+      "loss": 0.0012,
+      "step": 10351
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5691432634198364e-06,
+      "loss": 0.0012,
+      "step": 10352
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5691014407896112e-06,
+      "loss": 0.0123,
+      "step": 10353
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569059618159386e-06,
+      "loss": 0.0021,
+      "step": 10354
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5690177955291608e-06,
+      "loss": 0.0012,
+      "step": 10355
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5689759728989355e-06,
+      "loss": 0.0035,
+      "step": 10356
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5689341502687103e-06,
+      "loss": 0.0245,
+      "step": 10357
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5688923276384851e-06,
+      "loss": 0.6314,
+      "step": 10358
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.56885050500826e-06,
+      "loss": 0.0717,
+      "step": 10359
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5688086823780347e-06,
+      "loss": 0.0018,
+      "step": 10360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5687668597478095e-06,
+      "loss": 1.5936,
+      "step": 10361
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5687250371175843e-06,
+      "loss": 0.0177,
+      "step": 10362
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568683214487359e-06,
+      "loss": 0.0111,
+      "step": 10363
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5686413918571338e-06,
+      "loss": 0.0047,
+      "step": 10364
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5685995692269084e-06,
+      "loss": 0.0613,
+      "step": 10365
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5685577465966834e-06,
+      "loss": 0.7494,
+      "step": 10366
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5685159239664582e-06,
+      "loss": 2.6506,
+      "step": 10367
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568474101336233e-06,
+      "loss": 0.0103,
+      "step": 10368
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5684322787060078e-06,
+      "loss": 0.0009,
+      "step": 10369
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5683904560757825e-06,
+      "loss": 0.0025,
+      "step": 10370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5683486334455573e-06,
+      "loss": 0.0012,
+      "step": 10371
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5683068108153321e-06,
+      "loss": 0.0014,
+      "step": 10372
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568264988185107e-06,
+      "loss": 0.0027,
+      "step": 10373
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5682231655548817e-06,
+      "loss": 0.0036,
+      "step": 10374
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5681813429246565e-06,
+      "loss": 1.3825,
+      "step": 10375
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5681395202944313e-06,
+      "loss": 0.0497,
+      "step": 10376
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568097697664206e-06,
+      "loss": 2.5301,
+      "step": 10377
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5680558750339808e-06,
+      "loss": 1.6335,
+      "step": 10378
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5680140524037554e-06,
+      "loss": 0.0124,
+      "step": 10379
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5679722297735304e-06,
+      "loss": 0.0017,
+      "step": 10380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5679304071433052e-06,
+      "loss": 0.6644,
+      "step": 10381
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.56788858451308e-06,
+      "loss": 0.5175,
+      "step": 10382
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5678467618828548e-06,
+      "loss": 2.3922,
+      "step": 10383
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5678049392526296e-06,
+      "loss": 0.0079,
+      "step": 10384
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5677631166224043e-06,
+      "loss": 0.0091,
+      "step": 10385
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5677212939921791e-06,
+      "loss": 0.0053,
+      "step": 10386
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567679471361954e-06,
+      "loss": 0.0073,
+      "step": 10387
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5676376487317287e-06,
+      "loss": 0.0915,
+      "step": 10388
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5675958261015035e-06,
+      "loss": 0.0011,
+      "step": 10389
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5675540034712783e-06,
+      "loss": 0.177,
+      "step": 10390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567512180841053e-06,
+      "loss": 0.0021,
+      "step": 10391
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5674703582108276e-06,
+      "loss": 0.0028,
+      "step": 10392
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5674285355806024e-06,
+      "loss": 0.0228,
+      "step": 10393
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5673867129503772e-06,
+      "loss": 0.0033,
+      "step": 10394
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5673448903201522e-06,
+      "loss": 0.2756,
+      "step": 10395
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567303067689927e-06,
+      "loss": 0.001,
+      "step": 10396
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5672612450597018e-06,
+      "loss": 0.0051,
+      "step": 10397
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5672194224294766e-06,
+      "loss": 0.005,
+      "step": 10398
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5671775997992513e-06,
+      "loss": 0.2325,
+      "step": 10399
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5671357771690261e-06,
+      "loss": 0.0023,
+      "step": 10400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567093954538801e-06,
+      "loss": 0.0017,
+      "step": 10401
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5670521319085757e-06,
+      "loss": 0.005,
+      "step": 10402
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5670103092783505e-06,
+      "loss": 0.0056,
+      "step": 10403
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5669684866481253e-06,
+      "loss": 0.0283,
+      "step": 10404
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5669266640179e-06,
+      "loss": 0.0164,
+      "step": 10405
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5668848413876746e-06,
+      "loss": 0.0039,
+      "step": 10406
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5668430187574494e-06,
+      "loss": 0.0031,
+      "step": 10407
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5668011961272242e-06,
+      "loss": 0.4964,
+      "step": 10408
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566759373496999e-06,
+      "loss": 0.0035,
+      "step": 10409
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566717550866774e-06,
+      "loss": 0.0009,
+      "step": 10410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5666757282365488e-06,
+      "loss": 0.0062,
+      "step": 10411
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5666339056063236e-06,
+      "loss": 0.004,
+      "step": 10412
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5665920829760984e-06,
+      "loss": 1.5652,
+      "step": 10413
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5665502603458731e-06,
+      "loss": 0.0054,
+      "step": 10414
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566508437715648e-06,
+      "loss": 0.0012,
+      "step": 10415
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5664666150854227e-06,
+      "loss": 0.0007,
+      "step": 10416
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5664247924551975e-06,
+      "loss": 0.007,
+      "step": 10417
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5663829698249723e-06,
+      "loss": 0.002,
+      "step": 10418
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5663411471947469e-06,
+      "loss": 0.0008,
+      "step": 10419
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5662993245645217e-06,
+      "loss": 0.0021,
+      "step": 10420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5662575019342964e-06,
+      "loss": 0.0005,
+      "step": 10421
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5662156793040712e-06,
+      "loss": 0.0185,
+      "step": 10422
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566173856673846e-06,
+      "loss": 1.7951,
+      "step": 10423
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566132034043621e-06,
+      "loss": 0.0007,
+      "step": 10424
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5660902114133958e-06,
+      "loss": 0.0007,
+      "step": 10425
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5660483887831706e-06,
+      "loss": 0.0026,
+      "step": 10426
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5660065661529454e-06,
+      "loss": 2.8458,
+      "step": 10427
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5659647435227202e-06,
+      "loss": 1.2615,
+      "step": 10428
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565922920892495e-06,
+      "loss": 0.0021,
+      "step": 10429
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5658810982622697e-06,
+      "loss": 0.0009,
+      "step": 10430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5658392756320445e-06,
+      "loss": 0.6236,
+      "step": 10431
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5657974530018193e-06,
+      "loss": 0.0041,
+      "step": 10432
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5657556303715939e-06,
+      "loss": 0.0008,
+      "step": 10433
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5657138077413687e-06,
+      "loss": 0.0074,
+      "step": 10434
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5656719851111434e-06,
+      "loss": 0.0006,
+      "step": 10435
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5656301624809182e-06,
+      "loss": 0.0125,
+      "step": 10436
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565588339850693e-06,
+      "loss": 0.3188,
+      "step": 10437
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5655465172204678e-06,
+      "loss": 0.021,
+      "step": 10438
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5655046945902428e-06,
+      "loss": 0.2318,
+      "step": 10439
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5654628719600176e-06,
+      "loss": 2.8949,
+      "step": 10440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5654210493297924e-06,
+      "loss": 0.0526,
+      "step": 10441
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5653792266995672e-06,
+      "loss": 0.0005,
+      "step": 10442
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565337404069342e-06,
+      "loss": 2.5009,
+      "step": 10443
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5652955814391167e-06,
+      "loss": 0.0077,
+      "step": 10444
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5652537588088915e-06,
+      "loss": 0.0007,
+      "step": 10445
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5652119361786663e-06,
+      "loss": 2.2452,
+      "step": 10446
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5651701135484409e-06,
+      "loss": 0.0011,
+      "step": 10447
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5651282909182157e-06,
+      "loss": 2.1192,
+      "step": 10448
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5650864682879905e-06,
+      "loss": 0.0009,
+      "step": 10449
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5650446456577652e-06,
+      "loss": 0.0025,
+      "step": 10450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.56500282302754e-06,
+      "loss": 0.0078,
+      "step": 10451
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5649610003973148e-06,
+      "loss": 1.2959,
+      "step": 10452
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5649191777670896e-06,
+      "loss": 0.0013,
+      "step": 10453
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5648773551368646e-06,
+      "loss": 0.0074,
+      "step": 10454
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5648355325066394e-06,
+      "loss": 0.0038,
+      "step": 10455
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5647937098764142e-06,
+      "loss": 0.0009,
+      "step": 10456
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.564751887246189e-06,
+      "loss": 1.4699,
+      "step": 10457
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5647100646159637e-06,
+      "loss": 0.0008,
+      "step": 10458
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5646682419857385e-06,
+      "loss": 0.0022,
+      "step": 10459
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.564626419355513e-06,
+      "loss": 0.2542,
+      "step": 10460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5645845967252879e-06,
+      "loss": 0.0724,
+      "step": 10461
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5645427740950627e-06,
+      "loss": 0.0005,
+      "step": 10462
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5645009514648375e-06,
+      "loss": 0.0027,
+      "step": 10463
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5644591288346122e-06,
+      "loss": 2.7698,
+      "step": 10464
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.564417306204387e-06,
+      "loss": 0.0009,
+      "step": 10465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5643754835741618e-06,
+      "loss": 0.0114,
+      "step": 10466
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5643336609439366e-06,
+      "loss": 0.0042,
+      "step": 10467
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5642918383137116e-06,
+      "loss": 0.0013,
+      "step": 10468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5642500156834864e-06,
+      "loss": 0.673,
+      "step": 10469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5642081930532612e-06,
+      "loss": 0.0023,
+      "step": 10470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.564166370423036e-06,
+      "loss": 0.003,
+      "step": 10471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5641245477928108e-06,
+      "loss": 0.0007,
+      "step": 10472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5640827251625855e-06,
+      "loss": 1.1151,
+      "step": 10473
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5640409025323601e-06,
+      "loss": 0.0006,
+      "step": 10474
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563999079902135e-06,
+      "loss": 0.0015,
+      "step": 10475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5639572572719097e-06,
+      "loss": 2.5926,
+      "step": 10476
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5639154346416845e-06,
+      "loss": 0.1814,
+      "step": 10477
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5638736120114593e-06,
+      "loss": 0.0035,
+      "step": 10478
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563831789381234e-06,
+      "loss": 0.0026,
+      "step": 10479
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5637899667510088e-06,
+      "loss": 0.0015,
+      "step": 10480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5637481441207836e-06,
+      "loss": 0.0008,
+      "step": 10481
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5637063214905584e-06,
+      "loss": 0.0019,
+      "step": 10482
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5636644988603334e-06,
+      "loss": 0.0064,
+      "step": 10483
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5636226762301082e-06,
+      "loss": 2.2702,
+      "step": 10484
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563580853599883e-06,
+      "loss": 0.0474,
+      "step": 10485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5635390309696578e-06,
+      "loss": 0.0105,
+      "step": 10486
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5634972083394323e-06,
+      "loss": 1.2118,
+      "step": 10487
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5634553857092071e-06,
+      "loss": 0.0017,
+      "step": 10488
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563413563078982e-06,
+      "loss": 0.0032,
+      "step": 10489
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5633717404487567e-06,
+      "loss": 0.0034,
+      "step": 10490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5633299178185315e-06,
+      "loss": 0.1297,
+      "step": 10491
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5632880951883063e-06,
+      "loss": 0.6909,
+      "step": 10492
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563246272558081e-06,
+      "loss": 0.0081,
+      "step": 10493
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5632044499278558e-06,
+      "loss": 0.9662,
+      "step": 10494
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5631626272976306e-06,
+      "loss": 0.0051,
+      "step": 10495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5631208046674054e-06,
+      "loss": 0.0009,
+      "step": 10496
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5630789820371802e-06,
+      "loss": 1.1297,
+      "step": 10497
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5630371594069552e-06,
+      "loss": 0.0012,
+      "step": 10498
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.56299533677673e-06,
+      "loss": 0.0005,
+      "step": 10499
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5629535141465048e-06,
+      "loss": 0.0092,
+      "step": 10500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5629116915162793e-06,
+      "loss": 0.0007,
+      "step": 10501
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5628698688860541e-06,
+      "loss": 1.9252,
+      "step": 10502
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562828046255829e-06,
+      "loss": 0.1218,
+      "step": 10503
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5627862236256037e-06,
+      "loss": 0.0048,
+      "step": 10504
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5627444009953785e-06,
+      "loss": 0.6766,
+      "step": 10505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5627025783651533e-06,
+      "loss": 1.5435,
+      "step": 10506
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562660755734928e-06,
+      "loss": 0.0015,
+      "step": 10507
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5626189331047028e-06,
+      "loss": 0.0321,
+      "step": 10508
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5625771104744776e-06,
+      "loss": 0.0268,
+      "step": 10509
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5625352878442524e-06,
+      "loss": 0.0005,
+      "step": 10510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5624934652140272e-06,
+      "loss": 0.001,
+      "step": 10511
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5624516425838022e-06,
+      "loss": 0.0017,
+      "step": 10512
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562409819953577e-06,
+      "loss": 0.0053,
+      "step": 10513
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5623679973233518e-06,
+      "loss": 0.0023,
+      "step": 10514
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5623261746931264e-06,
+      "loss": 2.5893,
+      "step": 10515
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5622843520629011e-06,
+      "loss": 0.0029,
+      "step": 10516
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562242529432676e-06,
+      "loss": 0.0015,
+      "step": 10517
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5622007068024507e-06,
+      "loss": 0.0011,
+      "step": 10518
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5621588841722255e-06,
+      "loss": 0.013,
+      "step": 10519
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5621170615420003e-06,
+      "loss": 0.099,
+      "step": 10520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562075238911775e-06,
+      "loss": 0.0088,
+      "step": 10521
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5620334162815499e-06,
+      "loss": 0.0027,
+      "step": 10522
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5619915936513246e-06,
+      "loss": 0.0014,
+      "step": 10523
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5619497710210994e-06,
+      "loss": 0.001,
+      "step": 10524
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5619079483908742e-06,
+      "loss": 0.0014,
+      "step": 10525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561866125760649e-06,
+      "loss": 0.6821,
+      "step": 10526
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561824303130424e-06,
+      "loss": 0.0008,
+      "step": 10527
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5617824805001986e-06,
+      "loss": 0.0009,
+      "step": 10528
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5617406578699734e-06,
+      "loss": 0.0023,
+      "step": 10529
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5616988352397481e-06,
+      "loss": 0.0012,
+      "step": 10530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561657012609523e-06,
+      "loss": 0.9675,
+      "step": 10531
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5616151899792977e-06,
+      "loss": 0.0172,
+      "step": 10532
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5615733673490725e-06,
+      "loss": 0.8373,
+      "step": 10533
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5615315447188473e-06,
+      "loss": 0.003,
+      "step": 10534
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561489722088622e-06,
+      "loss": 0.0007,
+      "step": 10535
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5614478994583969e-06,
+      "loss": 0.872,
+      "step": 10536
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5614060768281717e-06,
+      "loss": 0.0251,
+      "step": 10537
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5613642541979464e-06,
+      "loss": 0.0103,
+      "step": 10538
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5613224315677212e-06,
+      "loss": 0.0332,
+      "step": 10539
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561280608937496e-06,
+      "loss": 0.0028,
+      "step": 10540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5612387863072706e-06,
+      "loss": 0.0018,
+      "step": 10541
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5611969636770456e-06,
+      "loss": 0.0012,
+      "step": 10542
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5611551410468204e-06,
+      "loss": 0.0015,
+      "step": 10543
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5611133184165952e-06,
+      "loss": 0.337,
+      "step": 10544
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.56107149578637e-06,
+      "loss": 4.3089,
+      "step": 10545
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5610296731561447e-06,
+      "loss": 0.0019,
+      "step": 10546
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5609878505259195e-06,
+      "loss": 0.0028,
+      "step": 10547
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5609460278956943e-06,
+      "loss": 0.0029,
+      "step": 10548
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560904205265469e-06,
+      "loss": 0.001,
+      "step": 10549
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5608623826352439e-06,
+      "loss": 0.0006,
+      "step": 10550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5608205600050187e-06,
+      "loss": 0.0007,
+      "step": 10551
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5607787373747934e-06,
+      "loss": 1.1443,
+      "step": 10552
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5607369147445682e-06,
+      "loss": 0.0007,
+      "step": 10553
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560695092114343e-06,
+      "loss": 0.0006,
+      "step": 10554
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5606532694841176e-06,
+      "loss": 0.0007,
+      "step": 10555
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5606114468538926e-06,
+      "loss": 1.2858,
+      "step": 10556
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5605696242236674e-06,
+      "loss": 0.0005,
+      "step": 10557
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5605278015934422e-06,
+      "loss": 0.7531,
+      "step": 10558
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560485978963217e-06,
+      "loss": 0.0014,
+      "step": 10559
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5604441563329917e-06,
+      "loss": 0.3752,
+      "step": 10560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5604023337027665e-06,
+      "loss": 0.0022,
+      "step": 10561
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5603605110725413e-06,
+      "loss": 0.0032,
+      "step": 10562
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560318688442316e-06,
+      "loss": 0.0009,
+      "step": 10563
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5602768658120909e-06,
+      "loss": 0.0027,
+      "step": 10564
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5602350431818657e-06,
+      "loss": 0.8123,
+      "step": 10565
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5601932205516405e-06,
+      "loss": 1.3557,
+      "step": 10566
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5601513979214152e-06,
+      "loss": 0.8211,
+      "step": 10567
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5601095752911898e-06,
+      "loss": 1.5575,
+      "step": 10568
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5600677526609646e-06,
+      "loss": 0.0182,
+      "step": 10569
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5600259300307394e-06,
+      "loss": 0.0056,
+      "step": 10570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5599841074005144e-06,
+      "loss": 0.4744,
+      "step": 10571
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5599422847702892e-06,
+      "loss": 0.0018,
+      "step": 10572
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559900462140064e-06,
+      "loss": 0.0057,
+      "step": 10573
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5598586395098387e-06,
+      "loss": 1.6623,
+      "step": 10574
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5598168168796135e-06,
+      "loss": 0.0267,
+      "step": 10575
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5597749942493883e-06,
+      "loss": 0.0008,
+      "step": 10576
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559733171619163e-06,
+      "loss": 0.0011,
+      "step": 10577
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5596913489889379e-06,
+      "loss": 0.0012,
+      "step": 10578
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5596495263587127e-06,
+      "loss": 0.0011,
+      "step": 10579
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5596077037284875e-06,
+      "loss": 0.0005,
+      "step": 10580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5595658810982622e-06,
+      "loss": 0.0012,
+      "step": 10581
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5595240584680368e-06,
+      "loss": 1.2844,
+      "step": 10582
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5594822358378116e-06,
+      "loss": 0.0006,
+      "step": 10583
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5594404132075864e-06,
+      "loss": 0.376,
+      "step": 10584
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5593985905773614e-06,
+      "loss": 0.0014,
+      "step": 10585
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5593567679471362e-06,
+      "loss": 0.0006,
+      "step": 10586
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559314945316911e-06,
+      "loss": 0.0009,
+      "step": 10587
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5592731226866858e-06,
+      "loss": 2.5067,
+      "step": 10588
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5592313000564605e-06,
+      "loss": 0.0012,
+      "step": 10589
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5591894774262353e-06,
+      "loss": 0.0064,
+      "step": 10590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5591476547960101e-06,
+      "loss": 0.0009,
+      "step": 10591
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559105832165785e-06,
+      "loss": 0.0066,
+      "step": 10592
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5590640095355597e-06,
+      "loss": 0.0032,
+      "step": 10593
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5590221869053345e-06,
+      "loss": 0.0027,
+      "step": 10594
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558980364275109e-06,
+      "loss": 0.0021,
+      "step": 10595
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5589385416448838e-06,
+      "loss": 0.0123,
+      "step": 10596
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5588967190146586e-06,
+      "loss": 0.0018,
+      "step": 10597
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5588548963844334e-06,
+      "loss": 0.0008,
+      "step": 10598
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5588130737542082e-06,
+      "loss": 0.0028,
+      "step": 10599
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5587712511239832e-06,
+      "loss": 0.0005,
+      "step": 10600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558729428493758e-06,
+      "loss": 0.4316,
+      "step": 10601
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5586876058635328e-06,
+      "loss": 0.0007,
+      "step": 10602
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5586457832333075e-06,
+      "loss": 0.0031,
+      "step": 10603
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5586039606030823e-06,
+      "loss": 0.0011,
+      "step": 10604
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5585621379728571e-06,
+      "loss": 0.0032,
+      "step": 10605
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558520315342632e-06,
+      "loss": 0.0023,
+      "step": 10606
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5584784927124067e-06,
+      "loss": 0.0007,
+      "step": 10607
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5584366700821815e-06,
+      "loss": 0.2288,
+      "step": 10608
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558394847451956e-06,
+      "loss": 0.0098,
+      "step": 10609
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5583530248217308e-06,
+      "loss": 0.7976,
+      "step": 10610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5583112021915056e-06,
+      "loss": 0.002,
+      "step": 10611
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5582693795612804e-06,
+      "loss": 0.001,
+      "step": 10612
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5582275569310552e-06,
+      "loss": 0.0005,
+      "step": 10613
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.55818573430083e-06,
+      "loss": 1.5503,
+      "step": 10614
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.558143911670605e-06,
+      "loss": 0.0058,
+      "step": 10615
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5581020890403798e-06,
+      "loss": 0.0006,
+      "step": 10616
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5580602664101546e-06,
+      "loss": 1.7682,
+      "step": 10617
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5580184437799293e-06,
+      "loss": 0.0006,
+      "step": 10618
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5579766211497041e-06,
+      "loss": 0.003,
+      "step": 10619
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.557934798519479e-06,
+      "loss": 0.0034,
+      "step": 10620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5578929758892537e-06,
+      "loss": 0.0062,
+      "step": 10621
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5578511532590285e-06,
+      "loss": 0.0047,
+      "step": 10622
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.557809330628803e-06,
+      "loss": 0.6883,
+      "step": 10623
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5577675079985779e-06,
+      "loss": 0.5699,
+      "step": 10624
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5577256853683526e-06,
+      "loss": 0.0013,
+      "step": 10625
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5576838627381274e-06,
+      "loss": 0.0006,
+      "step": 10626
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5576420401079022e-06,
+      "loss": 0.0041,
+      "step": 10627
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.557600217477677e-06,
+      "loss": 0.0011,
+      "step": 10628
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.557558394847452e-06,
+      "loss": 0.0041,
+      "step": 10629
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5575165722172268e-06,
+      "loss": 0.0066,
+      "step": 10630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5574747495870016e-06,
+      "loss": 0.0025,
+      "step": 10631
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5574329269567764e-06,
+      "loss": 0.004,
+      "step": 10632
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5573911043265511e-06,
+      "loss": 0.596,
+      "step": 10633
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.557349281696326e-06,
+      "loss": 0.0004,
+      "step": 10634
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5573074590661007e-06,
+      "loss": 0.001,
+      "step": 10635
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5572656364358753e-06,
+      "loss": 2.2455,
+      "step": 10636
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.55722381380565e-06,
+      "loss": 0.003,
+      "step": 10637
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5571819911754249e-06,
+      "loss": 0.6007,
+      "step": 10638
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5571401685451996e-06,
+      "loss": 0.0093,
+      "step": 10639
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5570983459149744e-06,
+      "loss": 0.001,
+      "step": 10640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5570565232847492e-06,
+      "loss": 0.0008,
+      "step": 10641
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.557014700654524e-06,
+      "loss": 1.6565,
+      "step": 10642
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5569728780242988e-06,
+      "loss": 0.907,
+      "step": 10643
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5569310553940738e-06,
+      "loss": 0.0007,
+      "step": 10644
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5568892327638486e-06,
+      "loss": 0.0007,
+      "step": 10645
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5568474101336234e-06,
+      "loss": 0.0012,
+      "step": 10646
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5568055875033981e-06,
+      "loss": 0.0011,
+      "step": 10647
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556763764873173e-06,
+      "loss": 0.0019,
+      "step": 10648
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5567219422429477e-06,
+      "loss": 0.0225,
+      "step": 10649
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5566801196127223e-06,
+      "loss": 0.8778,
+      "step": 10650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556638296982497e-06,
+      "loss": 0.0006,
+      "step": 10651
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5565964743522719e-06,
+      "loss": 0.2299,
+      "step": 10652
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5565546517220467e-06,
+      "loss": 0.002,
+      "step": 10653
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5565128290918214e-06,
+      "loss": 0.0009,
+      "step": 10654
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5564710064615962e-06,
+      "loss": 0.001,
+      "step": 10655
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556429183831371e-06,
+      "loss": 0.7483,
+      "step": 10656
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5563873612011458e-06,
+      "loss": 0.006,
+      "step": 10657
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5563455385709206e-06,
+      "loss": 0.2669,
+      "step": 10658
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5563037159406956e-06,
+      "loss": 0.001,
+      "step": 10659
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5562618933104704e-06,
+      "loss": 0.003,
+      "step": 10660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5562200706802452e-06,
+      "loss": 0.0825,
+      "step": 10661
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.55617824805002e-06,
+      "loss": 0.0056,
+      "step": 10662
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5561364254197945e-06,
+      "loss": 0.0006,
+      "step": 10663
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5560946027895693e-06,
+      "loss": 0.0008,
+      "step": 10664
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556052780159344e-06,
+      "loss": 0.0004,
+      "step": 10665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5560109575291189e-06,
+      "loss": 1.9226,
+      "step": 10666
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5559691348988937e-06,
+      "loss": 0.0007,
+      "step": 10667
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5559273122686684e-06,
+      "loss": 0.0005,
+      "step": 10668
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5558854896384432e-06,
+      "loss": 0.0009,
+      "step": 10669
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555843667008218e-06,
+      "loss": 0.0008,
+      "step": 10670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5558018443779928e-06,
+      "loss": 0.0049,
+      "step": 10671
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5557600217477676e-06,
+      "loss": 1.4184,
+      "step": 10672
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5557181991175426e-06,
+      "loss": 0.0041,
+      "step": 10673
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5556763764873174e-06,
+      "loss": 0.0004,
+      "step": 10674
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5556345538570922e-06,
+      "loss": 0.0032,
+      "step": 10675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555592731226867e-06,
+      "loss": 0.0666,
+      "step": 10676
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5555509085966415e-06,
+      "loss": 3.1983,
+      "step": 10677
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5555090859664163e-06,
+      "loss": 0.001,
+      "step": 10678
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555467263336191e-06,
+      "loss": 0.001,
+      "step": 10679
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5554254407059659e-06,
+      "loss": 0.0005,
+      "step": 10680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5553836180757407e-06,
+      "loss": 0.2608,
+      "step": 10681
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5553417954455155e-06,
+      "loss": 0.0005,
+      "step": 10682
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5552999728152902e-06,
+      "loss": 0.0012,
+      "step": 10683
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555258150185065e-06,
+      "loss": 0.4937,
+      "step": 10684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5552163275548398e-06,
+      "loss": 0.5228,
+      "step": 10685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5551745049246146e-06,
+      "loss": 0.0312,
+      "step": 10686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5551326822943894e-06,
+      "loss": 0.0472,
+      "step": 10687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5550908596641644e-06,
+      "loss": 0.0018,
+      "step": 10688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5550490370339392e-06,
+      "loss": 0.0004,
+      "step": 10689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5550072144037137e-06,
+      "loss": 0.0017,
+      "step": 10690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5549653917734885e-06,
+      "loss": 0.0012,
+      "step": 10691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5549235691432633e-06,
+      "loss": 0.0008,
+      "step": 10692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5548817465130381e-06,
+      "loss": 1.2369,
+      "step": 10693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5548399238828129e-06,
+      "loss": 0.0009,
+      "step": 10694
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5547981012525877e-06,
+      "loss": 0.7843,
+      "step": 10695
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5547562786223625e-06,
+      "loss": 0.2492,
+      "step": 10696
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5547144559921373e-06,
+      "loss": 0.0004,
+      "step": 10697
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554672633361912e-06,
+      "loss": 0.0008,
+      "step": 10698
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5546308107316868e-06,
+      "loss": 0.0009,
+      "step": 10699
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5545889881014616e-06,
+      "loss": 0.0012,
+      "step": 10700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5545471654712364e-06,
+      "loss": 1.8902,
+      "step": 10701
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5545053428410112e-06,
+      "loss": 0.0021,
+      "step": 10702
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5544635202107862e-06,
+      "loss": 0.0004,
+      "step": 10703
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5544216975805608e-06,
+      "loss": 0.0033,
+      "step": 10704
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5543798749503355e-06,
+      "loss": 0.0027,
+      "step": 10705
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5543380523201103e-06,
+      "loss": 0.004,
+      "step": 10706
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5542962296898851e-06,
+      "loss": 0.0045,
+      "step": 10707
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.55425440705966e-06,
+      "loss": 0.0075,
+      "step": 10708
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5542125844294347e-06,
+      "loss": 0.0018,
+      "step": 10709
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5541707617992095e-06,
+      "loss": 2.1122,
+      "step": 10710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5541289391689843e-06,
+      "loss": 2.5786,
+      "step": 10711
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554087116538759e-06,
+      "loss": 0.0129,
+      "step": 10712
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5540452939085338e-06,
+      "loss": 0.0006,
+      "step": 10713
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5540034712783086e-06,
+      "loss": 1.9081,
+      "step": 10714
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5539616486480834e-06,
+      "loss": 0.0038,
+      "step": 10715
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5539198260178582e-06,
+      "loss": 0.3413,
+      "step": 10716
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5538780033876332e-06,
+      "loss": 0.0006,
+      "step": 10717
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5538361807574078e-06,
+      "loss": 0.5242,
+      "step": 10718
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5537943581271826e-06,
+      "loss": 2.168,
+      "step": 10719
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5537525354969573e-06,
+      "loss": 0.0004,
+      "step": 10720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5537107128667321e-06,
+      "loss": 0.2236,
+      "step": 10721
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553668890236507e-06,
+      "loss": 0.0534,
+      "step": 10722
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5536270676062817e-06,
+      "loss": 0.0009,
+      "step": 10723
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5535852449760565e-06,
+      "loss": 0.9914,
+      "step": 10724
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5535434223458313e-06,
+      "loss": 0.2886,
+      "step": 10725
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553501599715606e-06,
+      "loss": 0.0018,
+      "step": 10726
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5534597770853808e-06,
+      "loss": 1.9289,
+      "step": 10727
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5534179544551556e-06,
+      "loss": 0.2796,
+      "step": 10728
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5533761318249304e-06,
+      "loss": 0.7741,
+      "step": 10729
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553334309194705e-06,
+      "loss": 0.0011,
+      "step": 10730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5532924865644798e-06,
+      "loss": 0.0009,
+      "step": 10731
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5532506639342548e-06,
+      "loss": 0.2127,
+      "step": 10732
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5532088413040296e-06,
+      "loss": 0.0022,
+      "step": 10733
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5531670186738043e-06,
+      "loss": 0.012,
+      "step": 10734
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5531251960435791e-06,
+      "loss": 0.273,
+      "step": 10735
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553083373413354e-06,
+      "loss": 0.0019,
+      "step": 10736
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5530415507831287e-06,
+      "loss": 0.0006,
+      "step": 10737
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5529997281529035e-06,
+      "loss": 0.008,
+      "step": 10738
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5529579055226783e-06,
+      "loss": 0.1101,
+      "step": 10739
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552916082892453e-06,
+      "loss": 2.0435,
+      "step": 10740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5528742602622279e-06,
+      "loss": 0.0008,
+      "step": 10741
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5528324376320026e-06,
+      "loss": 0.0901,
+      "step": 10742
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5527906150017774e-06,
+      "loss": 0.386,
+      "step": 10743
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552748792371552e-06,
+      "loss": 0.0017,
+      "step": 10744
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5527069697413268e-06,
+      "loss": 0.0011,
+      "step": 10745
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5526651471111016e-06,
+      "loss": 0.03,
+      "step": 10746
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5526233244808766e-06,
+      "loss": 1.2805,
+      "step": 10747
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5525815018506514e-06,
+      "loss": 0.0025,
+      "step": 10748
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5525396792204261e-06,
+      "loss": 0.0025,
+      "step": 10749
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552497856590201e-06,
+      "loss": 0.0006,
+      "step": 10750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5524560339599757e-06,
+      "loss": 0.0009,
+      "step": 10751
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5524142113297505e-06,
+      "loss": 2.7551,
+      "step": 10752
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5523723886995253e-06,
+      "loss": 0.0082,
+      "step": 10753
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5523305660693e-06,
+      "loss": 0.0009,
+      "step": 10754
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5522887434390749e-06,
+      "loss": 0.0008,
+      "step": 10755
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5522469208088496e-06,
+      "loss": 0.0178,
+      "step": 10756
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5522050981786244e-06,
+      "loss": 1.9602,
+      "step": 10757
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552163275548399e-06,
+      "loss": 0.0056,
+      "step": 10758
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5521214529181738e-06,
+      "loss": 0.0008,
+      "step": 10759
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5520796302879486e-06,
+      "loss": 0.001,
+      "step": 10760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5520378076577236e-06,
+      "loss": 0.0005,
+      "step": 10761
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5519959850274984e-06,
+      "loss": 0.0035,
+      "step": 10762
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5519541623972731e-06,
+      "loss": 0.005,
+      "step": 10763
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551912339767048e-06,
+      "loss": 0.0038,
+      "step": 10764
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5518705171368227e-06,
+      "loss": 0.1128,
+      "step": 10765
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5518286945065975e-06,
+      "loss": 1.9367,
+      "step": 10766
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5517868718763723e-06,
+      "loss": 0.1057,
+      "step": 10767
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551745049246147e-06,
+      "loss": 0.002,
+      "step": 10768
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5517032266159219e-06,
+      "loss": 0.0021,
+      "step": 10769
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5516614039856967e-06,
+      "loss": 0.0035,
+      "step": 10770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5516195813554712e-06,
+      "loss": 0.0062,
+      "step": 10771
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.551577758725246e-06,
+      "loss": 0.0009,
+      "step": 10772
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5515359360950208e-06,
+      "loss": 0.0008,
+      "step": 10773
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5514941134647956e-06,
+      "loss": 0.0034,
+      "step": 10774
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5514522908345704e-06,
+      "loss": 0.002,
+      "step": 10775
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5514104682043454e-06,
+      "loss": 0.0035,
+      "step": 10776
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5513686455741202e-06,
+      "loss": 0.0019,
+      "step": 10777
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.551326822943895e-06,
+      "loss": 0.0027,
+      "step": 10778
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5512850003136697e-06,
+      "loss": 0.0086,
+      "step": 10779
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5512431776834445e-06,
+      "loss": 0.0024,
+      "step": 10780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5512013550532193e-06,
+      "loss": 0.0016,
+      "step": 10781
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.551159532422994e-06,
+      "loss": 0.4136,
+      "step": 10782
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5511177097927689e-06,
+      "loss": 0.0028,
+      "step": 10783
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5510758871625437e-06,
+      "loss": 0.0042,
+      "step": 10784
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5510340645323182e-06,
+      "loss": 0.0011,
+      "step": 10785
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.550992241902093e-06,
+      "loss": 0.0069,
+      "step": 10786
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5509504192718678e-06,
+      "loss": 0.4242,
+      "step": 10787
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5509085966416426e-06,
+      "loss": 0.0018,
+      "step": 10788
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5508667740114174e-06,
+      "loss": 0.0012,
+      "step": 10789
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5508249513811924e-06,
+      "loss": 0.002,
+      "step": 10790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5507831287509672e-06,
+      "loss": 0.0021,
+      "step": 10791
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.550741306120742e-06,
+      "loss": 0.0096,
+      "step": 10792
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5506994834905167e-06,
+      "loss": 0.596,
+      "step": 10793
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5506576608602915e-06,
+      "loss": 0.006,
+      "step": 10794
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5506158382300663e-06,
+      "loss": 0.6695,
+      "step": 10795
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.550574015599841e-06,
+      "loss": 0.009,
+      "step": 10796
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5505321929696159e-06,
+      "loss": 0.0004,
+      "step": 10797
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5504903703393905e-06,
+      "loss": 0.0017,
+      "step": 10798
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5504485477091652e-06,
+      "loss": 2.0046,
+      "step": 10799
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.55040672507894e-06,
+      "loss": 2.8612,
+      "step": 10800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5503649024487148e-06,
+      "loss": 0.0006,
+      "step": 10801
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5503230798184896e-06,
+      "loss": 0.0013,
+      "step": 10802
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5502812571882644e-06,
+      "loss": 0.0012,
+      "step": 10803
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5502394345580392e-06,
+      "loss": 0.0027,
+      "step": 10804
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5501976119278142e-06,
+      "loss": 0.3455,
+      "step": 10805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.550155789297589e-06,
+      "loss": 0.0048,
+      "step": 10806
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5501139666673637e-06,
+      "loss": 0.053,
+      "step": 10807
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5500721440371385e-06,
+      "loss": 2.1515,
+      "step": 10808
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5500303214069133e-06,
+      "loss": 0.002,
+      "step": 10809
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5499884987766881e-06,
+      "loss": 1.94,
+      "step": 10810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5499466761464629e-06,
+      "loss": 0.0045,
+      "step": 10811
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5499048535162375e-06,
+      "loss": 0.0006,
+      "step": 10812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5498630308860123e-06,
+      "loss": 0.0026,
+      "step": 10813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549821208255787e-06,
+      "loss": 0.0014,
+      "step": 10814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5497793856255618e-06,
+      "loss": 0.0022,
+      "step": 10815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5497375629953366e-06,
+      "loss": 2.3594,
+      "step": 10816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5496957403651114e-06,
+      "loss": 0.5671,
+      "step": 10817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5496539177348862e-06,
+      "loss": 0.0008,
+      "step": 10818
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549612095104661e-06,
+      "loss": 0.0014,
+      "step": 10819
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549570272474436e-06,
+      "loss": 0.003,
+      "step": 10820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5495284498442108e-06,
+      "loss": 0.0045,
+      "step": 10821
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5494866272139855e-06,
+      "loss": 0.0018,
+      "step": 10822
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5494448045837603e-06,
+      "loss": 0.0014,
+      "step": 10823
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5494029819535351e-06,
+      "loss": 2.9189,
+      "step": 10824
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.54936115932331e-06,
+      "loss": 0.0028,
+      "step": 10825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5493193366930845e-06,
+      "loss": 0.8714,
+      "step": 10826
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5492775140628593e-06,
+      "loss": 1.0766,
+      "step": 10827
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549235691432634e-06,
+      "loss": 0.2992,
+      "step": 10828
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5491938688024088e-06,
+      "loss": 0.3271,
+      "step": 10829
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5491520461721836e-06,
+      "loss": 0.5141,
+      "step": 10830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5491102235419584e-06,
+      "loss": 0.0068,
+      "step": 10831
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5490684009117332e-06,
+      "loss": 0.0099,
+      "step": 10832
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549026578281508e-06,
+      "loss": 0.0013,
+      "step": 10833
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548984755651283e-06,
+      "loss": 0.0021,
+      "step": 10834
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5489429330210578e-06,
+      "loss": 0.0016,
+      "step": 10835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5489011103908326e-06,
+      "loss": 0.0085,
+      "step": 10836
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5488592877606073e-06,
+      "loss": 0.0019,
+      "step": 10837
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5488174651303821e-06,
+      "loss": 0.0005,
+      "step": 10838
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5487756425001567e-06,
+      "loss": 0.0004,
+      "step": 10839
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5487338198699315e-06,
+      "loss": 0.0015,
+      "step": 10840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5486919972397063e-06,
+      "loss": 1.7264,
+      "step": 10841
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548650174609481e-06,
+      "loss": 0.0042,
+      "step": 10842
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5486083519792558e-06,
+      "loss": 0.0019,
+      "step": 10843
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5485665293490306e-06,
+      "loss": 0.0016,
+      "step": 10844
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5485247067188054e-06,
+      "loss": 2.3504,
+      "step": 10845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5484828840885802e-06,
+      "loss": 1.0302,
+      "step": 10846
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548441061458355e-06,
+      "loss": 0.2961,
+      "step": 10847
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5483992388281298e-06,
+      "loss": 0.003,
+      "step": 10848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5483574161979048e-06,
+      "loss": 0.0025,
+      "step": 10849
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5483155935676796e-06,
+      "loss": 1.0472,
+      "step": 10850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5482737709374543e-06,
+      "loss": 0.0031,
+      "step": 10851
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5482319483072291e-06,
+      "loss": 2.4204,
+      "step": 10852
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5481901256770037e-06,
+      "loss": 0.0684,
+      "step": 10853
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5481483030467785e-06,
+      "loss": 1.9591,
+      "step": 10854
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5481064804165533e-06,
+      "loss": 0.0008,
+      "step": 10855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548064657786328e-06,
+      "loss": 2.211,
+      "step": 10856
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5480228351561029e-06,
+      "loss": 0.0015,
+      "step": 10857
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5479810125258776e-06,
+      "loss": 0.001,
+      "step": 10858
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5479391898956524e-06,
+      "loss": 0.0027,
+      "step": 10859
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5478973672654272e-06,
+      "loss": 0.0006,
+      "step": 10860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547855544635202e-06,
+      "loss": 0.4654,
+      "step": 10861
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5478137220049768e-06,
+      "loss": 0.0064,
+      "step": 10862
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5477718993747516e-06,
+      "loss": 0.225,
+      "step": 10863
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5477300767445266e-06,
+      "loss": 0.224,
+      "step": 10864
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5476882541143014e-06,
+      "loss": 0.001,
+      "step": 10865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547646431484076e-06,
+      "loss": 0.0051,
+      "step": 10866
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5476046088538507e-06,
+      "loss": 0.003,
+      "step": 10867
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5475627862236255e-06,
+      "loss": 0.0008,
+      "step": 10868
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5475209635934003e-06,
+      "loss": 0.0281,
+      "step": 10869
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547479140963175e-06,
+      "loss": 0.0012,
+      "step": 10870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5474373183329499e-06,
+      "loss": 0.0431,
+      "step": 10871
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5473954957027246e-06,
+      "loss": 0.0167,
+      "step": 10872
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5473536730724994e-06,
+      "loss": 0.118,
+      "step": 10873
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5473118504422742e-06,
+      "loss": 1.6629,
+      "step": 10874
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547270027812049e-06,
+      "loss": 0.0035,
+      "step": 10875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5472282051818238e-06,
+      "loss": 0.001,
+      "step": 10876
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5471863825515986e-06,
+      "loss": 0.0023,
+      "step": 10877
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5471445599213736e-06,
+      "loss": 0.0026,
+      "step": 10878
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5471027372911484e-06,
+      "loss": 0.0683,
+      "step": 10879
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547060914660923e-06,
+      "loss": 0.0021,
+      "step": 10880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5470190920306977e-06,
+      "loss": 0.195,
+      "step": 10881
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5469772694004725e-06,
+      "loss": 0.0024,
+      "step": 10882
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5469354467702473e-06,
+      "loss": 0.0058,
+      "step": 10883
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546893624140022e-06,
+      "loss": 0.3091,
+      "step": 10884
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5468518015097969e-06,
+      "loss": 0.0422,
+      "step": 10885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5468099788795717e-06,
+      "loss": 0.0009,
+      "step": 10886
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5467681562493464e-06,
+      "loss": 0.2001,
+      "step": 10887
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5467263336191212e-06,
+      "loss": 0.4738,
+      "step": 10888
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546684510988896e-06,
+      "loss": 1.9072,
+      "step": 10889
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5466426883586708e-06,
+      "loss": 0.0007,
+      "step": 10890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5466008657284456e-06,
+      "loss": 0.6374,
+      "step": 10891
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5465590430982204e-06,
+      "loss": 0.0056,
+      "step": 10892
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5465172204679952e-06,
+      "loss": 0.0035,
+      "step": 10893
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.54647539783777e-06,
+      "loss": 0.1431,
+      "step": 10894
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5464335752075447e-06,
+      "loss": 0.0009,
+      "step": 10895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5463917525773195e-06,
+      "loss": 0.0013,
+      "step": 10896
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5463499299470943e-06,
+      "loss": 0.0044,
+      "step": 10897
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546308107316869e-06,
+      "loss": 0.0013,
+      "step": 10898
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5462662846866439e-06,
+      "loss": 0.0019,
+      "step": 10899
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5462244620564187e-06,
+      "loss": 1.1772,
+      "step": 10900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5461826394261935e-06,
+      "loss": 0.0493,
+      "step": 10901
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5461408167959682e-06,
+      "loss": 0.0006,
+      "step": 10902
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546098994165743e-06,
+      "loss": 2.6579,
+      "step": 10903
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5460571715355178e-06,
+      "loss": 0.0012,
+      "step": 10904
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5460153489052926e-06,
+      "loss": 0.0006,
+      "step": 10905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5459735262750672e-06,
+      "loss": 0.235,
+      "step": 10906
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545931703644842e-06,
+      "loss": 0.0403,
+      "step": 10907
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545889881014617e-06,
+      "loss": 0.011,
+      "step": 10908
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5458480583843917e-06,
+      "loss": 0.0005,
+      "step": 10909
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5458062357541665e-06,
+      "loss": 0.0011,
+      "step": 10910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5457644131239413e-06,
+      "loss": 1.2234,
+      "step": 10911
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545722590493716e-06,
+      "loss": 0.0028,
+      "step": 10912
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5456807678634909e-06,
+      "loss": 0.0084,
+      "step": 10913
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5456389452332657e-06,
+      "loss": 0.0084,
+      "step": 10914
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5455971226030405e-06,
+      "loss": 0.0023,
+      "step": 10915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5455552999728152e-06,
+      "loss": 0.9915,
+      "step": 10916
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.54551347734259e-06,
+      "loss": 2.1755,
+      "step": 10917
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5454716547123648e-06,
+      "loss": 0.0013,
+      "step": 10918
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5454298320821396e-06,
+      "loss": 0.0016,
+      "step": 10919
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5453880094519142e-06,
+      "loss": 0.0021,
+      "step": 10920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545346186821689e-06,
+      "loss": 0.0005,
+      "step": 10921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545304364191464e-06,
+      "loss": 0.4409,
+      "step": 10922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5452625415612388e-06,
+      "loss": 0.0027,
+      "step": 10923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5452207189310135e-06,
+      "loss": 0.0018,
+      "step": 10924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5451788963007883e-06,
+      "loss": 0.8393,
+      "step": 10925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5451370736705631e-06,
+      "loss": 0.0013,
+      "step": 10926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545095251040338e-06,
+      "loss": 0.0019,
+      "step": 10927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5450534284101127e-06,
+      "loss": 0.9946,
+      "step": 10928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5450116057798875e-06,
+      "loss": 0.0096,
+      "step": 10929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5449697831496623e-06,
+      "loss": 0.001,
+      "step": 10930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.544927960519437e-06,
+      "loss": 0.0017,
+      "step": 10931
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5448861378892118e-06,
+      "loss": 0.0005,
+      "step": 10932
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5448443152589864e-06,
+      "loss": 0.0042,
+      "step": 10933
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5448024926287612e-06,
+      "loss": 0.0017,
+      "step": 10934
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.544760669998536e-06,
+      "loss": 1.9782,
+      "step": 10935
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5447188473683108e-06,
+      "loss": 0.003,
+      "step": 10936
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5446770247380858e-06,
+      "loss": 0.0124,
+      "step": 10937
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5446352021078605e-06,
+      "loss": 0.0147,
+      "step": 10938
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5445933794776353e-06,
+      "loss": 2.3414,
+      "step": 10939
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5445515568474101e-06,
+      "loss": 0.0005,
+      "step": 10940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.544509734217185e-06,
+      "loss": 0.0018,
+      "step": 10941
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5444679115869597e-06,
+      "loss": 1.4305,
+      "step": 10942
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5444260889567345e-06,
+      "loss": 1.1597,
+      "step": 10943
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5443842663265093e-06,
+      "loss": 0.0007,
+      "step": 10944
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.544342443696284e-06,
+      "loss": 0.0025,
+      "step": 10945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5443006210660588e-06,
+      "loss": 0.0008,
+      "step": 10946
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5442587984358334e-06,
+      "loss": 0.0022,
+      "step": 10947
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5442169758056082e-06,
+      "loss": 0.001,
+      "step": 10948
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.544175153175383e-06,
+      "loss": 0.0018,
+      "step": 10949
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5441333305451578e-06,
+      "loss": 0.5413,
+      "step": 10950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5440915079149326e-06,
+      "loss": 0.5658,
+      "step": 10951
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5440496852847076e-06,
+      "loss": 0.0037,
+      "step": 10952
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5440078626544823e-06,
+      "loss": 0.7173,
+      "step": 10953
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5439660400242571e-06,
+      "loss": 0.0061,
+      "step": 10954
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543924217394032e-06,
+      "loss": 0.0004,
+      "step": 10955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5438823947638067e-06,
+      "loss": 0.0164,
+      "step": 10956
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5438405721335815e-06,
+      "loss": 0.0015,
+      "step": 10957
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5437987495033563e-06,
+      "loss": 2.7351,
+      "step": 10958
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543756926873131e-06,
+      "loss": 0.004,
+      "step": 10959
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5437151042429058e-06,
+      "loss": 0.0021,
+      "step": 10960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5436732816126804e-06,
+      "loss": 0.0066,
+      "step": 10961
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5436314589824552e-06,
+      "loss": 0.003,
+      "step": 10962
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.54358963635223e-06,
+      "loss": 0.0011,
+      "step": 10963
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5435478137220048e-06,
+      "loss": 0.0008,
+      "step": 10964
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5435059910917796e-06,
+      "loss": 0.0007,
+      "step": 10965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5434641684615546e-06,
+      "loss": 0.0009,
+      "step": 10966
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5434223458313293e-06,
+      "loss": 1.1542,
+      "step": 10967
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5433805232011041e-06,
+      "loss": 0.0071,
+      "step": 10968
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543338700570879e-06,
+      "loss": 0.0058,
+      "step": 10969
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5432968779406537e-06,
+      "loss": 0.001,
+      "step": 10970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5432550553104285e-06,
+      "loss": 0.001,
+      "step": 10971
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5432132326802033e-06,
+      "loss": 0.0005,
+      "step": 10972
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543171410049978e-06,
+      "loss": 0.0015,
+      "step": 10973
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5431295874197526e-06,
+      "loss": 2.431,
+      "step": 10974
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5430877647895274e-06,
+      "loss": 0.0022,
+      "step": 10975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5430459421593022e-06,
+      "loss": 0.0008,
+      "step": 10976
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543004119529077e-06,
+      "loss": 0.0011,
+      "step": 10977
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5429622968988518e-06,
+      "loss": 0.0177,
+      "step": 10978
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5429204742686266e-06,
+      "loss": 0.0008,
+      "step": 10979
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5428786516384014e-06,
+      "loss": 0.0006,
+      "step": 10980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5428368290081764e-06,
+      "loss": 0.0016,
+      "step": 10981
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5427950063779511e-06,
+      "loss": 0.0004,
+      "step": 10982
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542753183747726e-06,
+      "loss": 0.0024,
+      "step": 10983
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5427113611175007e-06,
+      "loss": 0.0008,
+      "step": 10984
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5426695384872755e-06,
+      "loss": 0.001,
+      "step": 10985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5426277158570503e-06,
+      "loss": 0.0012,
+      "step": 10986
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542585893226825e-06,
+      "loss": 0.2752,
+      "step": 10987
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5425440705965997e-06,
+      "loss": 0.2041,
+      "step": 10988
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5425022479663744e-06,
+      "loss": 0.0415,
+      "step": 10989
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5424604253361492e-06,
+      "loss": 0.2959,
+      "step": 10990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542418602705924e-06,
+      "loss": 0.069,
+      "step": 10991
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5423767800756988e-06,
+      "loss": 0.0009,
+      "step": 10992
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5423349574454736e-06,
+      "loss": 0.0068,
+      "step": 10993
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5422931348152484e-06,
+      "loss": 0.0008,
+      "step": 10994
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5422513121850234e-06,
+      "loss": 0.0006,
+      "step": 10995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5422094895547982e-06,
+      "loss": 0.0067,
+      "step": 10996
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542167666924573e-06,
+      "loss": 0.0202,
+      "step": 10997
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5421258442943477e-06,
+      "loss": 0.0013,
+      "step": 10998
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5420840216641225e-06,
+      "loss": 0.002,
+      "step": 10999
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5420421990338973e-06,
+      "loss": 1.4773,
+      "step": 11000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5420003764036719e-06,
+      "loss": 0.0006,
+      "step": 11001
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5419585537734467e-06,
+      "loss": 3.3382,
+      "step": 11002
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5419167311432214e-06,
+      "loss": 0.0105,
+      "step": 11003
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5418749085129962e-06,
+      "loss": 0.0019,
+      "step": 11004
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541833085882771e-06,
+      "loss": 0.0004,
+      "step": 11005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5417912632525458e-06,
+      "loss": 1.7434,
+      "step": 11006
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5417494406223206e-06,
+      "loss": 0.0366,
+      "step": 11007
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5417076179920954e-06,
+      "loss": 0.9753,
+      "step": 11008
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5416657953618702e-06,
+      "loss": 1.1693,
+      "step": 11009
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5416239727316452e-06,
+      "loss": 3.0815,
+      "step": 11010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.54158215010142e-06,
+      "loss": 0.0008,
+      "step": 11011
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5415403274711947e-06,
+      "loss": 0.0006,
+      "step": 11012
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5414985048409695e-06,
+      "loss": 0.0006,
+      "step": 11013
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5414566822107443e-06,
+      "loss": 0.0017,
+      "step": 11014
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5414148595805189e-06,
+      "loss": 0.0243,
+      "step": 11015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5413730369502937e-06,
+      "loss": 0.2957,
+      "step": 11016
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5413312143200685e-06,
+      "loss": 0.0006,
+      "step": 11017
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5412893916898432e-06,
+      "loss": 0.0046,
+      "step": 11018
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541247569059618e-06,
+      "loss": 0.0005,
+      "step": 11019
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5412057464293928e-06,
+      "loss": 0.004,
+      "step": 11020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5411639237991676e-06,
+      "loss": 0.0031,
+      "step": 11021
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5411221011689424e-06,
+      "loss": 0.0008,
+      "step": 11022
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5410802785387172e-06,
+      "loss": 0.0061,
+      "step": 11023
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541038455908492e-06,
+      "loss": 0.0007,
+      "step": 11024
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540996633278267e-06,
+      "loss": 0.0407,
+      "step": 11025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5409548106480417e-06,
+      "loss": 0.0045,
+      "step": 11026
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5409129880178165e-06,
+      "loss": 0.3762,
+      "step": 11027
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5408711653875913e-06,
+      "loss": 0.001,
+      "step": 11028
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5408293427573659e-06,
+      "loss": 0.127,
+      "step": 11029
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5407875201271407e-06,
+      "loss": 0.5727,
+      "step": 11030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5407456974969155e-06,
+      "loss": 2.8037,
+      "step": 11031
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5407038748666902e-06,
+      "loss": 0.218,
+      "step": 11032
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540662052236465e-06,
+      "loss": 0.0009,
+      "step": 11033
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5406202296062398e-06,
+      "loss": 0.0032,
+      "step": 11034
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5405784069760146e-06,
+      "loss": 0.509,
+      "step": 11035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5405365843457894e-06,
+      "loss": 0.0052,
+      "step": 11036
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5404947617155642e-06,
+      "loss": 0.4857,
+      "step": 11037
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540452939085339e-06,
+      "loss": 0.0005,
+      "step": 11038
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540411116455114e-06,
+      "loss": 0.0016,
+      "step": 11039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5403692938248888e-06,
+      "loss": 0.0066,
+      "step": 11040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5403274711946635e-06,
+      "loss": 0.0156,
+      "step": 11041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5402856485644381e-06,
+      "loss": 0.4567,
+      "step": 11042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540243825934213e-06,
+      "loss": 0.0025,
+      "step": 11043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5402020033039877e-06,
+      "loss": 0.0004,
+      "step": 11044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5401601806737625e-06,
+      "loss": 0.0009,
+      "step": 11045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5401183580435373e-06,
+      "loss": 0.0018,
+      "step": 11046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540076535413312e-06,
+      "loss": 2.8829,
+      "step": 11047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5400347127830868e-06,
+      "loss": 0.001,
+      "step": 11048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5399928901528616e-06,
+      "loss": 0.1112,
+      "step": 11049
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5399510675226364e-06,
+      "loss": 0.0006,
+      "step": 11050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5399092448924112e-06,
+      "loss": 0.0343,
+      "step": 11051
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539867422262186e-06,
+      "loss": 0.0005,
+      "step": 11052
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5398255996319608e-06,
+      "loss": 1.2463,
+      "step": 11053
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5397837770017358e-06,
+      "loss": 0.0037,
+      "step": 11054
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5397419543715105e-06,
+      "loss": 0.0007,
+      "step": 11055
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5397001317412851e-06,
+      "loss": 0.0013,
+      "step": 11056
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.53965830911106e-06,
+      "loss": 0.0023,
+      "step": 11057
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5396164864808347e-06,
+      "loss": 0.0007,
+      "step": 11058
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5395746638506095e-06,
+      "loss": 0.002,
+      "step": 11059
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5395328412203843e-06,
+      "loss": 0.2669,
+      "step": 11060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539491018590159e-06,
+      "loss": 0.0023,
+      "step": 11061
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5394491959599338e-06,
+      "loss": 0.1231,
+      "step": 11062
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5394073733297086e-06,
+      "loss": 0.2148,
+      "step": 11063
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5393655506994834e-06,
+      "loss": 0.9576,
+      "step": 11064
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5393237280692582e-06,
+      "loss": 0.0004,
+      "step": 11065
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539281905439033e-06,
+      "loss": 0.011,
+      "step": 11066
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5392400828088078e-06,
+      "loss": 0.0005,
+      "step": 11067
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5391982601785826e-06,
+      "loss": 0.0006,
+      "step": 11068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5391564375483573e-06,
+      "loss": 0.0003,
+      "step": 11069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5391146149181321e-06,
+      "loss": 0.0007,
+      "step": 11070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539072792287907e-06,
+      "loss": 0.0051,
+      "step": 11071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5390309696576817e-06,
+      "loss": 0.0017,
+      "step": 11072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5389891470274565e-06,
+      "loss": 0.0011,
+      "step": 11073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5389473243972313e-06,
+      "loss": 2.538,
+      "step": 11074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538905501767006e-06,
+      "loss": 0.0829,
+      "step": 11075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5388636791367808e-06,
+      "loss": 0.0042,
+      "step": 11076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5388218565065556e-06,
+      "loss": 0.0005,
+      "step": 11077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5387800338763304e-06,
+      "loss": 0.0031,
+      "step": 11078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5387382112461052e-06,
+      "loss": 0.0028,
+      "step": 11079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.53869638861588e-06,
+      "loss": 0.0007,
+      "step": 11080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5386545659856548e-06,
+      "loss": 0.9556,
+      "step": 11081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5386127433554294e-06,
+      "loss": 0.6372,
+      "step": 11082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5385709207252044e-06,
+      "loss": 0.003,
+      "step": 11083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5385290980949791e-06,
+      "loss": 0.5995,
+      "step": 11084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538487275464754e-06,
+      "loss": 0.0005,
+      "step": 11085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5384454528345287e-06,
+      "loss": 0.0006,
+      "step": 11086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5384036302043035e-06,
+      "loss": 3.2459,
+      "step": 11087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5383618075740783e-06,
+      "loss": 0.0011,
+      "step": 11088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538319984943853e-06,
+      "loss": 0.0711,
+      "step": 11089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5382781623136279e-06,
+      "loss": 0.0141,
+      "step": 11090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5382363396834026e-06,
+      "loss": 0.0037,
+      "step": 11091
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5381945170531774e-06,
+      "loss": 0.0014,
+      "step": 11092
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5381526944229522e-06,
+      "loss": 2.8548,
+      "step": 11093
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.538110871792727e-06,
+      "loss": 0.0188,
+      "step": 11094
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5380690491625018e-06,
+      "loss": 0.0005,
+      "step": 11095
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5380272265322764e-06,
+      "loss": 2.2241,
+      "step": 11096
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5379854039020511e-06,
+      "loss": 3.2422,
+      "step": 11097
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5379435812718261e-06,
+      "loss": 0.0011,
+      "step": 11098
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.537901758641601e-06,
+      "loss": 2.2718,
+      "step": 11099
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5378599360113757e-06,
+      "loss": 0.0006,
+      "step": 11100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5378181133811505e-06,
+      "loss": 0.0714,
+      "step": 11101
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5377762907509253e-06,
+      "loss": 0.0014,
+      "step": 11102
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5377344681207e-06,
+      "loss": 0.0038,
+      "step": 11103
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5376926454904749e-06,
+      "loss": 0.0049,
+      "step": 11104
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5376508228602497e-06,
+      "loss": 0.0008,
+      "step": 11105
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5376090002300244e-06,
+      "loss": 0.0028,
+      "step": 11106
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5375671775997992e-06,
+      "loss": 0.0011,
+      "step": 11107
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.537525354969574e-06,
+      "loss": 0.0019,
+      "step": 11108
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5374835323393486e-06,
+      "loss": 0.0004,
+      "step": 11109
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5374417097091234e-06,
+      "loss": 0.0005,
+      "step": 11110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5373998870788982e-06,
+      "loss": 0.7027,
+      "step": 11111
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.537358064448673e-06,
+      "loss": 2.0818,
+      "step": 11112
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.537316241818448e-06,
+      "loss": 0.0011,
+      "step": 11113
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5372744191882227e-06,
+      "loss": 0.0012,
+      "step": 11114
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5372325965579975e-06,
+      "loss": 2.3372,
+      "step": 11115
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5371907739277723e-06,
+      "loss": 0.0004,
+      "step": 11116
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.537148951297547e-06,
+      "loss": 0.0035,
+      "step": 11117
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5371071286673219e-06,
+      "loss": 0.0008,
+      "step": 11118
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5370653060370967e-06,
+      "loss": 0.0008,
+      "step": 11119
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5370234834068714e-06,
+      "loss": 0.0019,
+      "step": 11120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5369816607766462e-06,
+      "loss": 0.0008,
+      "step": 11121
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536939838146421e-06,
+      "loss": 0.0024,
+      "step": 11122
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5368980155161956e-06,
+      "loss": 0.0011,
+      "step": 11123
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5368561928859704e-06,
+      "loss": 0.3607,
+      "step": 11124
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5368143702557452e-06,
+      "loss": 2.3021,
+      "step": 11125
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53677254762552e-06,
+      "loss": 0.1568,
+      "step": 11126
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536730724995295e-06,
+      "loss": 2.5158,
+      "step": 11127
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5366889023650697e-06,
+      "loss": 0.0013,
+      "step": 11128
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5366470797348445e-06,
+      "loss": 0.0005,
+      "step": 11129
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5366052571046193e-06,
+      "loss": 0.8607,
+      "step": 11130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536563434474394e-06,
+      "loss": 0.0018,
+      "step": 11131
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5365216118441689e-06,
+      "loss": 0.001,
+      "step": 11132
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5364797892139437e-06,
+      "loss": 0.2738,
+      "step": 11133
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5364379665837185e-06,
+      "loss": 0.0041,
+      "step": 11134
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5363961439534932e-06,
+      "loss": 0.7153,
+      "step": 11135
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5363543213232678e-06,
+      "loss": 0.0009,
+      "step": 11136
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5363124986930426e-06,
+      "loss": 0.0128,
+      "step": 11137
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5362706760628174e-06,
+      "loss": 0.0009,
+      "step": 11138
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5362288534325922e-06,
+      "loss": 0.0059,
+      "step": 11139
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536187030802367e-06,
+      "loss": 0.0021,
+      "step": 11140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5361452081721417e-06,
+      "loss": 0.0012,
+      "step": 11141
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5361033855419167e-06,
+      "loss": 0.0032,
+      "step": 11142
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5360615629116915e-06,
+      "loss": 0.2929,
+      "step": 11143
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5360197402814663e-06,
+      "loss": 0.8205,
+      "step": 11144
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535977917651241e-06,
+      "loss": 0.484,
+      "step": 11145
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5359360950210159e-06,
+      "loss": 0.2911,
+      "step": 11146
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5358942723907907e-06,
+      "loss": 0.0005,
+      "step": 11147
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5358524497605655e-06,
+      "loss": 0.0017,
+      "step": 11148
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5358106271303402e-06,
+      "loss": 0.3873,
+      "step": 11149
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5357688045001148e-06,
+      "loss": 0.0014,
+      "step": 11150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5357269818698896e-06,
+      "loss": 0.0054,
+      "step": 11151
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5356851592396644e-06,
+      "loss": 0.1541,
+      "step": 11152
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5356433366094392e-06,
+      "loss": 0.0011,
+      "step": 11153
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535601513979214e-06,
+      "loss": 0.0026,
+      "step": 11154
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5355596913489888e-06,
+      "loss": 0.0008,
+      "step": 11155
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5355178687187635e-06,
+      "loss": 0.0014,
+      "step": 11156
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5354760460885385e-06,
+      "loss": 0.0007,
+      "step": 11157
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5354342234583133e-06,
+      "loss": 0.0019,
+      "step": 11158
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5353924008280881e-06,
+      "loss": 0.0012,
+      "step": 11159
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535350578197863e-06,
+      "loss": 0.0008,
+      "step": 11160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5353087555676377e-06,
+      "loss": 0.0013,
+      "step": 11161
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5352669329374125e-06,
+      "loss": 0.9659,
+      "step": 11162
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5352251103071873e-06,
+      "loss": 0.0004,
+      "step": 11163
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5351832876769618e-06,
+      "loss": 0.0061,
+      "step": 11164
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5351414650467366e-06,
+      "loss": 0.0004,
+      "step": 11165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5350996424165114e-06,
+      "loss": 1.5534,
+      "step": 11166
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5350578197862862e-06,
+      "loss": 0.6506,
+      "step": 11167
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535015997156061e-06,
+      "loss": 0.0059,
+      "step": 11168
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5349741745258358e-06,
+      "loss": 0.0009,
+      "step": 11169
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5349323518956106e-06,
+      "loss": 0.0106,
+      "step": 11170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5348905292653855e-06,
+      "loss": 0.0046,
+      "step": 11171
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5348487066351603e-06,
+      "loss": 0.0005,
+      "step": 11172
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5348068840049351e-06,
+      "loss": 0.006,
+      "step": 11173
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53476506137471e-06,
+      "loss": 0.0797,
+      "step": 11174
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5347232387444847e-06,
+      "loss": 0.0705,
+      "step": 11175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5346814161142595e-06,
+      "loss": 0.0005,
+      "step": 11176
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534639593484034e-06,
+      "loss": 0.8421,
+      "step": 11177
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5345977708538088e-06,
+      "loss": 0.0765,
+      "step": 11178
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5345559482235836e-06,
+      "loss": 0.0123,
+      "step": 11179
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5345141255933584e-06,
+      "loss": 3.0044,
+      "step": 11180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5344723029631332e-06,
+      "loss": 0.0039,
+      "step": 11181
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534430480332908e-06,
+      "loss": 0.0021,
+      "step": 11182
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5343886577026828e-06,
+      "loss": 1.9749,
+      "step": 11183
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5343468350724576e-06,
+      "loss": 3.5313,
+      "step": 11184
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5343050124422323e-06,
+      "loss": 0.0037,
+      "step": 11185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5342631898120073e-06,
+      "loss": 0.0645,
+      "step": 11186
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5342213671817821e-06,
+      "loss": 0.0025,
+      "step": 11187
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534179544551557e-06,
+      "loss": 0.002,
+      "step": 11188
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5341377219213317e-06,
+      "loss": 0.0012,
+      "step": 11189
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5340958992911065e-06,
+      "loss": 0.0009,
+      "step": 11190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534054076660881e-06,
+      "loss": 0.0009,
+      "step": 11191
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5340122540306558e-06,
+      "loss": 0.0059,
+      "step": 11192
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5339704314004306e-06,
+      "loss": 0.7233,
+      "step": 11193
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5339286087702054e-06,
+      "loss": 0.006,
+      "step": 11194
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5338867861399802e-06,
+      "loss": 2.5865,
+      "step": 11195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533844963509755e-06,
+      "loss": 0.0012,
+      "step": 11196
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5338031408795298e-06,
+      "loss": 1.0476,
+      "step": 11197
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5337613182493046e-06,
+      "loss": 0.005,
+      "step": 11198
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5337194956190794e-06,
+      "loss": 0.0018,
+      "step": 11199
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5336776729888544e-06,
+      "loss": 0.0011,
+      "step": 11200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5336358503586291e-06,
+      "loss": 0.2435,
+      "step": 11201
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533594027728404e-06,
+      "loss": 0.0013,
+      "step": 11202
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5335522050981787e-06,
+      "loss": 0.0013,
+      "step": 11203
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5335103824679533e-06,
+      "loss": 0.0411,
+      "step": 11204
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533468559837728e-06,
+      "loss": 0.0678,
+      "step": 11205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5334267372075029e-06,
+      "loss": 0.0039,
+      "step": 11206
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5333849145772776e-06,
+      "loss": 0.0027,
+      "step": 11207
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5333430919470524e-06,
+      "loss": 0.0013,
+      "step": 11208
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5333012693168272e-06,
+      "loss": 2.5684,
+      "step": 11209
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533259446686602e-06,
+      "loss": 0.0028,
+      "step": 11210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5332176240563768e-06,
+      "loss": 0.0007,
+      "step": 11211
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5331758014261516e-06,
+      "loss": 0.0036,
+      "step": 11212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5331339787959264e-06,
+      "loss": 0.0007,
+      "step": 11213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5330921561657011e-06,
+      "loss": 0.0239,
+      "step": 11214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5330503335354761e-06,
+      "loss": 0.005,
+      "step": 11215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533008510905251e-06,
+      "loss": 0.0085,
+      "step": 11216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5329666882750257e-06,
+      "loss": 0.5957,
+      "step": 11217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5329248656448003e-06,
+      "loss": 0.0086,
+      "step": 11218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532883043014575e-06,
+      "loss": 0.0008,
+      "step": 11219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5328412203843499e-06,
+      "loss": 0.0005,
+      "step": 11220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5327993977541247e-06,
+      "loss": 0.0013,
+      "step": 11221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5327575751238994e-06,
+      "loss": 0.0006,
+      "step": 11222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5327157524936742e-06,
+      "loss": 1.3197,
+      "step": 11223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532673929863449e-06,
+      "loss": 1.5848,
+      "step": 11224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5326321072332238e-06,
+      "loss": 0.0012,
+      "step": 11225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5325902846029986e-06,
+      "loss": 0.2499,
+      "step": 11226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5325484619727734e-06,
+      "loss": 0.0032,
+      "step": 11227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5325066393425482e-06,
+      "loss": 2.3455,
+      "step": 11228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532464816712323e-06,
+      "loss": 0.003,
+      "step": 11229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532422994082098e-06,
+      "loss": 0.0006,
+      "step": 11230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5323811714518727e-06,
+      "loss": 2.0803,
+      "step": 11231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5323393488216473e-06,
+      "loss": 0.0003,
+      "step": 11232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532297526191422e-06,
+      "loss": 2.1624,
+      "step": 11233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5322557035611969e-06,
+      "loss": 0.0118,
+      "step": 11234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5322138809309717e-06,
+      "loss": 0.0013,
+      "step": 11235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5321720583007464e-06,
+      "loss": 0.004,
+      "step": 11236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5321302356705212e-06,
+      "loss": 0.0011,
+      "step": 11237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532088413040296e-06,
+      "loss": 0.0082,
+      "step": 11238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5320465904100708e-06,
+      "loss": 0.0009,
+      "step": 11239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5320047677798456e-06,
+      "loss": 0.0012,
+      "step": 11240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5319629451496204e-06,
+      "loss": 0.0022,
+      "step": 11241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5319211225193952e-06,
+      "loss": 0.0023,
+      "step": 11242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53187929988917e-06,
+      "loss": 0.0017,
+      "step": 11243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531837477258945e-06,
+      "loss": 0.0053,
+      "step": 11244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5317956546287195e-06,
+      "loss": 0.0044,
+      "step": 11245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5317538319984943e-06,
+      "loss": 0.0111,
+      "step": 11246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531712009368269e-06,
+      "loss": 0.0004,
+      "step": 11247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5316701867380439e-06,
+      "loss": 0.0351,
+      "step": 11248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5316283641078187e-06,
+      "loss": 0.0526,
+      "step": 11249
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5315865414775935e-06,
+      "loss": 0.0006,
+      "step": 11250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5315447188473682e-06,
+      "loss": 2.3978,
+      "step": 11251
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.531502896217143e-06,
+      "loss": 0.5375,
+      "step": 11252
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5314610735869178e-06,
+      "loss": 1.9679,
+      "step": 11253
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5314192509566926e-06,
+      "loss": 0.0553,
+      "step": 11254
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5313774283264674e-06,
+      "loss": 0.0008,
+      "step": 11255
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5313356056962422e-06,
+      "loss": 1.6101,
+      "step": 11256
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.531293783066017e-06,
+      "loss": 0.1901,
+      "step": 11257
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5312519604357915e-06,
+      "loss": 0.3791,
+      "step": 11258
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5312101378055665e-06,
+      "loss": 0.9042,
+      "step": 11259
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5311683151753413e-06,
+      "loss": 0.0024,
+      "step": 11260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5311264925451161e-06,
+      "loss": 0.0012,
+      "step": 11261
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5310846699148909e-06,
+      "loss": 0.642,
+      "step": 11262
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5310428472846657e-06,
+      "loss": 0.7089,
+      "step": 11263
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5310010246544405e-06,
+      "loss": 0.0012,
+      "step": 11264
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5309592020242153e-06,
+      "loss": 0.0026,
+      "step": 11265
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.53091737939399e-06,
+      "loss": 0.0006,
+      "step": 11266
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5308755567637648e-06,
+      "loss": 0.0103,
+      "step": 11267
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5308337341335396e-06,
+      "loss": 0.148,
+      "step": 11268
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5307919115033144e-06,
+      "loss": 0.3071,
+      "step": 11269
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5307500888730892e-06,
+      "loss": 1.1257,
+      "step": 11270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.530708266242864e-06,
+      "loss": 0.0103,
+      "step": 11271
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5306664436126385e-06,
+      "loss": 1.8803,
+      "step": 11272
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5306246209824133e-06,
+      "loss": 0.0019,
+      "step": 11273
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5305827983521883e-06,
+      "loss": 0.0006,
+      "step": 11274
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5305409757219631e-06,
+      "loss": 2.8579,
+      "step": 11275
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.530499153091738e-06,
+      "loss": 0.002,
+      "step": 11276
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5304573304615127e-06,
+      "loss": 1.6475,
+      "step": 11277
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5304155078312875e-06,
+      "loss": 2.7229,
+      "step": 11278
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5303736852010623e-06,
+      "loss": 0.0018,
+      "step": 11279
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.530331862570837e-06,
+      "loss": 0.0013,
+      "step": 11280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5302900399406118e-06,
+      "loss": 0.0007,
+      "step": 11281
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5302482173103866e-06,
+      "loss": 0.0181,
+      "step": 11282
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5302063946801614e-06,
+      "loss": 0.0008,
+      "step": 11283
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5301645720499362e-06,
+      "loss": 0.0029,
+      "step": 11284
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5301227494197108e-06,
+      "loss": 0.0025,
+      "step": 11285
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5300809267894856e-06,
+      "loss": 0.9181,
+      "step": 11286
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5300391041592603e-06,
+      "loss": 0.4893,
+      "step": 11287
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5299972815290353e-06,
+      "loss": 0.0039,
+      "step": 11288
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5299554588988101e-06,
+      "loss": 0.0021,
+      "step": 11289
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529913636268585e-06,
+      "loss": 0.0074,
+      "step": 11290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5298718136383597e-06,
+      "loss": 0.0006,
+      "step": 11291
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5298299910081345e-06,
+      "loss": 0.007,
+      "step": 11292
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5297881683779093e-06,
+      "loss": 0.001,
+      "step": 11293
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529746345747684e-06,
+      "loss": 1.9439,
+      "step": 11294
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5297045231174588e-06,
+      "loss": 0.0063,
+      "step": 11295
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5296627004872336e-06,
+      "loss": 0.0087,
+      "step": 11296
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5296208778570084e-06,
+      "loss": 0.0029,
+      "step": 11297
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5295790552267832e-06,
+      "loss": 1.2049,
+      "step": 11298
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5295372325965578e-06,
+      "loss": 0.457,
+      "step": 11299
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5294954099663326e-06,
+      "loss": 0.0013,
+      "step": 11300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5294535873361073e-06,
+      "loss": 0.0006,
+      "step": 11301
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5294117647058821e-06,
+      "loss": 0.9689,
+      "step": 11302
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5293699420756571e-06,
+      "loss": 0.0017,
+      "step": 11303
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529328119445432e-06,
+      "loss": 0.0603,
+      "step": 11304
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5292862968152067e-06,
+      "loss": 0.0013,
+      "step": 11305
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5292444741849815e-06,
+      "loss": 0.0014,
+      "step": 11306
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5292026515547563e-06,
+      "loss": 0.1839,
+      "step": 11307
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529160828924531e-06,
+      "loss": 0.9557,
+      "step": 11308
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5291190062943059e-06,
+      "loss": 1.415,
+      "step": 11309
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5290771836640806e-06,
+      "loss": 0.0053,
+      "step": 11310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5290353610338554e-06,
+      "loss": 0.0026,
+      "step": 11311
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.52899353840363e-06,
+      "loss": 0.1304,
+      "step": 11312
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5289517157734048e-06,
+      "loss": 0.0034,
+      "step": 11313
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5289098931431796e-06,
+      "loss": 0.0038,
+      "step": 11314
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5288680705129544e-06,
+      "loss": 0.0064,
+      "step": 11315
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5288262478827291e-06,
+      "loss": 0.0022,
+      "step": 11316
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528784425252504e-06,
+      "loss": 0.0032,
+      "step": 11317
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528742602622279e-06,
+      "loss": 0.0012,
+      "step": 11318
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5287007799920537e-06,
+      "loss": 0.5205,
+      "step": 11319
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5286589573618285e-06,
+      "loss": 0.1124,
+      "step": 11320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5286171347316033e-06,
+      "loss": 0.0037,
+      "step": 11321
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528575312101378e-06,
+      "loss": 0.0171,
+      "step": 11322
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5285334894711529e-06,
+      "loss": 2.3636,
+      "step": 11323
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5284916668409276e-06,
+      "loss": 0.008,
+      "step": 11324
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5284498442107024e-06,
+      "loss": 0.0055,
+      "step": 11325
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528408021580477e-06,
+      "loss": 0.0026,
+      "step": 11326
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5283661989502518e-06,
+      "loss": 0.0019,
+      "step": 11327
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5283243763200266e-06,
+      "loss": 0.0013,
+      "step": 11328
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5282825536898014e-06,
+      "loss": 0.0064,
+      "step": 11329
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5282407310595762e-06,
+      "loss": 0.0018,
+      "step": 11330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528198908429351e-06,
+      "loss": 0.0007,
+      "step": 11331
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528157085799126e-06,
+      "loss": 0.0008,
+      "step": 11332
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5281152631689007e-06,
+      "loss": 0.0014,
+      "step": 11333
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5280734405386755e-06,
+      "loss": 0.0011,
+      "step": 11334
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5280316179084503e-06,
+      "loss": 0.0128,
+      "step": 11335
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527989795278225e-06,
+      "loss": 0.7973,
+      "step": 11336
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5279479726479999e-06,
+      "loss": 0.0022,
+      "step": 11337
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5279061500177747e-06,
+      "loss": 0.0006,
+      "step": 11338
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5278643273875494e-06,
+      "loss": 0.001,
+      "step": 11339
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527822504757324e-06,
+      "loss": 0.0038,
+      "step": 11340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5277806821270988e-06,
+      "loss": 0.0007,
+      "step": 11341
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5277388594968736e-06,
+      "loss": 1.7582,
+      "step": 11342
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5276970368666484e-06,
+      "loss": 0.0175,
+      "step": 11343
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5276552142364232e-06,
+      "loss": 2.2899,
+      "step": 11344
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527613391606198e-06,
+      "loss": 0.0015,
+      "step": 11345
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5275715689759727e-06,
+      "loss": 0.0022,
+      "step": 11346
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5275297463457477e-06,
+      "loss": 0.023,
+      "step": 11347
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5274879237155225e-06,
+      "loss": 0.0005,
+      "step": 11348
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5274461010852973e-06,
+      "loss": 0.0042,
+      "step": 11349
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527404278455072e-06,
+      "loss": 0.0052,
+      "step": 11350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5273624558248469e-06,
+      "loss": 1.8662,
+      "step": 11351
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5273206331946217e-06,
+      "loss": 0.0004,
+      "step": 11352
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5272788105643962e-06,
+      "loss": 0.0016,
+      "step": 11353
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527236987934171e-06,
+      "loss": 0.0007,
+      "step": 11354
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5271951653039458e-06,
+      "loss": 0.0088,
+      "step": 11355
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5271533426737206e-06,
+      "loss": 0.0007,
+      "step": 11356
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5271115200434954e-06,
+      "loss": 0.1251,
+      "step": 11357
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5270696974132702e-06,
+      "loss": 0.0014,
+      "step": 11358
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527027874783045e-06,
+      "loss": 1.5464,
+      "step": 11359
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5269860521528197e-06,
+      "loss": 0.0011,
+      "step": 11360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5269442295225947e-06,
+      "loss": 1.8389,
+      "step": 11361
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5269024068923695e-06,
+      "loss": 0.0005,
+      "step": 11362
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5268605842621443e-06,
+      "loss": 0.0041,
+      "step": 11363
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526818761631919e-06,
+      "loss": 0.0118,
+      "step": 11364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5267769390016939e-06,
+      "loss": 0.0005,
+      "step": 11365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5267351163714687e-06,
+      "loss": 0.0038,
+      "step": 11366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5266932937412432e-06,
+      "loss": 0.5935,
+      "step": 11367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526651471111018e-06,
+      "loss": 0.0009,
+      "step": 11368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5266096484807928e-06,
+      "loss": 0.0227,
+      "step": 11369
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5265678258505676e-06,
+      "loss": 0.0009,
+      "step": 11370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5265260032203424e-06,
+      "loss": 0.0005,
+      "step": 11371
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5264841805901172e-06,
+      "loss": 0.0006,
+      "step": 11372
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526442357959892e-06,
+      "loss": 0.0006,
+      "step": 11373
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5264005353296667e-06,
+      "loss": 0.4879,
+      "step": 11374
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5263587126994415e-06,
+      "loss": 0.001,
+      "step": 11375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5263168900692165e-06,
+      "loss": 3.4173,
+      "step": 11376
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5262750674389913e-06,
+      "loss": 0.0095,
+      "step": 11377
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5262332448087661e-06,
+      "loss": 0.0004,
+      "step": 11378
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5261914221785409e-06,
+      "loss": 0.0009,
+      "step": 11379
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5261495995483155e-06,
+      "loss": 0.0721,
+      "step": 11380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5261077769180903e-06,
+      "loss": 0.7895,
+      "step": 11381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526065954287865e-06,
+      "loss": 4.0096,
+      "step": 11382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5260241316576398e-06,
+      "loss": 0.0093,
+      "step": 11383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5259823090274146e-06,
+      "loss": 0.8037,
+      "step": 11384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5259404863971894e-06,
+      "loss": 0.5504,
+      "step": 11385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5258986637669642e-06,
+      "loss": 1.1535,
+      "step": 11386
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525856841136739e-06,
+      "loss": 0.0014,
+      "step": 11387
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5258150185065138e-06,
+      "loss": 0.0026,
+      "step": 11388
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5257731958762885e-06,
+      "loss": 0.0006,
+      "step": 11389
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5257313732460633e-06,
+      "loss": 0.0061,
+      "step": 11390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5256895506158383e-06,
+      "loss": 1.1062,
+      "step": 11391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5256477279856131e-06,
+      "loss": 0.4816,
+      "step": 11392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525605905355388e-06,
+      "loss": 0.0188,
+      "step": 11393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5255640827251625e-06,
+      "loss": 0.0014,
+      "step": 11394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5255222600949373e-06,
+      "loss": 0.0105,
+      "step": 11395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525480437464712e-06,
+      "loss": 0.0091,
+      "step": 11396
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5254386148344868e-06,
+      "loss": 0.006,
+      "step": 11397
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5253967922042616e-06,
+      "loss": 0.0018,
+      "step": 11398
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5253549695740364e-06,
+      "loss": 0.8901,
+      "step": 11399
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5253131469438112e-06,
+      "loss": 0.0015,
+      "step": 11400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525271324313586e-06,
+      "loss": 0.0006,
+      "step": 11401
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5252295016833608e-06,
+      "loss": 1.4795,
+      "step": 11402
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5251876790531356e-06,
+      "loss": 0.0014,
+      "step": 11403
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5251458564229103e-06,
+      "loss": 0.0112,
+      "step": 11404
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5251040337926853e-06,
+      "loss": 0.0046,
+      "step": 11405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5250622111624601e-06,
+      "loss": 0.0008,
+      "step": 11406
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5250203885322347e-06,
+      "loss": 0.008,
+      "step": 11407
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5249785659020095e-06,
+      "loss": 0.0021,
+      "step": 11408
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5249367432717843e-06,
+      "loss": 1.631,
+      "step": 11409
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.524894920641559e-06,
+      "loss": 0.0034,
+      "step": 11410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5248530980113338e-06,
+      "loss": 0.204,
+      "step": 11411
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5248112753811086e-06,
+      "loss": 0.0019,
+      "step": 11412
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5247694527508834e-06,
+      "loss": 0.0015,
+      "step": 11413
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5247276301206582e-06,
+      "loss": 0.5664,
+      "step": 11414
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.524685807490433e-06,
+      "loss": 1.1568,
+      "step": 11415
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5246439848602078e-06,
+      "loss": 0.0061,
+      "step": 11416
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5246021622299826e-06,
+      "loss": 0.0061,
+      "step": 11417
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5245603395997573e-06,
+      "loss": 0.0022,
+      "step": 11418
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5245185169695321e-06,
+      "loss": 0.0114,
+      "step": 11419
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5244766943393071e-06,
+      "loss": 0.0032,
+      "step": 11420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5244348717090817e-06,
+      "loss": 0.0067,
+      "step": 11421
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5243930490788565e-06,
+      "loss": 0.0028,
+      "step": 11422
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5243512264486313e-06,
+      "loss": 0.0016,
+      "step": 11423
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.524309403818406e-06,
+      "loss": 1.0866,
+      "step": 11424
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5242675811881809e-06,
+      "loss": 0.0172,
+      "step": 11425
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5242257585579556e-06,
+      "loss": 0.1129,
+      "step": 11426
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5241839359277304e-06,
+      "loss": 0.0025,
+      "step": 11427
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5241421132975052e-06,
+      "loss": 0.0216,
+      "step": 11428
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.52410029066728e-06,
+      "loss": 1.8685,
+      "step": 11429
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5240584680370548e-06,
+      "loss": 0.0023,
+      "step": 11430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5240166454068296e-06,
+      "loss": 0.002,
+      "step": 11431
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5239748227766044e-06,
+      "loss": 0.0013,
+      "step": 11432
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5239330001463791e-06,
+      "loss": 0.0413,
+      "step": 11433
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5238911775161537e-06,
+      "loss": 0.0011,
+      "step": 11434
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5238493548859287e-06,
+      "loss": 0.0005,
+      "step": 11435
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5238075322557035e-06,
+      "loss": 0.0005,
+      "step": 11436
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5237657096254783e-06,
+      "loss": 0.001,
+      "step": 11437
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.523723886995253e-06,
+      "loss": 0.0011,
+      "step": 11438
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5236820643650279e-06,
+      "loss": 0.0007,
+      "step": 11439
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5236402417348026e-06,
+      "loss": 0.2563,
+      "step": 11440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5235984191045774e-06,
+      "loss": 0.0453,
+      "step": 11441
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5235565964743522e-06,
+      "loss": 0.0015,
+      "step": 11442
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.523514773844127e-06,
+      "loss": 0.0009,
+      "step": 11443
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5234729512139018e-06,
+      "loss": 0.0044,
+      "step": 11444
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5234311285836766e-06,
+      "loss": 0.0022,
+      "step": 11445
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5233893059534514e-06,
+      "loss": 3.0706,
+      "step": 11446
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.523347483323226e-06,
+      "loss": 1.3101,
+      "step": 11447
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5233056606930007e-06,
+      "loss": 0.3516,
+      "step": 11448
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5232638380627757e-06,
+      "loss": 2.0284,
+      "step": 11449
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5232220154325505e-06,
+      "loss": 0.0011,
+      "step": 11450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5231801928023253e-06,
+      "loss": 0.0112,
+      "step": 11451
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5231383701721e-06,
+      "loss": 0.0004,
+      "step": 11452
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5230965475418749e-06,
+      "loss": 0.007,
+      "step": 11453
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5230547249116497e-06,
+      "loss": 0.0065,
+      "step": 11454
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5230129022814244e-06,
+      "loss": 0.0004,
+      "step": 11455
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5229710796511992e-06,
+      "loss": 0.0019,
+      "step": 11456
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522929257020974e-06,
+      "loss": 0.3416,
+      "step": 11457
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5228874343907488e-06,
+      "loss": 0.0009,
+      "step": 11458
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5228456117605236e-06,
+      "loss": 0.2775,
+      "step": 11459
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5228037891302984e-06,
+      "loss": 0.0025,
+      "step": 11460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522761966500073e-06,
+      "loss": 0.0087,
+      "step": 11461
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5227201438698477e-06,
+      "loss": 0.0005,
+      "step": 11462
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5226783212396225e-06,
+      "loss": 0.0058,
+      "step": 11463
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5226364986093975e-06,
+      "loss": 2.6427,
+      "step": 11464
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5225946759791723e-06,
+      "loss": 0.5057,
+      "step": 11465
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522552853348947e-06,
+      "loss": 0.4503,
+      "step": 11466
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5225110307187219e-06,
+      "loss": 0.0019,
+      "step": 11467
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5224692080884967e-06,
+      "loss": 0.0007,
+      "step": 11468
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5224273854582715e-06,
+      "loss": 0.0013,
+      "step": 11469
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5223855628280462e-06,
+      "loss": 0.8013,
+      "step": 11470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522343740197821e-06,
+      "loss": 0.0028,
+      "step": 11471
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5223019175675958e-06,
+      "loss": 0.0007,
+      "step": 11472
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5222600949373706e-06,
+      "loss": 0.0048,
+      "step": 11473
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5222182723071454e-06,
+      "loss": 0.0038,
+      "step": 11474
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.52217644967692e-06,
+      "loss": 0.0003,
+      "step": 11475
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5221346270466947e-06,
+      "loss": 0.0736,
+      "step": 11476
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5220928044164695e-06,
+      "loss": 1.9632,
+      "step": 11477
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5220509817862443e-06,
+      "loss": 0.0004,
+      "step": 11478
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5220091591560193e-06,
+      "loss": 0.9927,
+      "step": 11479
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.521967336525794e-06,
+      "loss": 0.0007,
+      "step": 11480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5219255138955689e-06,
+      "loss": 1.9457,
+      "step": 11481
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5218836912653437e-06,
+      "loss": 1.6723,
+      "step": 11482
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5218418686351185e-06,
+      "loss": 0.0013,
+      "step": 11483
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5218000460048932e-06,
+      "loss": 0.0004,
+      "step": 11484
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.521758223374668e-06,
+      "loss": 0.0019,
+      "step": 11485
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5217164007444428e-06,
+      "loss": 1.6762,
+      "step": 11486
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5216745781142176e-06,
+      "loss": 0.0008,
+      "step": 11487
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5216327554839922e-06,
+      "loss": 1.2651,
+      "step": 11488
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.521590932853767e-06,
+      "loss": 0.001,
+      "step": 11489
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5215491102235418e-06,
+      "loss": 0.005,
+      "step": 11490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5215072875933165e-06,
+      "loss": 0.6422,
+      "step": 11491
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5214654649630913e-06,
+      "loss": 0.0061,
+      "step": 11492
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5214236423328663e-06,
+      "loss": 0.0068,
+      "step": 11493
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5213818197026411e-06,
+      "loss": 0.0092,
+      "step": 11494
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.521339997072416e-06,
+      "loss": 0.0689,
+      "step": 11495
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5212981744421907e-06,
+      "loss": 0.001,
+      "step": 11496
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5212563518119655e-06,
+      "loss": 0.001,
+      "step": 11497
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5212145291817403e-06,
+      "loss": 0.0006,
+      "step": 11498
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.521172706551515e-06,
+      "loss": 0.0015,
+      "step": 11499
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5211308839212898e-06,
+      "loss": 0.0006,
+      "step": 11500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5210890612910646e-06,
+      "loss": 0.003,
+      "step": 11501
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5210472386608392e-06,
+      "loss": 0.0039,
+      "step": 11502
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.521005416030614e-06,
+      "loss": 0.0042,
+      "step": 11503
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5209635934003888e-06,
+      "loss": 0.0012,
+      "step": 11504
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5209217707701635e-06,
+      "loss": 2.356,
+      "step": 11505
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5208799481399383e-06,
+      "loss": 0.0028,
+      "step": 11506
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5208381255097131e-06,
+      "loss": 0.0007,
+      "step": 11507
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5207963028794881e-06,
+      "loss": 0.0154,
+      "step": 11508
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520754480249263e-06,
+      "loss": 0.0006,
+      "step": 11509
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5207126576190377e-06,
+      "loss": 0.0004,
+      "step": 11510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5206708349888125e-06,
+      "loss": 0.0011,
+      "step": 11511
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5206290123585873e-06,
+      "loss": 0.0004,
+      "step": 11512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520587189728362e-06,
+      "loss": 2.7154,
+      "step": 11513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5205453670981368e-06,
+      "loss": 0.0007,
+      "step": 11514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5205035444679114e-06,
+      "loss": 0.1573,
+      "step": 11515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5204617218376862e-06,
+      "loss": 0.3458,
+      "step": 11516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520419899207461e-06,
+      "loss": 0.0016,
+      "step": 11517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5203780765772358e-06,
+      "loss": 0.9566,
+      "step": 11518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5203362539470106e-06,
+      "loss": 0.0058,
+      "step": 11519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5202944313167853e-06,
+      "loss": 0.0061,
+      "step": 11520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5202526086865601e-06,
+      "loss": 0.0087,
+      "step": 11521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520210786056335e-06,
+      "loss": 0.0056,
+      "step": 11522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.52016896342611e-06,
+      "loss": 1.2371,
+      "step": 11523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5201271407958847e-06,
+      "loss": 0.0012,
+      "step": 11524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5200853181656595e-06,
+      "loss": 0.0039,
+      "step": 11525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5200434955354343e-06,
+      "loss": 0.0057,
+      "step": 11526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520001672905209e-06,
+      "loss": 0.0005,
+      "step": 11527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5199598502749838e-06,
+      "loss": 0.0092,
+      "step": 11528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5199180276447584e-06,
+      "loss": 0.0048,
+      "step": 11529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5198762050145332e-06,
+      "loss": 0.0028,
+      "step": 11530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519834382384308e-06,
+      "loss": 0.3287,
+      "step": 11531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5197925597540828e-06,
+      "loss": 0.0122,
+      "step": 11532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5197507371238576e-06,
+      "loss": 2.6752,
+      "step": 11533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5197089144936324e-06,
+      "loss": 1.8043,
+      "step": 11534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5196670918634071e-06,
+      "loss": 0.0021,
+      "step": 11535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519625269233182e-06,
+      "loss": 0.0036,
+      "step": 11536
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519583446602957e-06,
+      "loss": 0.0004,
+      "step": 11537
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5195416239727317e-06,
+      "loss": 1.1492,
+      "step": 11538
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5194998013425065e-06,
+      "loss": 0.0049,
+      "step": 11539
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5194579787122813e-06,
+      "loss": 0.0005,
+      "step": 11540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519416156082056e-06,
+      "loss": 0.0332,
+      "step": 11541
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5193743334518309e-06,
+      "loss": 0.0008,
+      "step": 11542
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5193325108216054e-06,
+      "loss": 0.0011,
+      "step": 11543
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5192906881913802e-06,
+      "loss": 0.0004,
+      "step": 11544
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519248865561155e-06,
+      "loss": 0.0006,
+      "step": 11545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5192070429309298e-06,
+      "loss": 0.0365,
+      "step": 11546
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5191652203007046e-06,
+      "loss": 0.0024,
+      "step": 11547
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5191233976704794e-06,
+      "loss": 0.0005,
+      "step": 11548
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5190815750402541e-06,
+      "loss": 0.0009,
+      "step": 11549
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519039752410029e-06,
+      "loss": 0.0009,
+      "step": 11550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5189979297798037e-06,
+      "loss": 0.0037,
+      "step": 11551
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5189561071495787e-06,
+      "loss": 0.0006,
+      "step": 11552
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5189142845193535e-06,
+      "loss": 0.0014,
+      "step": 11553
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5188724618891283e-06,
+      "loss": 0.0005,
+      "step": 11554
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518830639258903e-06,
+      "loss": 0.0059,
+      "step": 11555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5187888166286777e-06,
+      "loss": 0.0007,
+      "step": 11556
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5187469939984524e-06,
+      "loss": 0.0084,
+      "step": 11557
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5187051713682272e-06,
+      "loss": 0.0081,
+      "step": 11558
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518663348738002e-06,
+      "loss": 0.002,
+      "step": 11559
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5186215261077768e-06,
+      "loss": 0.0061,
+      "step": 11560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5185797034775516e-06,
+      "loss": 1.2087,
+      "step": 11561
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5185378808473264e-06,
+      "loss": 0.0074,
+      "step": 11562
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5184960582171012e-06,
+      "loss": 0.001,
+      "step": 11563
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518454235586876e-06,
+      "loss": 0.0009,
+      "step": 11564
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5184124129566507e-06,
+      "loss": 1.5595,
+      "step": 11565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5183705903264257e-06,
+      "loss": 0.0004,
+      "step": 11566
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5183287676962005e-06,
+      "loss": 0.0017,
+      "step": 11567
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5182869450659753e-06,
+      "loss": 0.0037,
+      "step": 11568
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.51824512243575e-06,
+      "loss": 0.0014,
+      "step": 11569
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5182032998055247e-06,
+      "loss": 0.0004,
+      "step": 11570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5181614771752994e-06,
+      "loss": 0.0185,
+      "step": 11571
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5181196545450742e-06,
+      "loss": 0.0017,
+      "step": 11572
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.518077831914849e-06,
+      "loss": 0.0005,
+      "step": 11573
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5180360092846238e-06,
+      "loss": 0.0035,
+      "step": 11574
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5179941866543986e-06,
+      "loss": 3.7266,
+      "step": 11575
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5179523640241734e-06,
+      "loss": 0.225,
+      "step": 11576
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5179105413939482e-06,
+      "loss": 0.0012,
+      "step": 11577
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.517868718763723e-06,
+      "loss": 0.0009,
+      "step": 11578
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5178268961334977e-06,
+      "loss": 0.0019,
+      "step": 11579
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5177850735032725e-06,
+      "loss": 0.001,
+      "step": 11580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5177432508730475e-06,
+      "loss": 0.5751,
+      "step": 11581
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5177014282428223e-06,
+      "loss": 0.0013,
+      "step": 11582
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5176596056125969e-06,
+      "loss": 0.1755,
+      "step": 11583
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5176177829823717e-06,
+      "loss": 0.024,
+      "step": 11584
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5175759603521465e-06,
+      "loss": 0.0005,
+      "step": 11585
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5175341377219212e-06,
+      "loss": 0.0004,
+      "step": 11586
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.517492315091696e-06,
+      "loss": 1.5502,
+      "step": 11587
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5174504924614708e-06,
+      "loss": 0.2132,
+      "step": 11588
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5174086698312456e-06,
+      "loss": 0.0009,
+      "step": 11589
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5173668472010204e-06,
+      "loss": 0.0012,
+      "step": 11590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5173250245707952e-06,
+      "loss": 0.3092,
+      "step": 11591
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.51728320194057e-06,
+      "loss": 0.8862,
+      "step": 11592
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5172413793103447e-06,
+      "loss": 0.0064,
+      "step": 11593
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5171995566801195e-06,
+      "loss": 0.0011,
+      "step": 11594
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5171577340498943e-06,
+      "loss": 0.9529,
+      "step": 11595
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5171159114196693e-06,
+      "loss": 0.0003,
+      "step": 11596
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5170740887894439e-06,
+      "loss": 0.0005,
+      "step": 11597
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5170322661592187e-06,
+      "loss": 0.025,
+      "step": 11598
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5169904435289935e-06,
+      "loss": 0.1842,
+      "step": 11599
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5169486208987682e-06,
+      "loss": 0.3479,
+      "step": 11600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.516906798268543e-06,
+      "loss": 2.0507,
+      "step": 11601
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5168649756383178e-06,
+      "loss": 0.0166,
+      "step": 11602
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5168231530080926e-06,
+      "loss": 0.0178,
+      "step": 11603
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5167813303778674e-06,
+      "loss": 0.0047,
+      "step": 11604
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5167395077476422e-06,
+      "loss": 0.005,
+      "step": 11605
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.516697685117417e-06,
+      "loss": 0.0008,
+      "step": 11606
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5166558624871918e-06,
+      "loss": 0.0006,
+      "step": 11607
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5166140398569665e-06,
+      "loss": 1.0968,
+      "step": 11608
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5165722172267413e-06,
+      "loss": 0.0006,
+      "step": 11609
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5165303945965161e-06,
+      "loss": 1.5143,
+      "step": 11610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.516488571966291e-06,
+      "loss": 0.013,
+      "step": 11611
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5164467493360657e-06,
+      "loss": 0.0037,
+      "step": 11612
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5164049267058405e-06,
+      "loss": 0.0023,
+      "step": 11613
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5163631040756153e-06,
+      "loss": 0.3144,
+      "step": 11614
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.51632128144539e-06,
+      "loss": 0.0057,
+      "step": 11615
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5162794588151648e-06,
+      "loss": 0.0008,
+      "step": 11616
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5162376361849396e-06,
+      "loss": 0.0009,
+      "step": 11617
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5161958135547144e-06,
+      "loss": 0.0162,
+      "step": 11618
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5161539909244892e-06,
+      "loss": 0.1572,
+      "step": 11619
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.516112168294264e-06,
+      "loss": 0.0212,
+      "step": 11620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5160703456640388e-06,
+      "loss": 0.115,
+      "step": 11621
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5160285230338135e-06,
+      "loss": 1.4223,
+      "step": 11622
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5159867004035881e-06,
+      "loss": 0.0031,
+      "step": 11623
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.515944877773363e-06,
+      "loss": 0.0014,
+      "step": 11624
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.515903055143138e-06,
+      "loss": 0.0033,
+      "step": 11625
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5158612325129127e-06,
+      "loss": 0.014,
+      "step": 11626
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5158194098826875e-06,
+      "loss": 1.108,
+      "step": 11627
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5157775872524623e-06,
+      "loss": 0.0009,
+      "step": 11628
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.515735764622237e-06,
+      "loss": 0.0009,
+      "step": 11629
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5156939419920118e-06,
+      "loss": 0.0006,
+      "step": 11630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5156521193617866e-06,
+      "loss": 0.0015,
+      "step": 11631
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5156102967315614e-06,
+      "loss": 0.0165,
+      "step": 11632
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5155684741013362e-06,
+      "loss": 0.0019,
+      "step": 11633
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.515526651471111e-06,
+      "loss": 0.0008,
+      "step": 11634
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5154848288408858e-06,
+      "loss": 1.6718,
+      "step": 11635
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5154430062106606e-06,
+      "loss": 0.0046,
+      "step": 11636
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5154011835804351e-06,
+      "loss": 0.0021,
+      "step": 11637
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.51535936095021e-06,
+      "loss": 0.0014,
+      "step": 11638
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5153175383199847e-06,
+      "loss": 0.0004,
+      "step": 11639
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5152757156897597e-06,
+      "loss": 0.0006,
+      "step": 11640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5152338930595345e-06,
+      "loss": 2.4935,
+      "step": 11641
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5151920704293093e-06,
+      "loss": 0.0028,
+      "step": 11642
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.515150247799084e-06,
+      "loss": 0.0027,
+      "step": 11643
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5151084251688588e-06,
+      "loss": 0.0011,
+      "step": 11644
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5150666025386336e-06,
+      "loss": 2.5374,
+      "step": 11645
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5150247799084084e-06,
+      "loss": 0.0006,
+      "step": 11646
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5149829572781832e-06,
+      "loss": 0.0004,
+      "step": 11647
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.514941134647958e-06,
+      "loss": 0.002,
+      "step": 11648
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5148993120177328e-06,
+      "loss": 0.4137,
+      "step": 11649
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5148574893875074e-06,
+      "loss": 0.1337,
+      "step": 11650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5148156667572821e-06,
+      "loss": 4.303,
+      "step": 11651
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.514773844127057e-06,
+      "loss": 0.0006,
+      "step": 11652
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5147320214968317e-06,
+      "loss": 0.0015,
+      "step": 11653
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5146901988666067e-06,
+      "loss": 0.8846,
+      "step": 11654
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5146483762363815e-06,
+      "loss": 0.0019,
+      "step": 11655
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5146065536061563e-06,
+      "loss": 0.0832,
+      "step": 11656
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.514564730975931e-06,
+      "loss": 0.6058,
+      "step": 11657
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5145229083457059e-06,
+      "loss": 0.0013,
+      "step": 11658
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5144810857154806e-06,
+      "loss": 0.0005,
+      "step": 11659
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5144392630852554e-06,
+      "loss": 1.3115,
+      "step": 11660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5143974404550302e-06,
+      "loss": 2.3452,
+      "step": 11661
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.514355617824805e-06,
+      "loss": 1.3002,
+      "step": 11662
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5143137951945798e-06,
+      "loss": 2.4064,
+      "step": 11663
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5142719725643544e-06,
+      "loss": 0.0005,
+      "step": 11664
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5142301499341291e-06,
+      "loss": 2.116,
+      "step": 11665
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.514188327303904e-06,
+      "loss": 0.0009,
+      "step": 11666
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5141465046736787e-06,
+      "loss": 0.004,
+      "step": 11667
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5141046820434535e-06,
+      "loss": 0.0047,
+      "step": 11668
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5140628594132285e-06,
+      "loss": 1.5214,
+      "step": 11669
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5140210367830033e-06,
+      "loss": 0.0003,
+      "step": 11670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513979214152778e-06,
+      "loss": 2.7663,
+      "step": 11671
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5139373915225529e-06,
+      "loss": 0.4068,
+      "step": 11672
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5138955688923277e-06,
+      "loss": 0.005,
+      "step": 11673
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5138537462621024e-06,
+      "loss": 0.0059,
+      "step": 11674
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5138119236318772e-06,
+      "loss": 0.3163,
+      "step": 11675
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513770101001652e-06,
+      "loss": 0.0007,
+      "step": 11676
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5137282783714268e-06,
+      "loss": 0.0009,
+      "step": 11677
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5136864557412014e-06,
+      "loss": 0.0016,
+      "step": 11678
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5136446331109762e-06,
+      "loss": 1.5826,
+      "step": 11679
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513602810480751e-06,
+      "loss": 0.0051,
+      "step": 11680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5135609878505257e-06,
+      "loss": 0.186,
+      "step": 11681
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5135191652203005e-06,
+      "loss": 0.0005,
+      "step": 11682
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5134773425900753e-06,
+      "loss": 0.0022,
+      "step": 11683
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5134355199598503e-06,
+      "loss": 0.2517,
+      "step": 11684
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513393697329625e-06,
+      "loss": 0.0602,
+      "step": 11685
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5133518746993999e-06,
+      "loss": 0.0006,
+      "step": 11686
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5133100520691747e-06,
+      "loss": 0.0059,
+      "step": 11687
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5132682294389494e-06,
+      "loss": 0.0211,
+      "step": 11688
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5132264068087242e-06,
+      "loss": 0.0018,
+      "step": 11689
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513184584178499e-06,
+      "loss": 0.001,
+      "step": 11690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5131427615482736e-06,
+      "loss": 0.0054,
+      "step": 11691
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5131009389180484e-06,
+      "loss": 0.0018,
+      "step": 11692
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5130591162878232e-06,
+      "loss": 0.0404,
+      "step": 11693
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513017293657598e-06,
+      "loss": 0.6812,
+      "step": 11694
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5129754710273727e-06,
+      "loss": 0.0055,
+      "step": 11695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5129336483971475e-06,
+      "loss": 1.7136,
+      "step": 11696
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5128918257669223e-06,
+      "loss": 0.2099,
+      "step": 11697
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5128500031366973e-06,
+      "loss": 0.4975,
+      "step": 11698
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512808180506472e-06,
+      "loss": 0.0013,
+      "step": 11699
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5127663578762469e-06,
+      "loss": 0.0008,
+      "step": 11700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5127245352460217e-06,
+      "loss": 0.0031,
+      "step": 11701
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5126827126157965e-06,
+      "loss": 0.0041,
+      "step": 11702
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5126408899855712e-06,
+      "loss": 0.2524,
+      "step": 11703
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512599067355346e-06,
+      "loss": 0.0492,
+      "step": 11704
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5125572447251206e-06,
+      "loss": 0.1831,
+      "step": 11705
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5125154220948954e-06,
+      "loss": 0.0011,
+      "step": 11706
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5124735994646702e-06,
+      "loss": 0.0008,
+      "step": 11707
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512431776834445e-06,
+      "loss": 0.0023,
+      "step": 11708
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5123899542042197e-06,
+      "loss": 0.3274,
+      "step": 11709
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5123481315739945e-06,
+      "loss": 0.8108,
+      "step": 11710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5123063089437693e-06,
+      "loss": 0.0028,
+      "step": 11711
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512264486313544e-06,
+      "loss": 0.0079,
+      "step": 11712
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512222663683319e-06,
+      "loss": 0.0107,
+      "step": 11713
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5121808410530939e-06,
+      "loss": 0.6857,
+      "step": 11714
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5121390184228687e-06,
+      "loss": 0.0021,
+      "step": 11715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5120971957926435e-06,
+      "loss": 0.0016,
+      "step": 11716
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5120553731624182e-06,
+      "loss": 0.0027,
+      "step": 11717
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5120135505321928e-06,
+      "loss": 0.0066,
+      "step": 11718
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5119717279019676e-06,
+      "loss": 0.0109,
+      "step": 11719
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5119299052717424e-06,
+      "loss": 0.0891,
+      "step": 11720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5118880826415172e-06,
+      "loss": 0.0011,
+      "step": 11721
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.511846260011292e-06,
+      "loss": 0.0029,
+      "step": 11722
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5118044373810668e-06,
+      "loss": 0.0021,
+      "step": 11723
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5117626147508415e-06,
+      "loss": 0.2885,
+      "step": 11724
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5117207921206163e-06,
+      "loss": 0.0013,
+      "step": 11725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5116789694903911e-06,
+      "loss": 1.5916,
+      "step": 11726
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.511637146860166e-06,
+      "loss": 0.1933,
+      "step": 11727
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.511595324229941e-06,
+      "loss": 0.0301,
+      "step": 11728
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5115535015997157e-06,
+      "loss": 0.0057,
+      "step": 11729
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5115116789694905e-06,
+      "loss": 2.0792,
+      "step": 11730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5114698563392653e-06,
+      "loss": 0.0034,
+      "step": 11731
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5114280337090398e-06,
+      "loss": 1.9296,
+      "step": 11732
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5113862110788146e-06,
+      "loss": 0.0009,
+      "step": 11733
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5113443884485894e-06,
+      "loss": 1.5937,
+      "step": 11734
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5113025658183642e-06,
+      "loss": 0.0006,
+      "step": 11735
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.511260743188139e-06,
+      "loss": 0.0042,
+      "step": 11736
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5112189205579138e-06,
+      "loss": 0.0014,
+      "step": 11737
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5111770979276886e-06,
+      "loss": 0.0027,
+      "step": 11738
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5111352752974633e-06,
+      "loss": 0.0071,
+      "step": 11739
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5110934526672381e-06,
+      "loss": 0.0077,
+      "step": 11740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.511051630037013e-06,
+      "loss": 0.0068,
+      "step": 11741
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.511009807406788e-06,
+      "loss": 0.0037,
+      "step": 11742
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5109679847765627e-06,
+      "loss": 0.0026,
+      "step": 11743
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5109261621463375e-06,
+      "loss": 1.3854,
+      "step": 11744
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5108843395161123e-06,
+      "loss": 0.0005,
+      "step": 11745
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5108425168858868e-06,
+      "loss": 0.016,
+      "step": 11746
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5108006942556616e-06,
+      "loss": 0.0033,
+      "step": 11747
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5107588716254364e-06,
+      "loss": 0.2148,
+      "step": 11748
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5107170489952112e-06,
+      "loss": 1.6799,
+      "step": 11749
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.510675226364986e-06,
+      "loss": 0.0005,
+      "step": 11750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5106334037347608e-06,
+      "loss": 0.0022,
+      "step": 11751
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5105915811045356e-06,
+      "loss": 0.001,
+      "step": 11752
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5105497584743103e-06,
+      "loss": 1.2911,
+      "step": 11753
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5105079358440851e-06,
+      "loss": 2.2952,
+      "step": 11754
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.51046611321386e-06,
+      "loss": 0.0058,
+      "step": 11755
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5104242905836347e-06,
+      "loss": 0.0008,
+      "step": 11756
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5103824679534097e-06,
+      "loss": 0.39,
+      "step": 11757
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5103406453231845e-06,
+      "loss": 0.0031,
+      "step": 11758
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.510298822692959e-06,
+      "loss": 0.0009,
+      "step": 11759
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5102570000627338e-06,
+      "loss": 0.1195,
+      "step": 11760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5102151774325086e-06,
+      "loss": 0.0028,
+      "step": 11761
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5101733548022834e-06,
+      "loss": 0.0033,
+      "step": 11762
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5101315321720582e-06,
+      "loss": 0.0041,
+      "step": 11763
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.510089709541833e-06,
+      "loss": 0.0045,
+      "step": 11764
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5100478869116078e-06,
+      "loss": 0.0025,
+      "step": 11765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5100060642813826e-06,
+      "loss": 0.0616,
+      "step": 11766
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5099642416511574e-06,
+      "loss": 0.0016,
+      "step": 11767
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5099224190209321e-06,
+      "loss": 0.0003,
+      "step": 11768
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509880596390707e-06,
+      "loss": 0.0009,
+      "step": 11769
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5098387737604817e-06,
+      "loss": 0.0008,
+      "step": 11770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5097969511302567e-06,
+      "loss": 0.0009,
+      "step": 11771
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5097551285000315e-06,
+      "loss": 0.0011,
+      "step": 11772
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509713305869806e-06,
+      "loss": 0.0031,
+      "step": 11773
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5096714832395809e-06,
+      "loss": 0.0019,
+      "step": 11774
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5096296606093556e-06,
+      "loss": 0.0008,
+      "step": 11775
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5095878379791304e-06,
+      "loss": 0.0005,
+      "step": 11776
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5095460153489052e-06,
+      "loss": 0.0005,
+      "step": 11777
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.50950419271868e-06,
+      "loss": 0.1227,
+      "step": 11778
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5094623700884548e-06,
+      "loss": 0.0059,
+      "step": 11779
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5094205474582296e-06,
+      "loss": 0.0034,
+      "step": 11780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5093787248280044e-06,
+      "loss": 0.0005,
+      "step": 11781
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5093369021977791e-06,
+      "loss": 0.0004,
+      "step": 11782
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509295079567554e-06,
+      "loss": 0.0006,
+      "step": 11783
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5092532569373287e-06,
+      "loss": 0.9135,
+      "step": 11784
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5092114343071035e-06,
+      "loss": 0.0042,
+      "step": 11785
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5091696116768783e-06,
+      "loss": 0.9037,
+      "step": 11786
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509127789046653e-06,
+      "loss": 0.0029,
+      "step": 11787
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5090859664164279e-06,
+      "loss": 0.0024,
+      "step": 11788
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5090441437862027e-06,
+      "loss": 1.0308,
+      "step": 11789
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5090023211559774e-06,
+      "loss": 0.0032,
+      "step": 11790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5089604985257522e-06,
+      "loss": 0.2494,
+      "step": 11791
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508918675895527e-06,
+      "loss": 1.134,
+      "step": 11792
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5088768532653018e-06,
+      "loss": 0.3913,
+      "step": 11793
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5088350306350766e-06,
+      "loss": 0.0639,
+      "step": 11794
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5087932080048514e-06,
+      "loss": 1.5521,
+      "step": 11795
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5087513853746262e-06,
+      "loss": 1.7288,
+      "step": 11796
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508709562744401e-06,
+      "loss": 0.0087,
+      "step": 11797
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5086677401141757e-06,
+      "loss": 0.0179,
+      "step": 11798
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5086259174839503e-06,
+      "loss": 0.463,
+      "step": 11799
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508584094853725e-06,
+      "loss": 0.1584,
+      "step": 11800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5085422722235e-06,
+      "loss": 0.0247,
+      "step": 11801
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5085004495932749e-06,
+      "loss": 0.0055,
+      "step": 11802
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5084586269630497e-06,
+      "loss": 0.0019,
+      "step": 11803
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5084168043328244e-06,
+      "loss": 0.0033,
+      "step": 11804
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5083749817025992e-06,
+      "loss": 3.0789,
+      "step": 11805
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508333159072374e-06,
+      "loss": 0.0013,
+      "step": 11806
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5082913364421488e-06,
+      "loss": 0.0009,
+      "step": 11807
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5082495138119236e-06,
+      "loss": 3.3194,
+      "step": 11808
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5082076911816984e-06,
+      "loss": 3.6037,
+      "step": 11809
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5081658685514732e-06,
+      "loss": 0.001,
+      "step": 11810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508124045921248e-06,
+      "loss": 0.001,
+      "step": 11811
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5080822232910227e-06,
+      "loss": 0.0052,
+      "step": 11812
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5080404006607973e-06,
+      "loss": 0.0025,
+      "step": 11813
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.507998578030572e-06,
+      "loss": 1.2921,
+      "step": 11814
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.507956755400347e-06,
+      "loss": 0.0015,
+      "step": 11815
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5079149327701219e-06,
+      "loss": 0.0037,
+      "step": 11816
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5078731101398967e-06,
+      "loss": 1.9444,
+      "step": 11817
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5078312875096715e-06,
+      "loss": 0.001,
+      "step": 11818
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5077894648794462e-06,
+      "loss": 0.0055,
+      "step": 11819
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.507747642249221e-06,
+      "loss": 0.0685,
+      "step": 11820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5077058196189958e-06,
+      "loss": 1.5109,
+      "step": 11821
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5076639969887706e-06,
+      "loss": 0.0009,
+      "step": 11822
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5076221743585454e-06,
+      "loss": 0.0068,
+      "step": 11823
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5075803517283202e-06,
+      "loss": 0.1862,
+      "step": 11824
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.507538529098095e-06,
+      "loss": 0.0031,
+      "step": 11825
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5074967064678695e-06,
+      "loss": 0.0976,
+      "step": 11826
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5074548838376443e-06,
+      "loss": 0.4136,
+      "step": 11827
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5074130612074191e-06,
+      "loss": 0.0007,
+      "step": 11828
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.507371238577194e-06,
+      "loss": 0.0009,
+      "step": 11829
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5073294159469689e-06,
+      "loss": 0.0008,
+      "step": 11830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5072875933167437e-06,
+      "loss": 0.223,
+      "step": 11831
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5072457706865185e-06,
+      "loss": 0.347,
+      "step": 11832
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5072039480562933e-06,
+      "loss": 0.0004,
+      "step": 11833
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.507162125426068e-06,
+      "loss": 0.7625,
+      "step": 11834
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5071203027958428e-06,
+      "loss": 0.0012,
+      "step": 11835
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5070784801656176e-06,
+      "loss": 0.0029,
+      "step": 11836
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5070366575353924e-06,
+      "loss": 0.0079,
+      "step": 11837
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5069948349051672e-06,
+      "loss": 0.0025,
+      "step": 11838
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506953012274942e-06,
+      "loss": 0.0053,
+      "step": 11839
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5069111896447165e-06,
+      "loss": 0.0019,
+      "step": 11840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5068693670144913e-06,
+      "loss": 0.0011,
+      "step": 11841
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5068275443842661e-06,
+      "loss": 0.0177,
+      "step": 11842
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506785721754041e-06,
+      "loss": 0.299,
+      "step": 11843
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5067438991238157e-06,
+      "loss": 1.4267,
+      "step": 11844
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5067020764935907e-06,
+      "loss": 1.7681,
+      "step": 11845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5066602538633655e-06,
+      "loss": 0.0025,
+      "step": 11846
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5066184312331403e-06,
+      "loss": 2.4384,
+      "step": 11847
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506576608602915e-06,
+      "loss": 0.0021,
+      "step": 11848
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5065347859726898e-06,
+      "loss": 0.2191,
+      "step": 11849
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5064929633424646e-06,
+      "loss": 0.0009,
+      "step": 11850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5064511407122394e-06,
+      "loss": 0.5082,
+      "step": 11851
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5064093180820142e-06,
+      "loss": 1.0461,
+      "step": 11852
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5063674954517888e-06,
+      "loss": 0.0084,
+      "step": 11853
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5063256728215636e-06,
+      "loss": 0.0004,
+      "step": 11854
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5062838501913383e-06,
+      "loss": 0.0006,
+      "step": 11855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5062420275611131e-06,
+      "loss": 0.0018,
+      "step": 11856
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506200204930888e-06,
+      "loss": 0.0017,
+      "step": 11857
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5061583823006627e-06,
+      "loss": 0.0165,
+      "step": 11858
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5061165596704377e-06,
+      "loss": 0.0032,
+      "step": 11859
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5060747370402125e-06,
+      "loss": 0.0078,
+      "step": 11860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5060329144099873e-06,
+      "loss": 0.0028,
+      "step": 11861
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505991091779762e-06,
+      "loss": 2.8513,
+      "step": 11862
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5059492691495368e-06,
+      "loss": 0.0011,
+      "step": 11863
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5059074465193116e-06,
+      "loss": 1.6582,
+      "step": 11864
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5058656238890864e-06,
+      "loss": 0.0013,
+      "step": 11865
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5058238012588612e-06,
+      "loss": 0.4586,
+      "step": 11866
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5057819786286358e-06,
+      "loss": 0.0015,
+      "step": 11867
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5057401559984106e-06,
+      "loss": 0.0052,
+      "step": 11868
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5056983333681853e-06,
+      "loss": 0.5845,
+      "step": 11869
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5056565107379601e-06,
+      "loss": 0.0006,
+      "step": 11870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505614688107735e-06,
+      "loss": 0.0122,
+      "step": 11871
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5055728654775097e-06,
+      "loss": 0.0028,
+      "step": 11872
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5055310428472845e-06,
+      "loss": 0.0015,
+      "step": 11873
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5054892202170595e-06,
+      "loss": 0.0004,
+      "step": 11874
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5054473975868343e-06,
+      "loss": 0.0006,
+      "step": 11875
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505405574956609e-06,
+      "loss": 0.0004,
+      "step": 11876
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5053637523263839e-06,
+      "loss": 2.1591,
+      "step": 11877
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5053219296961586e-06,
+      "loss": 0.0042,
+      "step": 11878
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5052801070659334e-06,
+      "loss": 0.0005,
+      "step": 11879
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5052382844357082e-06,
+      "loss": 0.0078,
+      "step": 11880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5051964618054828e-06,
+      "loss": 0.7973,
+      "step": 11881
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5051546391752576e-06,
+      "loss": 0.4441,
+      "step": 11882
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5051128165450324e-06,
+      "loss": 0.3009,
+      "step": 11883
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5050709939148071e-06,
+      "loss": 0.0021,
+      "step": 11884
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505029171284582e-06,
+      "loss": 0.0013,
+      "step": 11885
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5049873486543567e-06,
+      "loss": 0.0005,
+      "step": 11886
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5049455260241315e-06,
+      "loss": 0.0018,
+      "step": 11887
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5049037033939063e-06,
+      "loss": 0.0023,
+      "step": 11888
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5048618807636813e-06,
+      "loss": 0.0109,
+      "step": 11889
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.504820058133456e-06,
+      "loss": 0.0033,
+      "step": 11890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5047782355032309e-06,
+      "loss": 0.0003,
+      "step": 11891
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5047364128730056e-06,
+      "loss": 0.002,
+      "step": 11892
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5046945902427804e-06,
+      "loss": 1.5156,
+      "step": 11893
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.504652767612555e-06,
+      "loss": 0.0012,
+      "step": 11894
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5046109449823298e-06,
+      "loss": 2.2057,
+      "step": 11895
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5045691223521046e-06,
+      "loss": 0.7572,
+      "step": 11896
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5045272997218794e-06,
+      "loss": 2.1848,
+      "step": 11897
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5044854770916542e-06,
+      "loss": 0.1166,
+      "step": 11898
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.504443654461429e-06,
+      "loss": 0.0258,
+      "step": 11899
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5044018318312037e-06,
+      "loss": 0.0005,
+      "step": 11900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5043600092009785e-06,
+      "loss": 0.0009,
+      "step": 11901
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5043181865707533e-06,
+      "loss": 0.005,
+      "step": 11902
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5042763639405283e-06,
+      "loss": 0.0019,
+      "step": 11903
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.504234541310303e-06,
+      "loss": 0.1279,
+      "step": 11904
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5041927186800779e-06,
+      "loss": 0.0008,
+      "step": 11905
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5041508960498527e-06,
+      "loss": 0.0044,
+      "step": 11906
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5041090734196274e-06,
+      "loss": 0.0008,
+      "step": 11907
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.504067250789402e-06,
+      "loss": 0.0014,
+      "step": 11908
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5040254281591768e-06,
+      "loss": 0.001,
+      "step": 11909
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5039836055289516e-06,
+      "loss": 0.0059,
+      "step": 11910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5039417828987264e-06,
+      "loss": 0.002,
+      "step": 11911
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5038999602685012e-06,
+      "loss": 0.0525,
+      "step": 11912
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503858137638276e-06,
+      "loss": 0.0016,
+      "step": 11913
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5038163150080507e-06,
+      "loss": 0.8039,
+      "step": 11914
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5037744923778255e-06,
+      "loss": 0.0153,
+      "step": 11915
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5037326697476003e-06,
+      "loss": 0.0007,
+      "step": 11916
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503690847117375e-06,
+      "loss": 2.8657,
+      "step": 11917
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.50364902448715e-06,
+      "loss": 0.0212,
+      "step": 11918
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5036072018569249e-06,
+      "loss": 1.743,
+      "step": 11919
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5035653792266997e-06,
+      "loss": 1.6429,
+      "step": 11920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5035235565964742e-06,
+      "loss": 2.5425,
+      "step": 11921
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503481733966249e-06,
+      "loss": 0.0105,
+      "step": 11922
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5034399113360238e-06,
+      "loss": 0.0009,
+      "step": 11923
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5033980887057986e-06,
+      "loss": 0.0022,
+      "step": 11924
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5033562660755734e-06,
+      "loss": 0.0005,
+      "step": 11925
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5033144434453482e-06,
+      "loss": 0.0034,
+      "step": 11926
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503272620815123e-06,
+      "loss": 0.3681,
+      "step": 11927
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5032307981848977e-06,
+      "loss": 0.2117,
+      "step": 11928
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5031889755546725e-06,
+      "loss": 0.0005,
+      "step": 11929
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5031471529244473e-06,
+      "loss": 0.0013,
+      "step": 11930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503105330294222e-06,
+      "loss": 0.0016,
+      "step": 11931
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5030635076639969e-06,
+      "loss": 0.0623,
+      "step": 11932
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5030216850337719e-06,
+      "loss": 0.0282,
+      "step": 11933
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5029798624035467e-06,
+      "loss": 0.0103,
+      "step": 11934
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5029380397733212e-06,
+      "loss": 0.0011,
+      "step": 11935
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502896217143096e-06,
+      "loss": 0.4141,
+      "step": 11936
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5028543945128708e-06,
+      "loss": 0.0005,
+      "step": 11937
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5028125718826456e-06,
+      "loss": 0.0367,
+      "step": 11938
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5027707492524204e-06,
+      "loss": 0.0034,
+      "step": 11939
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5027289266221952e-06,
+      "loss": 0.0439,
+      "step": 11940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.50268710399197e-06,
+      "loss": 0.0012,
+      "step": 11941
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5026452813617447e-06,
+      "loss": 0.0021,
+      "step": 11942
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5026034587315195e-06,
+      "loss": 0.0053,
+      "step": 11943
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5025616361012943e-06,
+      "loss": 0.0018,
+      "step": 11944
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5025198134710691e-06,
+      "loss": 0.008,
+      "step": 11945
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502477990840844e-06,
+      "loss": 0.0085,
+      "step": 11946
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5024361682106189e-06,
+      "loss": 0.0041,
+      "step": 11947
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5023943455803937e-06,
+      "loss": 0.002,
+      "step": 11948
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5023525229501683e-06,
+      "loss": 0.053,
+      "step": 11949
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502310700319943e-06,
+      "loss": 0.0067,
+      "step": 11950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5022688776897178e-06,
+      "loss": 0.0043,
+      "step": 11951
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5022270550594926e-06,
+      "loss": 0.0004,
+      "step": 11952
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5021852324292674e-06,
+      "loss": 0.0026,
+      "step": 11953
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5021434097990422e-06,
+      "loss": 3.025,
+      "step": 11954
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502101587168817e-06,
+      "loss": 0.0049,
+      "step": 11955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5020597645385918e-06,
+      "loss": 0.0004,
+      "step": 11956
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5020179419083665e-06,
+      "loss": 0.0006,
+      "step": 11957
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5019761192781413e-06,
+      "loss": 0.1043,
+      "step": 11958
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5019342966479161e-06,
+      "loss": 0.3824,
+      "step": 11959
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501892474017691e-06,
+      "loss": 0.0712,
+      "step": 11960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5018506513874655e-06,
+      "loss": 0.1835,
+      "step": 11961
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5018088287572405e-06,
+      "loss": 0.0012,
+      "step": 11962
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5017670061270153e-06,
+      "loss": 0.029,
+      "step": 11963
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.50172518349679e-06,
+      "loss": 0.0081,
+      "step": 11964
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5016833608665648e-06,
+      "loss": 0.1519,
+      "step": 11965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5016415382363396e-06,
+      "loss": 0.0004,
+      "step": 11966
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5015997156061144e-06,
+      "loss": 0.0006,
+      "step": 11967
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5015578929758892e-06,
+      "loss": 0.0196,
+      "step": 11968
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501516070345664e-06,
+      "loss": 0.0004,
+      "step": 11969
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5014742477154388e-06,
+      "loss": 0.3987,
+      "step": 11970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5014324250852136e-06,
+      "loss": 0.0008,
+      "step": 11971
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5013906024549883e-06,
+      "loss": 0.139,
+      "step": 11972
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5013487798247631e-06,
+      "loss": 0.0007,
+      "step": 11973
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501306957194538e-06,
+      "loss": 0.0076,
+      "step": 11974
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5012651345643125e-06,
+      "loss": 0.0097,
+      "step": 11975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5012233119340875e-06,
+      "loss": 0.0043,
+      "step": 11976
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5011814893038623e-06,
+      "loss": 0.1345,
+      "step": 11977
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501139666673637e-06,
+      "loss": 0.0008,
+      "step": 11978
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5010978440434118e-06,
+      "loss": 0.0006,
+      "step": 11979
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5010560214131866e-06,
+      "loss": 0.0113,
+      "step": 11980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5010141987829614e-06,
+      "loss": 0.0007,
+      "step": 11981
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5009723761527362e-06,
+      "loss": 0.001,
+      "step": 11982
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500930553522511e-06,
+      "loss": 0.0051,
+      "step": 11983
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5008887308922858e-06,
+      "loss": 1.9044,
+      "step": 11984
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5008469082620606e-06,
+      "loss": 0.0178,
+      "step": 11985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5008050856318353e-06,
+      "loss": 0.0011,
+      "step": 11986
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5007632630016101e-06,
+      "loss": 0.0008,
+      "step": 11987
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500721440371385e-06,
+      "loss": 0.0007,
+      "step": 11988
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5006796177411595e-06,
+      "loss": 2.4975,
+      "step": 11989
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5006377951109343e-06,
+      "loss": 0.0009,
+      "step": 11990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5005959724807093e-06,
+      "loss": 0.0019,
+      "step": 11991
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500554149850484e-06,
+      "loss": 0.0031,
+      "step": 11992
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5005123272202589e-06,
+      "loss": 0.012,
+      "step": 11993
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5004705045900336e-06,
+      "loss": 0.315,
+      "step": 11994
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5004286819598084e-06,
+      "loss": 0.0081,
+      "step": 11995
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5003868593295832e-06,
+      "loss": 0.0009,
+      "step": 11996
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500345036699358e-06,
+      "loss": 0.0031,
+      "step": 11997
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5003032140691328e-06,
+      "loss": 0.0005,
+      "step": 11998
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5002613914389076e-06,
+      "loss": 0.2467,
+      "step": 11999
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5002195688086824e-06,
+      "loss": 0.0053,
+      "step": 12000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5001777461784571e-06,
+      "loss": 0.0015,
+      "step": 12001
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5001359235482317e-06,
+      "loss": 0.0003,
+      "step": 12002
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5000941009180065e-06,
+      "loss": 0.0052,
+      "step": 12003
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5000522782877813e-06,
+      "loss": 0.0006,
+      "step": 12004
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500010455657556e-06,
+      "loss": 0.0003,
+      "step": 12005
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499968633027331e-06,
+      "loss": 0.0006,
+      "step": 12006
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4999268103971059e-06,
+      "loss": 0.0003,
+      "step": 12007
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4998849877668806e-06,
+      "loss": 0.0016,
+      "step": 12008
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4998431651366554e-06,
+      "loss": 0.0005,
+      "step": 12009
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4998013425064302e-06,
+      "loss": 0.0043,
+      "step": 12010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499759519876205e-06,
+      "loss": 0.3189,
+      "step": 12011
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4997176972459798e-06,
+      "loss": 0.0004,
+      "step": 12012
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4996758746157546e-06,
+      "loss": 0.0007,
+      "step": 12013
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4996340519855294e-06,
+      "loss": 0.0008,
+      "step": 12014
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4995922293553042e-06,
+      "loss": 0.0368,
+      "step": 12015
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4995504067250787e-06,
+      "loss": 0.9143,
+      "step": 12016
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4995085840948535e-06,
+      "loss": 0.0031,
+      "step": 12017
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4994667614646283e-06,
+      "loss": 1.4071,
+      "step": 12018
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499424938834403e-06,
+      "loss": 0.005,
+      "step": 12019
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499383116204178e-06,
+      "loss": 3.5299,
+      "step": 12020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4993412935739529e-06,
+      "loss": 4.2703,
+      "step": 12021
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4992994709437277e-06,
+      "loss": 0.0241,
+      "step": 12022
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4992576483135024e-06,
+      "loss": 0.0444,
+      "step": 12023
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4992158256832772e-06,
+      "loss": 0.1426,
+      "step": 12024
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499174003053052e-06,
+      "loss": 0.0022,
+      "step": 12025
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4991321804228268e-06,
+      "loss": 0.0028,
+      "step": 12026
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4990903577926016e-06,
+      "loss": 0.0022,
+      "step": 12027
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4990485351623764e-06,
+      "loss": 1.7315,
+      "step": 12028
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499006712532151e-06,
+      "loss": 0.4248,
+      "step": 12029
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4989648899019257e-06,
+      "loss": 0.6572,
+      "step": 12030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4989230672717005e-06,
+      "loss": 0.0006,
+      "step": 12031
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4988812446414753e-06,
+      "loss": 0.0006,
+      "step": 12032
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.49883942201125e-06,
+      "loss": 0.0009,
+      "step": 12033
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4987975993810249e-06,
+      "loss": 1.2245,
+      "step": 12034
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4987557767507999e-06,
+      "loss": 0.002,
+      "step": 12035
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4987139541205747e-06,
+      "loss": 1.9285,
+      "step": 12036
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4986721314903495e-06,
+      "loss": 2.3575,
+      "step": 12037
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4986303088601242e-06,
+      "loss": 0.0144,
+      "step": 12038
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498588486229899e-06,
+      "loss": 0.001,
+      "step": 12039
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4985466635996738e-06,
+      "loss": 0.0046,
+      "step": 12040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4985048409694486e-06,
+      "loss": 0.1497,
+      "step": 12041
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4984630183392234e-06,
+      "loss": 0.0005,
+      "step": 12042
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498421195708998e-06,
+      "loss": 0.0036,
+      "step": 12043
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4983793730787727e-06,
+      "loss": 0.0012,
+      "step": 12044
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4983375504485475e-06,
+      "loss": 0.0557,
+      "step": 12045
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4982957278183223e-06,
+      "loss": 0.0007,
+      "step": 12046
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498253905188097e-06,
+      "loss": 0.0003,
+      "step": 12047
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4982120825578719e-06,
+      "loss": 2.1413,
+      "step": 12048
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4981702599276467e-06,
+      "loss": 0.003,
+      "step": 12049
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4981284372974217e-06,
+      "loss": 0.0022,
+      "step": 12050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4980866146671965e-06,
+      "loss": 0.0008,
+      "step": 12051
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4980447920369712e-06,
+      "loss": 0.0012,
+      "step": 12052
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.498002969406746e-06,
+      "loss": 0.0004,
+      "step": 12053
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4979611467765208e-06,
+      "loss": 1.9706,
+      "step": 12054
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4979193241462956e-06,
+      "loss": 0.0003,
+      "step": 12055
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4978775015160704e-06,
+      "loss": 1.1295,
+      "step": 12056
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.497835678885845e-06,
+      "loss": 0.829,
+      "step": 12057
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4977938562556198e-06,
+      "loss": 0.4761,
+      "step": 12058
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4977520336253945e-06,
+      "loss": 0.0012,
+      "step": 12059
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4977102109951693e-06,
+      "loss": 0.0003,
+      "step": 12060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4976683883649441e-06,
+      "loss": 0.9032,
+      "step": 12061
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.497626565734719e-06,
+      "loss": 0.0013,
+      "step": 12062
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4975847431044937e-06,
+      "loss": 0.0003,
+      "step": 12063
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4975429204742687e-06,
+      "loss": 0.0046,
+      "step": 12064
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4975010978440435e-06,
+      "loss": 0.0008,
+      "step": 12065
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4974592752138183e-06,
+      "loss": 1.6515,
+      "step": 12066
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.497417452583593e-06,
+      "loss": 0.0235,
+      "step": 12067
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4973756299533678e-06,
+      "loss": 0.001,
+      "step": 12068
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4973338073231426e-06,
+      "loss": 0.003,
+      "step": 12069
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4972919846929172e-06,
+      "loss": 0.0008,
+      "step": 12070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.497250162062692e-06,
+      "loss": 0.2393,
+      "step": 12071
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4972083394324668e-06,
+      "loss": 0.0014,
+      "step": 12072
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4971665168022415e-06,
+      "loss": 0.0029,
+      "step": 12073
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4971246941720163e-06,
+      "loss": 0.4528,
+      "step": 12074
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4970828715417911e-06,
+      "loss": 0.0005,
+      "step": 12075
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.497041048911566e-06,
+      "loss": 0.4197,
+      "step": 12076
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4969992262813407e-06,
+      "loss": 0.0059,
+      "step": 12077
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4969574036511155e-06,
+      "loss": 0.0028,
+      "step": 12078
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4969155810208905e-06,
+      "loss": 0.0064,
+      "step": 12079
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4968737583906653e-06,
+      "loss": 0.0453,
+      "step": 12080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.49683193576044e-06,
+      "loss": 0.0014,
+      "step": 12081
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4967901131302148e-06,
+      "loss": 0.0022,
+      "step": 12082
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4967482904999896e-06,
+      "loss": 0.2108,
+      "step": 12083
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4967064678697642e-06,
+      "loss": 0.4764,
+      "step": 12084
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.496664645239539e-06,
+      "loss": 0.3875,
+      "step": 12085
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4966228226093138e-06,
+      "loss": 0.001,
+      "step": 12086
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4965809999790886e-06,
+      "loss": 0.0086,
+      "step": 12087
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4965391773488633e-06,
+      "loss": 0.0011,
+      "step": 12088
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4964973547186381e-06,
+      "loss": 1.8511,
+      "step": 12089
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.496455532088413e-06,
+      "loss": 0.3391,
+      "step": 12090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4964137094581877e-06,
+      "loss": 4.2081,
+      "step": 12091
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4963718868279625e-06,
+      "loss": 0.0005,
+      "step": 12092
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4963300641977373e-06,
+      "loss": 2.555,
+      "step": 12093
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4962882415675123e-06,
+      "loss": 0.0562,
+      "step": 12094
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.496246418937287e-06,
+      "loss": 0.002,
+      "step": 12095
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4962045963070618e-06,
+      "loss": 0.0052,
+      "step": 12096
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4961627736768364e-06,
+      "loss": 0.0017,
+      "step": 12097
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4961209510466112e-06,
+      "loss": 0.0033,
+      "step": 12098
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.496079128416386e-06,
+      "loss": 0.0007,
+      "step": 12099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4960373057861608e-06,
+      "loss": 0.0053,
+      "step": 12100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4959954831559356e-06,
+      "loss": 1.806,
+      "step": 12101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4959536605257104e-06,
+      "loss": 0.0018,
+      "step": 12102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4959118378954851e-06,
+      "loss": 0.4158,
+      "step": 12103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.49587001526526e-06,
+      "loss": 0.0016,
+      "step": 12104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4958281926350347e-06,
+      "loss": 0.0017,
+      "step": 12105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4957863700048095e-06,
+      "loss": 0.0008,
+      "step": 12106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4957445473745843e-06,
+      "loss": 0.3164,
+      "step": 12107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4957027247443593e-06,
+      "loss": 0.8334,
+      "step": 12108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495660902114134e-06,
+      "loss": 0.0008,
+      "step": 12109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4956190794839089e-06,
+      "loss": 0.0036,
+      "step": 12110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4955772568536834e-06,
+      "loss": 0.0008,
+      "step": 12111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4955354342234582e-06,
+      "loss": 0.0046,
+      "step": 12112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495493611593233e-06,
+      "loss": 0.0003,
+      "step": 12113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4954517889630078e-06,
+      "loss": 0.0006,
+      "step": 12114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4954099663327826e-06,
+      "loss": 0.0082,
+      "step": 12115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4953681437025574e-06,
+      "loss": 0.004,
+      "step": 12116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4953263210723321e-06,
+      "loss": 1.51,
+      "step": 12117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495284498442107e-06,
+      "loss": 0.0009,
+      "step": 12118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4952426758118817e-06,
+      "loss": 0.0017,
+      "step": 12119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4952008531816565e-06,
+      "loss": 0.0005,
+      "step": 12120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4951590305514313e-06,
+      "loss": 0.003,
+      "step": 12121
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495117207921206e-06,
+      "loss": 0.0021,
+      "step": 12122
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495075385290981e-06,
+      "loss": 0.0025,
+      "step": 12123
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4950335626607556e-06,
+      "loss": 0.0009,
+      "step": 12124
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4949917400305304e-06,
+      "loss": 0.0227,
+      "step": 12125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4949499174003052e-06,
+      "loss": 0.0016,
+      "step": 12126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.49490809477008e-06,
+      "loss": 0.0006,
+      "step": 12127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4948662721398548e-06,
+      "loss": 0.0005,
+      "step": 12128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4948244495096296e-06,
+      "loss": 0.0049,
+      "step": 12129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4947826268794044e-06,
+      "loss": 0.0011,
+      "step": 12130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4947408042491792e-06,
+      "loss": 1.1906,
+      "step": 12131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.494698981618954e-06,
+      "loss": 0.0015,
+      "step": 12132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4946571589887287e-06,
+      "loss": 0.0067,
+      "step": 12133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4946153363585035e-06,
+      "loss": 0.0042,
+      "step": 12134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4945735137282783e-06,
+      "loss": 0.6094,
+      "step": 12135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.494531691098053e-06,
+      "loss": 1.0957,
+      "step": 12136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4944898684678277e-06,
+      "loss": 0.0037,
+      "step": 12137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4944480458376027e-06,
+      "loss": 2.3436,
+      "step": 12138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4944062232073774e-06,
+      "loss": 0.0038,
+      "step": 12139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4943644005771522e-06,
+      "loss": 0.0007,
+      "step": 12140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.494322577946927e-06,
+      "loss": 0.0005,
+      "step": 12141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4942807553167018e-06,
+      "loss": 0.005,
+      "step": 12142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4942389326864766e-06,
+      "loss": 0.001,
+      "step": 12143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4941971100562514e-06,
+      "loss": 0.0061,
+      "step": 12144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4941552874260262e-06,
+      "loss": 0.0015,
+      "step": 12145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.494113464795801e-06,
+      "loss": 0.018,
+      "step": 12146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4940716421655757e-06,
+      "loss": 0.0005,
+      "step": 12147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4940298195353505e-06,
+      "loss": 0.001,
+      "step": 12148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4939879969051253e-06,
+      "loss": 0.0006,
+      "step": 12149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4939461742749e-06,
+      "loss": 0.0009,
+      "step": 12150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4939043516446747e-06,
+      "loss": 0.0023,
+      "step": 12151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4938625290144497e-06,
+      "loss": 0.0007,
+      "step": 12152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4938207063842245e-06,
+      "loss": 0.0064,
+      "step": 12153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4937788837539992e-06,
+      "loss": 0.0033,
+      "step": 12154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.493737061123774e-06,
+      "loss": 0.0006,
+      "step": 12155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4936952384935488e-06,
+      "loss": 0.003,
+      "step": 12156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4936534158633236e-06,
+      "loss": 0.0013,
+      "step": 12157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4936115932330984e-06,
+      "loss": 0.0133,
+      "step": 12158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4935697706028732e-06,
+      "loss": 0.0094,
+      "step": 12159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.493527947972648e-06,
+      "loss": 0.0003,
+      "step": 12160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4934861253424227e-06,
+      "loss": 0.0003,
+      "step": 12161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4934443027121975e-06,
+      "loss": 0.0036,
+      "step": 12162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4934024800819723e-06,
+      "loss": 0.0007,
+      "step": 12163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4933606574517469e-06,
+      "loss": 0.001,
+      "step": 12164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4933188348215217e-06,
+      "loss": 0.001,
+      "step": 12165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4932770121912965e-06,
+      "loss": 0.2469,
+      "step": 12166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4932351895610715e-06,
+      "loss": 0.0033,
+      "step": 12167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4931933669308462e-06,
+      "loss": 0.0032,
+      "step": 12168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.493151544300621e-06,
+      "loss": 0.4387,
+      "step": 12169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4931097216703958e-06,
+      "loss": 1.0757,
+      "step": 12170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4930678990401706e-06,
+      "loss": 0.0005,
+      "step": 12171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4930260764099454e-06,
+      "loss": 0.0005,
+      "step": 12172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4929842537797202e-06,
+      "loss": 0.001,
+      "step": 12173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492942431149495e-06,
+      "loss": 0.0004,
+      "step": 12174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4929006085192698e-06,
+      "loss": 0.0015,
+      "step": 12175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4928587858890445e-06,
+      "loss": 0.2819,
+      "step": 12176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4928169632588193e-06,
+      "loss": 0.0007,
+      "step": 12177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492775140628594e-06,
+      "loss": 0.0008,
+      "step": 12178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4927333179983687e-06,
+      "loss": 0.0031,
+      "step": 12179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4926914953681435e-06,
+      "loss": 0.3493,
+      "step": 12180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4926496727379185e-06,
+      "loss": 0.0206,
+      "step": 12181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4926078501076933e-06,
+      "loss": 1.4576,
+      "step": 12182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492566027477468e-06,
+      "loss": 0.0115,
+      "step": 12183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4925242048472428e-06,
+      "loss": 1.2185,
+      "step": 12184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4924823822170176e-06,
+      "loss": 0.1624,
+      "step": 12185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4924405595867924e-06,
+      "loss": 0.0003,
+      "step": 12186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4923987369565672e-06,
+      "loss": 0.0007,
+      "step": 12187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492356914326342e-06,
+      "loss": 0.0012,
+      "step": 12188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4923150916961168e-06,
+      "loss": 0.0264,
+      "step": 12189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4922732690658915e-06,
+      "loss": 0.2443,
+      "step": 12190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4922314464356663e-06,
+      "loss": 0.0108,
+      "step": 12191
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492189623805441e-06,
+      "loss": 0.0007,
+      "step": 12192
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4921478011752157e-06,
+      "loss": 3.1815,
+      "step": 12193
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4921059785449905e-06,
+      "loss": 0.0033,
+      "step": 12194
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4920641559147653e-06,
+      "loss": 2.5259,
+      "step": 12195
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4920223332845403e-06,
+      "loss": 0.0028,
+      "step": 12196
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491980510654315e-06,
+      "loss": 0.0013,
+      "step": 12197
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4919386880240898e-06,
+      "loss": 0.8194,
+      "step": 12198
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4918968653938646e-06,
+      "loss": 0.0005,
+      "step": 12199
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4918550427636394e-06,
+      "loss": 0.0004,
+      "step": 12200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4918132201334142e-06,
+      "loss": 0.0041,
+      "step": 12201
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491771397503189e-06,
+      "loss": 0.0093,
+      "step": 12202
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4917295748729638e-06,
+      "loss": 0.0018,
+      "step": 12203
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4916877522427386e-06,
+      "loss": 0.001,
+      "step": 12204
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4916459296125131e-06,
+      "loss": 0.0005,
+      "step": 12205
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491604106982288e-06,
+      "loss": 2.5685,
+      "step": 12206
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4915622843520627e-06,
+      "loss": 0.3142,
+      "step": 12207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4915204617218375e-06,
+      "loss": 0.0017,
+      "step": 12208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4914786390916123e-06,
+      "loss": 0.0004,
+      "step": 12209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.491436816461387e-06,
+      "loss": 0.013,
+      "step": 12210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.491394993831162e-06,
+      "loss": 0.0006,
+      "step": 12211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4913531712009368e-06,
+      "loss": 0.0004,
+      "step": 12212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4913113485707116e-06,
+      "loss": 0.0005,
+      "step": 12213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4912695259404864e-06,
+      "loss": 0.001,
+      "step": 12214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4912277033102612e-06,
+      "loss": 0.0022,
+      "step": 12215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.491185880680036e-06,
+      "loss": 0.3122,
+      "step": 12216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4911440580498108e-06,
+      "loss": 0.2226,
+      "step": 12217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4911022354195856e-06,
+      "loss": 0.0003,
+      "step": 12218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4910604127893601e-06,
+      "loss": 1.7382,
+      "step": 12219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.491018590159135e-06,
+      "loss": 0.0018,
+      "step": 12220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4909767675289097e-06,
+      "loss": 0.0006,
+      "step": 12221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4909349448986845e-06,
+      "loss": 0.0068,
+      "step": 12222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4908931222684593e-06,
+      "loss": 0.0004,
+      "step": 12223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.490851299638234e-06,
+      "loss": 0.0012,
+      "step": 12224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.490809477008009e-06,
+      "loss": 0.3339,
+      "step": 12225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4907676543777839e-06,
+      "loss": 1.7417,
+      "step": 12226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4907258317475586e-06,
+      "loss": 1.7754,
+      "step": 12227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4906840091173334e-06,
+      "loss": 0.0003,
+      "step": 12228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4906421864871082e-06,
+      "loss": 0.0043,
+      "step": 12229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.490600363856883e-06,
+      "loss": 0.0654,
+      "step": 12230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4905585412266578e-06,
+      "loss": 0.0298,
+      "step": 12231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4905167185964324e-06,
+      "loss": 0.0015,
+      "step": 12232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4904748959662071e-06,
+      "loss": 0.01,
+      "step": 12233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.490433073335982e-06,
+      "loss": 0.0012,
+      "step": 12234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4903912507057567e-06,
+      "loss": 0.8931,
+      "step": 12235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4903494280755315e-06,
+      "loss": 0.0009,
+      "step": 12236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4903076054453063e-06,
+      "loss": 0.0087,
+      "step": 12237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.490265782815081e-06,
+      "loss": 0.0006,
+      "step": 12238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4902239601848559e-06,
+      "loss": 0.0005,
+      "step": 12239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4901821375546309e-06,
+      "loss": 0.0004,
+      "step": 12240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4901403149244057e-06,
+      "loss": 0.0048,
+      "step": 12241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4900984922941804e-06,
+      "loss": 0.0023,
+      "step": 12242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4900566696639552e-06,
+      "loss": 0.0067,
+      "step": 12243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.49001484703373e-06,
+      "loss": 0.001,
+      "step": 12244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4899730244035048e-06,
+      "loss": 0.003,
+      "step": 12245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4899312017732794e-06,
+      "loss": 0.0005,
+      "step": 12246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4898893791430542e-06,
+      "loss": 0.0011,
+      "step": 12247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489847556512829e-06,
+      "loss": 0.0003,
+      "step": 12248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4898057338826037e-06,
+      "loss": 0.0005,
+      "step": 12249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4897639112523785e-06,
+      "loss": 0.002,
+      "step": 12250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4897220886221533e-06,
+      "loss": 0.0015,
+      "step": 12251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489680265991928e-06,
+      "loss": 0.0011,
+      "step": 12252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4896384433617029e-06,
+      "loss": 0.0019,
+      "step": 12253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4895966207314777e-06,
+      "loss": 0.0005,
+      "step": 12254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4895547981012527e-06,
+      "loss": 0.7091,
+      "step": 12255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4895129754710274e-06,
+      "loss": 0.1027,
+      "step": 12256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4894711528408022e-06,
+      "loss": 0.0445,
+      "step": 12257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489429330210577e-06,
+      "loss": 0.0025,
+      "step": 12258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4893875075803518e-06,
+      "loss": 0.1203,
+      "step": 12259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4893456849501264e-06,
+      "loss": 0.001,
+      "step": 12260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4893038623199012e-06,
+      "loss": 0.0055,
+      "step": 12261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489262039689676e-06,
+      "loss": 0.0212,
+      "step": 12262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4892202170594507e-06,
+      "loss": 0.331,
+      "step": 12263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4891783944292255e-06,
+      "loss": 0.0007,
+      "step": 12264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4891365717990003e-06,
+      "loss": 0.0002,
+      "step": 12265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489094749168775e-06,
+      "loss": 0.008,
+      "step": 12266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4890529265385499e-06,
+      "loss": 1.4943,
+      "step": 12267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4890111039083247e-06,
+      "loss": 0.0068,
+      "step": 12268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4889692812780997e-06,
+      "loss": 2.2501,
+      "step": 12269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4889274586478745e-06,
+      "loss": 1.121,
+      "step": 12270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4888856360176492e-06,
+      "loss": 0.0017,
+      "step": 12271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488843813387424e-06,
+      "loss": 0.0016,
+      "step": 12272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4888019907571986e-06,
+      "loss": 0.0032,
+      "step": 12273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4887601681269734e-06,
+      "loss": 0.0013,
+      "step": 12274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4887183454967482e-06,
+      "loss": 0.0051,
+      "step": 12275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488676522866523e-06,
+      "loss": 0.0024,
+      "step": 12276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4886347002362977e-06,
+      "loss": 0.0004,
+      "step": 12277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4885928776060725e-06,
+      "loss": 0.0023,
+      "step": 12278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4885510549758473e-06,
+      "loss": 0.0004,
+      "step": 12279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488509232345622e-06,
+      "loss": 0.0026,
+      "step": 12280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4884674097153969e-06,
+      "loss": 0.4645,
+      "step": 12281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4884255870851717e-06,
+      "loss": 0.0032,
+      "step": 12282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4883837644549465e-06,
+      "loss": 0.0146,
+      "step": 12283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4883419418247215e-06,
+      "loss": 0.0014,
+      "step": 12284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4883001191944962e-06,
+      "loss": 0.0005,
+      "step": 12285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488258296564271e-06,
+      "loss": 0.0084,
+      "step": 12286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4882164739340456e-06,
+      "loss": 3.0434,
+      "step": 12287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4881746513038204e-06,
+      "loss": 0.0013,
+      "step": 12288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4881328286735952e-06,
+      "loss": 0.0016,
+      "step": 12289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.48809100604337e-06,
+      "loss": 0.2377,
+      "step": 12290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4880491834131448e-06,
+      "loss": 0.8728,
+      "step": 12291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4880073607829195e-06,
+      "loss": 0.6713,
+      "step": 12292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4879655381526943e-06,
+      "loss": 0.01,
+      "step": 12293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4879237155224691e-06,
+      "loss": 0.0049,
+      "step": 12294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.487881892892244e-06,
+      "loss": 0.0005,
+      "step": 12295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4878400702620187e-06,
+      "loss": 0.0066,
+      "step": 12296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4877982476317935e-06,
+      "loss": 0.0005,
+      "step": 12297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4877564250015683e-06,
+      "loss": 0.0008,
+      "step": 12298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4877146023713433e-06,
+      "loss": 0.0011,
+      "step": 12299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4876727797411178e-06,
+      "loss": 0.0026,
+      "step": 12300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4876309571108926e-06,
+      "loss": 0.0018,
+      "step": 12301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4875891344806674e-06,
+      "loss": 0.0003,
+      "step": 12302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4875473118504422e-06,
+      "loss": 0.0243,
+      "step": 12303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.487505489220217e-06,
+      "loss": 0.0014,
+      "step": 12304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4874636665899918e-06,
+      "loss": 0.0011,
+      "step": 12305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4874218439597666e-06,
+      "loss": 0.0003,
+      "step": 12306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4873800213295413e-06,
+      "loss": 0.4392,
+      "step": 12307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4873381986993161e-06,
+      "loss": 0.1036,
+      "step": 12308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.487296376069091e-06,
+      "loss": 0.0023,
+      "step": 12309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4872545534388657e-06,
+      "loss": 0.0014,
+      "step": 12310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4872127308086405e-06,
+      "loss": 0.0033,
+      "step": 12311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4871709081784153e-06,
+      "loss": 0.8842,
+      "step": 12312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4871290855481903e-06,
+      "loss": 0.0008,
+      "step": 12313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4870872629179648e-06,
+      "loss": 0.0162,
+      "step": 12314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4870454402877396e-06,
+      "loss": 0.0009,
+      "step": 12315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4870036176575144e-06,
+      "loss": 0.002,
+      "step": 12316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4869617950272892e-06,
+      "loss": 0.3424,
+      "step": 12317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.486919972397064e-06,
+      "loss": 2.6974,
+      "step": 12318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4868781497668388e-06,
+      "loss": 0.0013,
+      "step": 12319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4868363271366136e-06,
+      "loss": 0.0068,
+      "step": 12320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4867945045063883e-06,
+      "loss": 0.0005,
+      "step": 12321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4867526818761631e-06,
+      "loss": 0.3063,
+      "step": 12322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.486710859245938e-06,
+      "loss": 0.0003,
+      "step": 12323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4866690366157127e-06,
+      "loss": 0.0009,
+      "step": 12324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4866272139854875e-06,
+      "loss": 0.0069,
+      "step": 12325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4865853913552623e-06,
+      "loss": 0.0137,
+      "step": 12326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4865435687250369e-06,
+      "loss": 0.0008,
+      "step": 12327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4865017460948118e-06,
+      "loss": 0.4302,
+      "step": 12328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4864599234645866e-06,
+      "loss": 0.0008,
+      "step": 12329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4864181008343614e-06,
+      "loss": 0.1776,
+      "step": 12330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4863762782041362e-06,
+      "loss": 0.0048,
+      "step": 12331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.486334455573911e-06,
+      "loss": 0.0035,
+      "step": 12332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4862926329436858e-06,
+      "loss": 0.656,
+      "step": 12333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4862508103134606e-06,
+      "loss": 0.1305,
+      "step": 12334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4862089876832354e-06,
+      "loss": 0.0003,
+      "step": 12335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4861671650530101e-06,
+      "loss": 4.8415,
+      "step": 12336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.486125342422785e-06,
+      "loss": 0.0025,
+      "step": 12337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4860835197925597e-06,
+      "loss": 2.199,
+      "step": 12338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4860416971623345e-06,
+      "loss": 2.298,
+      "step": 12339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485999874532109e-06,
+      "loss": 0.0009,
+      "step": 12340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4859580519018839e-06,
+      "loss": 0.0004,
+      "step": 12341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4859162292716586e-06,
+      "loss": 0.0011,
+      "step": 12342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4858744066414336e-06,
+      "loss": 0.0006,
+      "step": 12343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4858325840112084e-06,
+      "loss": 0.0013,
+      "step": 12344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4857907613809832e-06,
+      "loss": 0.5664,
+      "step": 12345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485748938750758e-06,
+      "loss": 0.0014,
+      "step": 12346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4857071161205328e-06,
+      "loss": 0.0023,
+      "step": 12347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4856652934903076e-06,
+      "loss": 2.8334,
+      "step": 12348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4856234708600824e-06,
+      "loss": 3.1262,
+      "step": 12349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4855816482298571e-06,
+      "loss": 0.0011,
+      "step": 12350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485539825599632e-06,
+      "loss": 0.0035,
+      "step": 12351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4854980029694067e-06,
+      "loss": 2.3869,
+      "step": 12352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4854561803391815e-06,
+      "loss": 0.0024,
+      "step": 12353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485414357708956e-06,
+      "loss": 0.0013,
+      "step": 12354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4853725350787309e-06,
+      "loss": 0.0016,
+      "step": 12355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4853307124485057e-06,
+      "loss": 0.0011,
+      "step": 12356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4852888898182807e-06,
+      "loss": 0.0005,
+      "step": 12357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4852470671880554e-06,
+      "loss": 0.4731,
+      "step": 12358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4852052445578302e-06,
+      "loss": 0.0005,
+      "step": 12359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485163421927605e-06,
+      "loss": 3.1007,
+      "step": 12360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4851215992973798e-06,
+      "loss": 0.0004,
+      "step": 12361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4850797766671546e-06,
+      "loss": 0.0007,
+      "step": 12362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4850379540369294e-06,
+      "loss": 1.7448,
+      "step": 12363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4849961314067042e-06,
+      "loss": 0.0072,
+      "step": 12364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.484954308776479e-06,
+      "loss": 0.2873,
+      "step": 12365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4849124861462537e-06,
+      "loss": 0.0005,
+      "step": 12366
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4848706635160283e-06,
+      "loss": 3.4837,
+      "step": 12367
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.484828840885803e-06,
+      "loss": 0.0008,
+      "step": 12368
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4847870182555779e-06,
+      "loss": 3.1488,
+      "step": 12369
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4847451956253527e-06,
+      "loss": 1.2313,
+      "step": 12370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4847033729951274e-06,
+      "loss": 0.0005,
+      "step": 12371
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4846615503649024e-06,
+      "loss": 0.0048,
+      "step": 12372
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4846197277346772e-06,
+      "loss": 0.0003,
+      "step": 12373
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.484577905104452e-06,
+      "loss": 0.0015,
+      "step": 12374
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4845360824742268e-06,
+      "loss": 0.102,
+      "step": 12375
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4844942598440016e-06,
+      "loss": 0.2352,
+      "step": 12376
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4844524372137764e-06,
+      "loss": 0.0005,
+      "step": 12377
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4844106145835512e-06,
+      "loss": 0.0007,
+      "step": 12378
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.484368791953326e-06,
+      "loss": 0.0004,
+      "step": 12379
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4843269693231007e-06,
+      "loss": 0.019,
+      "step": 12380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4842851466928753e-06,
+      "loss": 0.0303,
+      "step": 12381
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.48424332406265e-06,
+      "loss": 0.0011,
+      "step": 12382
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4842015014324249e-06,
+      "loss": 1.4401,
+      "step": 12383
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4841596788021997e-06,
+      "loss": 0.0053,
+      "step": 12384
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4841178561719745e-06,
+      "loss": 0.0009,
+      "step": 12385
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4840760335417495e-06,
+      "loss": 0.0015,
+      "step": 12386
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4840342109115242e-06,
+      "loss": 0.0022,
+      "step": 12387
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483992388281299e-06,
+      "loss": 0.0006,
+      "step": 12388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4839505656510738e-06,
+      "loss": 0.0004,
+      "step": 12389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4839087430208486e-06,
+      "loss": 0.2068,
+      "step": 12390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4838669203906234e-06,
+      "loss": 0.0007,
+      "step": 12391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4838250977603982e-06,
+      "loss": 0.003,
+      "step": 12392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483783275130173e-06,
+      "loss": 0.0028,
+      "step": 12393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4837414524999477e-06,
+      "loss": 0.0013,
+      "step": 12394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4836996298697223e-06,
+      "loss": 0.1945,
+      "step": 12395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4836578072394971e-06,
+      "loss": 0.7036,
+      "step": 12396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483615984609272e-06,
+      "loss": 0.0024,
+      "step": 12397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4835741619790467e-06,
+      "loss": 2.3514,
+      "step": 12398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4835323393488215e-06,
+      "loss": 0.0336,
+      "step": 12399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4834905167185963e-06,
+      "loss": 0.0007,
+      "step": 12400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4834486940883713e-06,
+      "loss": 0.684,
+      "step": 12401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483406871458146e-06,
+      "loss": 0.008,
+      "step": 12402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4833650488279208e-06,
+      "loss": 0.0011,
+      "step": 12403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4833232261976956e-06,
+      "loss": 1.947,
+      "step": 12404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4832814035674704e-06,
+      "loss": 0.004,
+      "step": 12405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4832395809372452e-06,
+      "loss": 0.1983,
+      "step": 12406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.48319775830702e-06,
+      "loss": 0.4536,
+      "step": 12407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4831559356767945e-06,
+      "loss": 0.0016,
+      "step": 12408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4831141130465693e-06,
+      "loss": 0.0014,
+      "step": 12409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4830722904163441e-06,
+      "loss": 0.0007,
+      "step": 12410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483030467786119e-06,
+      "loss": 1.432,
+      "step": 12411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4829886451558937e-06,
+      "loss": 2.5799,
+      "step": 12412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4829468225256685e-06,
+      "loss": 0.0013,
+      "step": 12413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4829049998954433e-06,
+      "loss": 0.0134,
+      "step": 12414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482863177265218e-06,
+      "loss": 0.0006,
+      "step": 12415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482821354634993e-06,
+      "loss": 0.0166,
+      "step": 12416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4827795320047678e-06,
+      "loss": 0.0092,
+      "step": 12417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4827377093745426e-06,
+      "loss": 0.0027,
+      "step": 12418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4826958867443174e-06,
+      "loss": 0.0094,
+      "step": 12419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4826540641140922e-06,
+      "loss": 0.0009,
+      "step": 12420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482612241483867e-06,
+      "loss": 0.0094,
+      "step": 12421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4825704188536416e-06,
+      "loss": 0.0124,
+      "step": 12422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4825285962234163e-06,
+      "loss": 2.2449,
+      "step": 12423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4824867735931911e-06,
+      "loss": 0.0017,
+      "step": 12424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482444950962966e-06,
+      "loss": 0.0046,
+      "step": 12425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4824031283327407e-06,
+      "loss": 0.3483,
+      "step": 12426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4823613057025155e-06,
+      "loss": 0.0038,
+      "step": 12427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4823194830722903e-06,
+      "loss": 0.0076,
+      "step": 12428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482277660442065e-06,
+      "loss": 0.008,
+      "step": 12429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.48223583781184e-06,
+      "loss": 0.0004,
+      "step": 12430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4821940151816148e-06,
+      "loss": 0.0011,
+      "step": 12431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4821521925513896e-06,
+      "loss": 0.4341,
+      "step": 12432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4821103699211644e-06,
+      "loss": 0.0022,
+      "step": 12433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4820685472909392e-06,
+      "loss": 0.0005,
+      "step": 12434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4820267246607138e-06,
+      "loss": 0.0048,
+      "step": 12435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4819849020304886e-06,
+      "loss": 0.0037,
+      "step": 12436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4819430794002633e-06,
+      "loss": 0.0056,
+      "step": 12437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4819012567700381e-06,
+      "loss": 0.0005,
+      "step": 12438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.481859434139813e-06,
+      "loss": 0.0004,
+      "step": 12439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4818176115095877e-06,
+      "loss": 0.0007,
+      "step": 12440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4817757888793625e-06,
+      "loss": 1.6848,
+      "step": 12441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4817339662491373e-06,
+      "loss": 4.2191,
+      "step": 12442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.481692143618912e-06,
+      "loss": 0.0013,
+      "step": 12443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4816503209886869e-06,
+      "loss": 0.0214,
+      "step": 12444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4816084983584618e-06,
+      "loss": 0.8275,
+      "step": 12445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4815666757282366e-06,
+      "loss": 0.1385,
+      "step": 12446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4815248530980114e-06,
+      "loss": 0.0006,
+      "step": 12447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4814830304677862e-06,
+      "loss": 0.0015,
+      "step": 12448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4814412078375608e-06,
+      "loss": 1.4563,
+      "step": 12449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4813993852073356e-06,
+      "loss": 0.0016,
+      "step": 12450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4813575625771104e-06,
+      "loss": 0.0011,
+      "step": 12451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4813157399468851e-06,
+      "loss": 1.6757,
+      "step": 12452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.48127391731666e-06,
+      "loss": 0.0131,
+      "step": 12453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4812320946864347e-06,
+      "loss": 2.8924,
+      "step": 12454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4811902720562095e-06,
+      "loss": 0.0413,
+      "step": 12455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4811484494259843e-06,
+      "loss": 0.753,
+      "step": 12456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.481106626795759e-06,
+      "loss": 0.0009,
+      "step": 12457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4810648041655339e-06,
+      "loss": 4.6706,
+      "step": 12458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4810229815353086e-06,
+      "loss": 0.0075,
+      "step": 12459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4809811589050836e-06,
+      "loss": 0.0007,
+      "step": 12460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4809393362748584e-06,
+      "loss": 1.3694,
+      "step": 12461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4808975136446332e-06,
+      "loss": 0.0013,
+      "step": 12462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4808556910144078e-06,
+      "loss": 0.2707,
+      "step": 12463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4808138683841826e-06,
+      "loss": 2.8314,
+      "step": 12464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4807720457539574e-06,
+      "loss": 0.3767,
+      "step": 12465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4807302231237322e-06,
+      "loss": 0.0037,
+      "step": 12466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.480688400493507e-06,
+      "loss": 0.0017,
+      "step": 12467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4806465778632817e-06,
+      "loss": 0.0013,
+      "step": 12468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4806047552330565e-06,
+      "loss": 0.0019,
+      "step": 12469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4805629326028313e-06,
+      "loss": 0.0041,
+      "step": 12470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.480521109972606e-06,
+      "loss": 0.0041,
+      "step": 12471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4804792873423809e-06,
+      "loss": 0.0012,
+      "step": 12472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4804374647121557e-06,
+      "loss": 0.5793,
+      "step": 12473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4803956420819307e-06,
+      "loss": 1.5964,
+      "step": 12474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4803538194517054e-06,
+      "loss": 0.0007,
+      "step": 12475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.48031199682148e-06,
+      "loss": 0.0005,
+      "step": 12476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4802701741912548e-06,
+      "loss": 0.0093,
+      "step": 12477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4802283515610296e-06,
+      "loss": 0.948,
+      "step": 12478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4801865289308044e-06,
+      "loss": 1.6656,
+      "step": 12479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4801447063005792e-06,
+      "loss": 0.0004,
+      "step": 12480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.480102883670354e-06,
+      "loss": 0.0058,
+      "step": 12481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4800610610401287e-06,
+      "loss": 0.0009,
+      "step": 12482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4800192384099035e-06,
+      "loss": 0.0012,
+      "step": 12483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4799774157796783e-06,
+      "loss": 0.0047,
+      "step": 12484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479935593149453e-06,
+      "loss": 0.0015,
+      "step": 12485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4798937705192279e-06,
+      "loss": 0.0012,
+      "step": 12486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4798519478890027e-06,
+      "loss": 0.0027,
+      "step": 12487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4798101252587775e-06,
+      "loss": 0.004,
+      "step": 12488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4797683026285524e-06,
+      "loss": 0.0013,
+      "step": 12489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479726479998327e-06,
+      "loss": 0.0034,
+      "step": 12490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4796846573681018e-06,
+      "loss": 1.2321,
+      "step": 12491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4796428347378766e-06,
+      "loss": 0.0005,
+      "step": 12492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4796010121076514e-06,
+      "loss": 0.0019,
+      "step": 12493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4795591894774262e-06,
+      "loss": 0.4843,
+      "step": 12494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479517366847201e-06,
+      "loss": 0.0011,
+      "step": 12495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4794755442169757e-06,
+      "loss": 0.0006,
+      "step": 12496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4794337215867505e-06,
+      "loss": 0.0119,
+      "step": 12497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4793918989565253e-06,
+      "loss": 0.0018,
+      "step": 12498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4793500763263e-06,
+      "loss": 0.0043,
+      "step": 12499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4793082536960749e-06,
+      "loss": 0.0171,
+      "step": 12500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4792664310658497e-06,
+      "loss": 0.0009,
+      "step": 12501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4792246084356245e-06,
+      "loss": 0.0017,
+      "step": 12502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479182785805399e-06,
+      "loss": 0.0018,
+      "step": 12503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479140963175174e-06,
+      "loss": 0.0004,
+      "step": 12504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4790991405449488e-06,
+      "loss": 0.0018,
+      "step": 12505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4790573179147236e-06,
+      "loss": 0.0007,
+      "step": 12506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4790154952844984e-06,
+      "loss": 0.0005,
+      "step": 12507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4789736726542732e-06,
+      "loss": 0.3169,
+      "step": 12508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478931850024048e-06,
+      "loss": 0.0015,
+      "step": 12509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4788900273938227e-06,
+      "loss": 0.272,
+      "step": 12510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4788482047635975e-06,
+      "loss": 0.0012,
+      "step": 12511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4788063821333723e-06,
+      "loss": 0.0005,
+      "step": 12512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4787645595031471e-06,
+      "loss": 1.9046,
+      "step": 12513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478722736872922e-06,
+      "loss": 0.0021,
+      "step": 12514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4786809142426967e-06,
+      "loss": 0.0018,
+      "step": 12515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4786390916124713e-06,
+      "loss": 0.422,
+      "step": 12516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478597268982246e-06,
+      "loss": 0.0003,
+      "step": 12517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478555446352021e-06,
+      "loss": 0.0008,
+      "step": 12518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4785136237217958e-06,
+      "loss": 0.0006,
+      "step": 12519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4784718010915706e-06,
+      "loss": 0.0103,
+      "step": 12520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4784299784613454e-06,
+      "loss": 0.0181,
+      "step": 12521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4783881558311202e-06,
+      "loss": 0.0192,
+      "step": 12522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478346333200895e-06,
+      "loss": 0.001,
+      "step": 12523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4783045105706698e-06,
+      "loss": 0.0018,
+      "step": 12524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4782626879404445e-06,
+      "loss": 0.4644,
+      "step": 12525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4782208653102193e-06,
+      "loss": 0.1554,
+      "step": 12526
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4781790426799941e-06,
+      "loss": 0.0008,
+      "step": 12527
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.478137220049769e-06,
+      "loss": 2.5892,
+      "step": 12528
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4780953974195437e-06,
+      "loss": 0.0034,
+      "step": 12529
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4780535747893183e-06,
+      "loss": 0.0004,
+      "step": 12530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.478011752159093e-06,
+      "loss": 0.001,
+      "step": 12531
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4779699295288678e-06,
+      "loss": 0.0031,
+      "step": 12532
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4779281068986428e-06,
+      "loss": 0.0008,
+      "step": 12533
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4778862842684176e-06,
+      "loss": 0.6866,
+      "step": 12534
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4778444616381924e-06,
+      "loss": 0.9193,
+      "step": 12535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4778026390079672e-06,
+      "loss": 0.0008,
+      "step": 12536
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.477760816377742e-06,
+      "loss": 2.5082,
+      "step": 12537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4777189937475168e-06,
+      "loss": 0.0007,
+      "step": 12538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4776771711172916e-06,
+      "loss": 0.0086,
+      "step": 12539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4776353484870663e-06,
+      "loss": 0.0007,
+      "step": 12540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4775935258568411e-06,
+      "loss": 0.0012,
+      "step": 12541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.477551703226616e-06,
+      "loss": 0.0019,
+      "step": 12542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4775098805963905e-06,
+      "loss": 0.0017,
+      "step": 12543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4774680579661653e-06,
+      "loss": 0.2888,
+      "step": 12544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.47742623533594e-06,
+      "loss": 0.5658,
+      "step": 12545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4773844127057148e-06,
+      "loss": 0.0507,
+      "step": 12546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4773425900754896e-06,
+      "loss": 0.8072,
+      "step": 12547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4773007674452646e-06,
+      "loss": 0.0022,
+      "step": 12548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4772589448150394e-06,
+      "loss": 0.0126,
+      "step": 12549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4772171221848142e-06,
+      "loss": 3.6677,
+      "step": 12550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.477175299554589e-06,
+      "loss": 0.1868,
+      "step": 12551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4771334769243638e-06,
+      "loss": 0.0004,
+      "step": 12552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4770916542941386e-06,
+      "loss": 0.7741,
+      "step": 12553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4770498316639133e-06,
+      "loss": 0.0011,
+      "step": 12554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4770080090336881e-06,
+      "loss": 0.0086,
+      "step": 12555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.476966186403463e-06,
+      "loss": 2.0917,
+      "step": 12556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4769243637732375e-06,
+      "loss": 0.0005,
+      "step": 12557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4768825411430123e-06,
+      "loss": 0.0045,
+      "step": 12558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.476840718512787e-06,
+      "loss": 0.0023,
+      "step": 12559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4767988958825619e-06,
+      "loss": 0.0039,
+      "step": 12560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4767570732523366e-06,
+      "loss": 0.0004,
+      "step": 12561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4767152506221116e-06,
+      "loss": 0.0004,
+      "step": 12562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4766734279918864e-06,
+      "loss": 0.876,
+      "step": 12563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4766316053616612e-06,
+      "loss": 0.0005,
+      "step": 12564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.476589782731436e-06,
+      "loss": 0.282,
+      "step": 12565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4765479601012108e-06,
+      "loss": 0.0006,
+      "step": 12566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4765061374709856e-06,
+      "loss": 0.0008,
+      "step": 12567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4764643148407604e-06,
+      "loss": 0.6463,
+      "step": 12568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4764224922105351e-06,
+      "loss": 0.0587,
+      "step": 12569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4763806695803097e-06,
+      "loss": 0.0015,
+      "step": 12570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4763388469500845e-06,
+      "loss": 0.0018,
+      "step": 12571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4762970243198593e-06,
+      "loss": 0.0029,
+      "step": 12572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.476255201689634e-06,
+      "loss": 0.0013,
+      "step": 12573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4762133790594089e-06,
+      "loss": 0.0008,
+      "step": 12574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4761715564291836e-06,
+      "loss": 1.3722,
+      "step": 12575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4761297337989584e-06,
+      "loss": 1.4293,
+      "step": 12576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4760879111687334e-06,
+      "loss": 0.407,
+      "step": 12577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4760460885385082e-06,
+      "loss": 0.0007,
+      "step": 12578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.476004265908283e-06,
+      "loss": 0.0022,
+      "step": 12579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4759624432780578e-06,
+      "loss": 0.0004,
+      "step": 12580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4759206206478326e-06,
+      "loss": 0.0033,
+      "step": 12581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4758787980176074e-06,
+      "loss": 0.0004,
+      "step": 12582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4758369753873822e-06,
+      "loss": 0.0017,
+      "step": 12583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4757951527571567e-06,
+      "loss": 0.0096,
+      "step": 12584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4757533301269315e-06,
+      "loss": 0.0032,
+      "step": 12585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4757115074967063e-06,
+      "loss": 0.1198,
+      "step": 12586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.475669684866481e-06,
+      "loss": 0.2607,
+      "step": 12587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4756278622362559e-06,
+      "loss": 0.0005,
+      "step": 12588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4755860396060307e-06,
+      "loss": 0.6045,
+      "step": 12589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4755442169758054e-06,
+      "loss": 0.0042,
+      "step": 12590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4755023943455804e-06,
+      "loss": 0.0011,
+      "step": 12591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4754605717153552e-06,
+      "loss": 0.001,
+      "step": 12592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.47541874908513e-06,
+      "loss": 0.0005,
+      "step": 12593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4753769264549048e-06,
+      "loss": 0.2843,
+      "step": 12594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4753351038246796e-06,
+      "loss": 0.0738,
+      "step": 12595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4752932811944544e-06,
+      "loss": 0.6716,
+      "step": 12596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4752514585642292e-06,
+      "loss": 0.0431,
+      "step": 12597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4752096359340037e-06,
+      "loss": 0.0008,
+      "step": 12598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4751678133037785e-06,
+      "loss": 0.0004,
+      "step": 12599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4751259906735533e-06,
+      "loss": 0.0134,
+      "step": 12600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.475084168043328e-06,
+      "loss": 0.0796,
+      "step": 12601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4750423454131029e-06,
+      "loss": 0.0055,
+      "step": 12602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4750005227828777e-06,
+      "loss": 0.0061,
+      "step": 12603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4749587001526525e-06,
+      "loss": 0.2922,
+      "step": 12604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4749168775224272e-06,
+      "loss": 0.0101,
+      "step": 12605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4748750548922022e-06,
+      "loss": 0.0027,
+      "step": 12606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474833232261977e-06,
+      "loss": 0.0912,
+      "step": 12607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4747914096317518e-06,
+      "loss": 0.0015,
+      "step": 12608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4747495870015266e-06,
+      "loss": 0.0028,
+      "step": 12609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4747077643713014e-06,
+      "loss": 0.0009,
+      "step": 12610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474665941741076e-06,
+      "loss": 2.6408,
+      "step": 12611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4746241191108507e-06,
+      "loss": 1.865,
+      "step": 12612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4745822964806255e-06,
+      "loss": 0.0404,
+      "step": 12613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4745404738504003e-06,
+      "loss": 2.7332,
+      "step": 12614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474498651220175e-06,
+      "loss": 0.0229,
+      "step": 12615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4744568285899499e-06,
+      "loss": 1.861,
+      "step": 12616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4744150059597247e-06,
+      "loss": 0.0017,
+      "step": 12617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4743731833294995e-06,
+      "loss": 0.0018,
+      "step": 12618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4743313606992742e-06,
+      "loss": 0.0092,
+      "step": 12619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474289538069049e-06,
+      "loss": 0.001,
+      "step": 12620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474247715438824e-06,
+      "loss": 0.0073,
+      "step": 12621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4742058928085988e-06,
+      "loss": 0.7064,
+      "step": 12622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4741640701783736e-06,
+      "loss": 3.0109,
+      "step": 12623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4741222475481484e-06,
+      "loss": 0.0023,
+      "step": 12624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474080424917923e-06,
+      "loss": 0.7188,
+      "step": 12625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4740386022876978e-06,
+      "loss": 0.0022,
+      "step": 12626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4739967796574725e-06,
+      "loss": 0.0006,
+      "step": 12627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4739549570272473e-06,
+      "loss": 0.5597,
+      "step": 12628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4739131343970221e-06,
+      "loss": 0.0049,
+      "step": 12629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.473871311766797e-06,
+      "loss": 0.2863,
+      "step": 12630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4738294891365717e-06,
+      "loss": 1.1468,
+      "step": 12631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4737876665063465e-06,
+      "loss": 0.1106,
+      "step": 12632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4737458438761213e-06,
+      "loss": 0.0005,
+      "step": 12633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.473704021245896e-06,
+      "loss": 0.0008,
+      "step": 12634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.473662198615671e-06,
+      "loss": 0.5968,
+      "step": 12635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4736203759854458e-06,
+      "loss": 0.0008,
+      "step": 12636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4735785533552206e-06,
+      "loss": 0.0042,
+      "step": 12637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4735367307249952e-06,
+      "loss": 0.021,
+      "step": 12638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.47349490809477e-06,
+      "loss": 0.6029,
+      "step": 12639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4734530854645448e-06,
+      "loss": 0.0007,
+      "step": 12640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4734112628343195e-06,
+      "loss": 0.0033,
+      "step": 12641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4733694402040943e-06,
+      "loss": 0.0039,
+      "step": 12642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4733276175738691e-06,
+      "loss": 0.0011,
+      "step": 12643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.473285794943644e-06,
+      "loss": 0.0229,
+      "step": 12644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4732439723134187e-06,
+      "loss": 0.013,
+      "step": 12645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4732021496831935e-06,
+      "loss": 0.0003,
+      "step": 12646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4731603270529683e-06,
+      "loss": 0.0003,
+      "step": 12647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.473118504422743e-06,
+      "loss": 0.2552,
+      "step": 12648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4730766817925178e-06,
+      "loss": 0.0036,
+      "step": 12649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4730348591622928e-06,
+      "loss": 0.001,
+      "step": 12650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4729930365320676e-06,
+      "loss": 0.0143,
+      "step": 12651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4729512139018422e-06,
+      "loss": 0.044,
+      "step": 12652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472909391271617e-06,
+      "loss": 0.0005,
+      "step": 12653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4728675686413918e-06,
+      "loss": 0.0008,
+      "step": 12654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4728257460111666e-06,
+      "loss": 0.0085,
+      "step": 12655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4727839233809413e-06,
+      "loss": 0.0066,
+      "step": 12656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4727421007507161e-06,
+      "loss": 0.0008,
+      "step": 12657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472700278120491e-06,
+      "loss": 0.001,
+      "step": 12658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4726584554902657e-06,
+      "loss": 0.0042,
+      "step": 12659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4726166328600405e-06,
+      "loss": 0.1895,
+      "step": 12660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4725748102298153e-06,
+      "loss": 0.2518,
+      "step": 12661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.47253298759959e-06,
+      "loss": 0.0012,
+      "step": 12662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4724911649693648e-06,
+      "loss": 0.0026,
+      "step": 12663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4724493423391396e-06,
+      "loss": 0.0011,
+      "step": 12664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4724075197089146e-06,
+      "loss": 0.0038,
+      "step": 12665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4723656970786892e-06,
+      "loss": 1.6986,
+      "step": 12666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472323874448464e-06,
+      "loss": 0.2342,
+      "step": 12667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4722820518182388e-06,
+      "loss": 0.2175,
+      "step": 12668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4722402291880136e-06,
+      "loss": 0.0024,
+      "step": 12669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4721984065577884e-06,
+      "loss": 0.0241,
+      "step": 12670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4721565839275631e-06,
+      "loss": 0.0017,
+      "step": 12671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472114761297338e-06,
+      "loss": 0.0104,
+      "step": 12672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4720729386671127e-06,
+      "loss": 0.0039,
+      "step": 12673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4720311160368875e-06,
+      "loss": 0.0015,
+      "step": 12674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4719892934066623e-06,
+      "loss": 0.002,
+      "step": 12675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.471947470776437e-06,
+      "loss": 0.0004,
+      "step": 12676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4719056481462119e-06,
+      "loss": 0.103,
+      "step": 12677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4718638255159864e-06,
+      "loss": 0.003,
+      "step": 12678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4718220028857614e-06,
+      "loss": 0.0003,
+      "step": 12679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4717801802555362e-06,
+      "loss": 2.5168,
+      "step": 12680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.471738357625311e-06,
+      "loss": 0.0006,
+      "step": 12681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4716965349950858e-06,
+      "loss": 1.296,
+      "step": 12682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4716547123648606e-06,
+      "loss": 0.0013,
+      "step": 12683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4716128897346354e-06,
+      "loss": 0.0031,
+      "step": 12684
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4715710671044101e-06,
+      "loss": 0.001,
+      "step": 12685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.471529244474185e-06,
+      "loss": 0.0004,
+      "step": 12686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4714874218439597e-06,
+      "loss": 0.0084,
+      "step": 12687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4714455992137345e-06,
+      "loss": 0.0006,
+      "step": 12688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4714037765835093e-06,
+      "loss": 0.0008,
+      "step": 12689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.471361953953284e-06,
+      "loss": 0.0303,
+      "step": 12690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4713201313230589e-06,
+      "loss": 0.0075,
+      "step": 12691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4712783086928334e-06,
+      "loss": 0.0013,
+      "step": 12692
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4712364860626082e-06,
+      "loss": 0.0003,
+      "step": 12693
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4711946634323832e-06,
+      "loss": 0.0019,
+      "step": 12694
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.471152840802158e-06,
+      "loss": 0.0151,
+      "step": 12695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4711110181719328e-06,
+      "loss": 0.0007,
+      "step": 12696
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4710691955417076e-06,
+      "loss": 0.554,
+      "step": 12697
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4710273729114824e-06,
+      "loss": 0.0009,
+      "step": 12698
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4709855502812572e-06,
+      "loss": 0.0007,
+      "step": 12699
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.470943727651032e-06,
+      "loss": 0.0013,
+      "step": 12700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4709019050208067e-06,
+      "loss": 0.0869,
+      "step": 12701
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4708600823905815e-06,
+      "loss": 0.0004,
+      "step": 12702
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4708182597603563e-06,
+      "loss": 0.0017,
+      "step": 12703
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.470776437130131e-06,
+      "loss": 2.38,
+      "step": 12704
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4707346144999059e-06,
+      "loss": 0.0003,
+      "step": 12705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4706927918696804e-06,
+      "loss": 0.0011,
+      "step": 12706
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4706509692394552e-06,
+      "loss": 2.5222,
+      "step": 12707
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.47060914660923e-06,
+      "loss": 0.0016,
+      "step": 12708
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.470567323979005e-06,
+      "loss": 0.1763,
+      "step": 12709
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4705255013487798e-06,
+      "loss": 0.0011,
+      "step": 12710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4704836787185546e-06,
+      "loss": 0.0003,
+      "step": 12711
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4704418560883294e-06,
+      "loss": 0.0016,
+      "step": 12712
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4704000334581042e-06,
+      "loss": 0.001,
+      "step": 12713
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.470358210827879e-06,
+      "loss": 0.0004,
+      "step": 12714
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4703163881976537e-06,
+      "loss": 0.7639,
+      "step": 12715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4702745655674285e-06,
+      "loss": 0.0005,
+      "step": 12716
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4702327429372033e-06,
+      "loss": 0.0039,
+      "step": 12717
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.470190920306978e-06,
+      "loss": 0.0164,
+      "step": 12718
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4701490976767527e-06,
+      "loss": 0.0004,
+      "step": 12719
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4701072750465275e-06,
+      "loss": 0.0177,
+      "step": 12720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4700654524163022e-06,
+      "loss": 0.0037,
+      "step": 12721
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.470023629786077e-06,
+      "loss": 0.001,
+      "step": 12722
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469981807155852e-06,
+      "loss": 3.0207,
+      "step": 12723
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4699399845256268e-06,
+      "loss": 0.0004,
+      "step": 12724
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4698981618954016e-06,
+      "loss": 0.0005,
+      "step": 12725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4698563392651764e-06,
+      "loss": 0.0011,
+      "step": 12726
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4698145166349512e-06,
+      "loss": 0.0786,
+      "step": 12727
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469772694004726e-06,
+      "loss": 0.0006,
+      "step": 12728
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4697308713745007e-06,
+      "loss": 0.0116,
+      "step": 12729
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4696890487442755e-06,
+      "loss": 0.0122,
+      "step": 12730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4696472261140503e-06,
+      "loss": 3.5522,
+      "step": 12731
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469605403483825e-06,
+      "loss": 0.0003,
+      "step": 12732
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4695635808535997e-06,
+      "loss": 0.0009,
+      "step": 12733
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4695217582233745e-06,
+      "loss": 0.0003,
+      "step": 12734
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4694799355931493e-06,
+      "loss": 0.0009,
+      "step": 12735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469438112962924e-06,
+      "loss": 0.0005,
+      "step": 12736
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4693962903326988e-06,
+      "loss": 2.2889,
+      "step": 12737
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4693544677024738e-06,
+      "loss": 0.0004,
+      "step": 12738
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4693126450722486e-06,
+      "loss": 0.0033,
+      "step": 12739
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4692708224420234e-06,
+      "loss": 0.0003,
+      "step": 12740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4692289998117982e-06,
+      "loss": 2.8052,
+      "step": 12741
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469187177181573e-06,
+      "loss": 0.0005,
+      "step": 12742
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4691453545513478e-06,
+      "loss": 0.0005,
+      "step": 12743
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4691035319211225e-06,
+      "loss": 2.6427,
+      "step": 12744
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4690617092908973e-06,
+      "loss": 0.0004,
+      "step": 12745
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469019886660672e-06,
+      "loss": 0.0029,
+      "step": 12746
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4689780640304467e-06,
+      "loss": 0.0005,
+      "step": 12747
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4689362414002215e-06,
+      "loss": 0.0196,
+      "step": 12748
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4688944187699963e-06,
+      "loss": 0.0023,
+      "step": 12749
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468852596139771e-06,
+      "loss": 0.0064,
+      "step": 12750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4688107735095458e-06,
+      "loss": 0.0003,
+      "step": 12751
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4687689508793208e-06,
+      "loss": 0.0003,
+      "step": 12752
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4687271282490956e-06,
+      "loss": 2.156,
+      "step": 12753
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4686853056188704e-06,
+      "loss": 0.858,
+      "step": 12754
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4686434829886452e-06,
+      "loss": 0.3676,
+      "step": 12755
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.46860166035842e-06,
+      "loss": 0.0004,
+      "step": 12756
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4685598377281948e-06,
+      "loss": 0.5664,
+      "step": 12757
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4685180150979695e-06,
+      "loss": 0.0037,
+      "step": 12758
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4684761924677443e-06,
+      "loss": 0.0113,
+      "step": 12759
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468434369837519e-06,
+      "loss": 0.0033,
+      "step": 12760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4683925472072937e-06,
+      "loss": 0.0013,
+      "step": 12761
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4683507245770685e-06,
+      "loss": 2.3398,
+      "step": 12762
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4683089019468433e-06,
+      "loss": 0.0083,
+      "step": 12763
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468267079316618e-06,
+      "loss": 0.0958,
+      "step": 12764
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4682252566863928e-06,
+      "loss": 0.0144,
+      "step": 12765
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4681834340561676e-06,
+      "loss": 0.3135,
+      "step": 12766
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4681416114259426e-06,
+      "loss": 0.0006,
+      "step": 12767
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4680997887957174e-06,
+      "loss": 2.1333,
+      "step": 12768
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4680579661654922e-06,
+      "loss": 0.2384,
+      "step": 12769
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468016143535267e-06,
+      "loss": 1.4611,
+      "step": 12770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4679743209050418e-06,
+      "loss": 0.2603,
+      "step": 12771
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4679324982748166e-06,
+      "loss": 0.0006,
+      "step": 12772
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4678906756445911e-06,
+      "loss": 0.0004,
+      "step": 12773
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467848853014366e-06,
+      "loss": 0.0005,
+      "step": 12774
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4678070303841407e-06,
+      "loss": 0.0014,
+      "step": 12775
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4677652077539155e-06,
+      "loss": 0.5882,
+      "step": 12776
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4677233851236903e-06,
+      "loss": 0.1262,
+      "step": 12777
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467681562493465e-06,
+      "loss": 0.0033,
+      "step": 12778
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4676397398632398e-06,
+      "loss": 2.7084,
+      "step": 12779
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4675979172330146e-06,
+      "loss": 0.0004,
+      "step": 12780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4675560946027894e-06,
+      "loss": 0.0284,
+      "step": 12781
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4675142719725644e-06,
+      "loss": 0.025,
+      "step": 12782
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4674724493423392e-06,
+      "loss": 0.5465,
+      "step": 12783
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467430626712114e-06,
+      "loss": 0.2286,
+      "step": 12784
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4673888040818888e-06,
+      "loss": 0.0017,
+      "step": 12785
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4673469814516636e-06,
+      "loss": 0.0004,
+      "step": 12786
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4673051588214381e-06,
+      "loss": 0.0021,
+      "step": 12787
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467263336191213e-06,
+      "loss": 0.5848,
+      "step": 12788
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4672215135609877e-06,
+      "loss": 0.0009,
+      "step": 12789
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4671796909307625e-06,
+      "loss": 0.0025,
+      "step": 12790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4671378683005373e-06,
+      "loss": 0.0009,
+      "step": 12791
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467096045670312e-06,
+      "loss": 0.003,
+      "step": 12792
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4670542230400869e-06,
+      "loss": 0.0006,
+      "step": 12793
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4670124004098616e-06,
+      "loss": 2.4906,
+      "step": 12794
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4669705777796364e-06,
+      "loss": 0.0024,
+      "step": 12795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4669287551494114e-06,
+      "loss": 0.1357,
+      "step": 12796
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4668869325191862e-06,
+      "loss": 0.0029,
+      "step": 12797
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.466845109888961e-06,
+      "loss": 0.0006,
+      "step": 12798
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4668032872587358e-06,
+      "loss": 0.0069,
+      "step": 12799
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4667614646285106e-06,
+      "loss": 0.0009,
+      "step": 12800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4667196419982851e-06,
+      "loss": 0.001,
+      "step": 12801
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.46667781936806e-06,
+      "loss": 0.3113,
+      "step": 12802
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4666359967378347e-06,
+      "loss": 0.0005,
+      "step": 12803
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4665941741076095e-06,
+      "loss": 0.0048,
+      "step": 12804
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4665523514773843e-06,
+      "loss": 0.1468,
+      "step": 12805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.466510528847159e-06,
+      "loss": 0.0031,
+      "step": 12806
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4664687062169339e-06,
+      "loss": 0.0039,
+      "step": 12807
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4664268835867087e-06,
+      "loss": 0.0978,
+      "step": 12808
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4663850609564834e-06,
+      "loss": 0.004,
+      "step": 12809
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4663432383262582e-06,
+      "loss": 3.5938,
+      "step": 12810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4663014156960332e-06,
+      "loss": 0.0006,
+      "step": 12811
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.466259593065808e-06,
+      "loss": 0.0835,
+      "step": 12812
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4662177704355828e-06,
+      "loss": 0.001,
+      "step": 12813
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4661759478053574e-06,
+      "loss": 0.0061,
+      "step": 12814
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4661341251751322e-06,
+      "loss": 0.0023,
+      "step": 12815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.466092302544907e-06,
+      "loss": 0.166,
+      "step": 12816
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4660504799146817e-06,
+      "loss": 0.0058,
+      "step": 12817
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4660086572844565e-06,
+      "loss": 0.4614,
+      "step": 12818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4659668346542313e-06,
+      "loss": 0.4596,
+      "step": 12819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465925012024006e-06,
+      "loss": 0.0168,
+      "step": 12820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4658831893937809e-06,
+      "loss": 0.0004,
+      "step": 12821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4658413667635557e-06,
+      "loss": 0.0092,
+      "step": 12822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4657995441333304e-06,
+      "loss": 0.0005,
+      "step": 12823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4657577215031052e-06,
+      "loss": 0.0035,
+      "step": 12824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.46571589887288e-06,
+      "loss": 0.0015,
+      "step": 12825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465674076242655e-06,
+      "loss": 0.0022,
+      "step": 12826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4656322536124298e-06,
+      "loss": 2.3127,
+      "step": 12827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4655904309822044e-06,
+      "loss": 0.0034,
+      "step": 12828
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4655486083519792e-06,
+      "loss": 1.024,
+      "step": 12829
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465506785721754e-06,
+      "loss": 0.0022,
+      "step": 12830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4654649630915287e-06,
+      "loss": 0.0059,
+      "step": 12831
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4654231404613035e-06,
+      "loss": 0.0007,
+      "step": 12832
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4653813178310783e-06,
+      "loss": 0.0028,
+      "step": 12833
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465339495200853e-06,
+      "loss": 2.8905,
+      "step": 12834
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4652976725706279e-06,
+      "loss": 0.901,
+      "step": 12835
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4652558499404027e-06,
+      "loss": 0.0005,
+      "step": 12836
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4652140273101775e-06,
+      "loss": 0.001,
+      "step": 12837
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4651722046799522e-06,
+      "loss": 0.0008,
+      "step": 12838
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465130382049727e-06,
+      "loss": 0.004,
+      "step": 12839
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465088559419502e-06,
+      "loss": 0.0007,
+      "step": 12840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4650467367892766e-06,
+      "loss": 0.0004,
+      "step": 12841
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4650049141590514e-06,
+      "loss": 0.0006,
+      "step": 12842
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4649630915288262e-06,
+      "loss": 0.0009,
+      "step": 12843
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.464921268898601e-06,
+      "loss": 0.0009,
+      "step": 12844
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4648794462683757e-06,
+      "loss": 0.0018,
+      "step": 12845
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4648376236381505e-06,
+      "loss": 2.2245,
+      "step": 12846
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4647958010079253e-06,
+      "loss": 1.7105,
+      "step": 12847
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4647539783777e-06,
+      "loss": 0.0059,
+      "step": 12848
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4647121557474749e-06,
+      "loss": 0.0044,
+      "step": 12849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4646703331172497e-06,
+      "loss": 0.3644,
+      "step": 12850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4646285104870245e-06,
+      "loss": 0.0019,
+      "step": 12851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4645866878567993e-06,
+      "loss": 0.0043,
+      "step": 12852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.464544865226574e-06,
+      "loss": 0.0022,
+      "step": 12853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4645030425963486e-06,
+      "loss": 0.0192,
+      "step": 12854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4644612199661236e-06,
+      "loss": 0.0008,
+      "step": 12855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4644193973358984e-06,
+      "loss": 0.2476,
+      "step": 12856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4643775747056732e-06,
+      "loss": 0.0004,
+      "step": 12857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.464335752075448e-06,
+      "loss": 0.0024,
+      "step": 12858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4642939294452228e-06,
+      "loss": 0.0503,
+      "step": 12859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4642521068149975e-06,
+      "loss": 0.0028,
+      "step": 12860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4642102841847723e-06,
+      "loss": 0.0015,
+      "step": 12861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4641684615545471e-06,
+      "loss": 0.0078,
+      "step": 12862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.464126638924322e-06,
+      "loss": 2.4543,
+      "step": 12863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4640848162940967e-06,
+      "loss": 0.013,
+      "step": 12864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4640429936638715e-06,
+      "loss": 1.604,
+      "step": 12865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4640011710336463e-06,
+      "loss": 0.0003,
+      "step": 12866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.463959348403421e-06,
+      "loss": 0.0088,
+      "step": 12867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4639175257731956e-06,
+      "loss": 0.821,
+      "step": 12868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4638757031429704e-06,
+      "loss": 0.0024,
+      "step": 12869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4638338805127454e-06,
+      "loss": 0.0019,
+      "step": 12870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4637920578825202e-06,
+      "loss": 0.0016,
+      "step": 12871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.463750235252295e-06,
+      "loss": 0.0073,
+      "step": 12872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4637084126220698e-06,
+      "loss": 0.0004,
+      "step": 12873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4636665899918445e-06,
+      "loss": 0.0012,
+      "step": 12874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4636247673616193e-06,
+      "loss": 0.0064,
+      "step": 12875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4635829447313941e-06,
+      "loss": 0.0063,
+      "step": 12876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.463541122101169e-06,
+      "loss": 0.0027,
+      "step": 12877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4634992994709437e-06,
+      "loss": 0.0038,
+      "step": 12878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4634574768407185e-06,
+      "loss": 2.0637,
+      "step": 12879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4634156542104933e-06,
+      "loss": 0.0058,
+      "step": 12880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4633738315802678e-06,
+      "loss": 1.4496,
+      "step": 12881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4633320089500426e-06,
+      "loss": 0.0009,
+      "step": 12882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4632901863198174e-06,
+      "loss": 0.2653,
+      "step": 12883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4632483636895924e-06,
+      "loss": 0.0004,
+      "step": 12884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4632065410593672e-06,
+      "loss": 0.0015,
+      "step": 12885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.463164718429142e-06,
+      "loss": 0.0027,
+      "step": 12886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4631228957989168e-06,
+      "loss": 0.0038,
+      "step": 12887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4630810731686916e-06,
+      "loss": 1.2952,
+      "step": 12888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4630392505384663e-06,
+      "loss": 0.0004,
+      "step": 12889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4629974279082411e-06,
+      "loss": 0.0008,
+      "step": 12890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462955605278016e-06,
+      "loss": 0.0075,
+      "step": 12891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4629137826477907e-06,
+      "loss": 0.0396,
+      "step": 12892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4628719600175655e-06,
+      "loss": 0.0023,
+      "step": 12893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4628301373873403e-06,
+      "loss": 0.0014,
+      "step": 12894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4627883147571149e-06,
+      "loss": 0.0014,
+      "step": 12895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4627464921268896e-06,
+      "loss": 0.0181,
+      "step": 12896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4627046694966644e-06,
+      "loss": 0.0008,
+      "step": 12897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4626628468664392e-06,
+      "loss": 0.0036,
+      "step": 12898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4626210242362142e-06,
+      "loss": 0.0019,
+      "step": 12899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462579201605989e-06,
+      "loss": 0.2069,
+      "step": 12900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4625373789757638e-06,
+      "loss": 0.0423,
+      "step": 12901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4624955563455386e-06,
+      "loss": 0.5871,
+      "step": 12902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4624537337153134e-06,
+      "loss": 0.0009,
+      "step": 12903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4624119110850881e-06,
+      "loss": 0.0234,
+      "step": 12904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462370088454863e-06,
+      "loss": 0.0058,
+      "step": 12905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4623282658246377e-06,
+      "loss": 0.0004,
+      "step": 12906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4622864431944125e-06,
+      "loss": 0.0008,
+      "step": 12907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4622446205641873e-06,
+      "loss": 0.0014,
+      "step": 12908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4622027979339619e-06,
+      "loss": 1.6964,
+      "step": 12909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4621609753037366e-06,
+      "loss": 0.0084,
+      "step": 12910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4621191526735114e-06,
+      "loss": 0.5362,
+      "step": 12911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4620773300432862e-06,
+      "loss": 0.0042,
+      "step": 12912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462035507413061e-06,
+      "loss": 0.0011,
+      "step": 12913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461993684782836e-06,
+      "loss": 2.7563,
+      "step": 12914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4619518621526108e-06,
+      "loss": 1.1198,
+      "step": 12915
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4619100395223856e-06,
+      "loss": 0.693,
+      "step": 12916
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4618682168921604e-06,
+      "loss": 0.0022,
+      "step": 12917
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4618263942619351e-06,
+      "loss": 0.0061,
+      "step": 12918
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.46178457163171e-06,
+      "loss": 0.2512,
+      "step": 12919
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4617427490014847e-06,
+      "loss": 0.0066,
+      "step": 12920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4617009263712595e-06,
+      "loss": 0.0005,
+      "step": 12921
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461659103741034e-06,
+      "loss": 3.6632,
+      "step": 12922
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4616172811108089e-06,
+      "loss": 0.0111,
+      "step": 12923
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4615754584805837e-06,
+      "loss": 0.0019,
+      "step": 12924
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4615336358503584e-06,
+      "loss": 0.0013,
+      "step": 12925
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4614918132201332e-06,
+      "loss": 0.0052,
+      "step": 12926
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461449990589908e-06,
+      "loss": 0.0026,
+      "step": 12927
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461408167959683e-06,
+      "loss": 0.0021,
+      "step": 12928
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4613663453294578e-06,
+      "loss": 0.1757,
+      "step": 12929
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4613245226992326e-06,
+      "loss": 0.0037,
+      "step": 12930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4612827000690074e-06,
+      "loss": 0.0006,
+      "step": 12931
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4612408774387822e-06,
+      "loss": 0.7489,
+      "step": 12932
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461199054808557e-06,
+      "loss": 0.0028,
+      "step": 12933
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4611572321783317e-06,
+      "loss": 0.0089,
+      "step": 12934
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4611154095481065e-06,
+      "loss": 0.0012,
+      "step": 12935
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461073586917881e-06,
+      "loss": 0.0067,
+      "step": 12936
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4610317642876559e-06,
+      "loss": 0.0011,
+      "step": 12937
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4609899416574307e-06,
+      "loss": 0.0029,
+      "step": 12938
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4609481190272054e-06,
+      "loss": 0.0009,
+      "step": 12939
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4609062963969802e-06,
+      "loss": 0.002,
+      "step": 12940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460864473766755e-06,
+      "loss": 0.0011,
+      "step": 12941
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4608226511365298e-06,
+      "loss": 1.6699,
+      "step": 12942
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4607808285063048e-06,
+      "loss": 0.0004,
+      "step": 12943
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4607390058760796e-06,
+      "loss": 0.0041,
+      "step": 12944
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4606971832458544e-06,
+      "loss": 0.6181,
+      "step": 12945
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4606553606156292e-06,
+      "loss": 1.4218,
+      "step": 12946
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460613537985404e-06,
+      "loss": 0.0155,
+      "step": 12947
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4605717153551787e-06,
+      "loss": 0.173,
+      "step": 12948
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4605298927249533e-06,
+      "loss": 0.0052,
+      "step": 12949
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460488070094728e-06,
+      "loss": 0.0006,
+      "step": 12950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4604462474645029e-06,
+      "loss": 0.0023,
+      "step": 12951
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4604044248342777e-06,
+      "loss": 0.0006,
+      "step": 12952
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4603626022040525e-06,
+      "loss": 0.0016,
+      "step": 12953
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4603207795738272e-06,
+      "loss": 0.0026,
+      "step": 12954
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460278956943602e-06,
+      "loss": 2.7212,
+      "step": 12955
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4602371343133768e-06,
+      "loss": 0.0016,
+      "step": 12956
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4601953116831518e-06,
+      "loss": 0.0017,
+      "step": 12957
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4601534890529266e-06,
+      "loss": 0.0084,
+      "step": 12958
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4601116664227014e-06,
+      "loss": 0.0064,
+      "step": 12959
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4600698437924762e-06,
+      "loss": 2.7932,
+      "step": 12960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460028021162251e-06,
+      "loss": 1.0552,
+      "step": 12961
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4599861985320257e-06,
+      "loss": 0.0016,
+      "step": 12962
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4599443759018003e-06,
+      "loss": 0.0026,
+      "step": 12963
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4599025532715751e-06,
+      "loss": 0.014,
+      "step": 12964
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4598607306413499e-06,
+      "loss": 0.0012,
+      "step": 12965
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4598189080111247e-06,
+      "loss": 0.0025,
+      "step": 12966
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4597770853808995e-06,
+      "loss": 0.0012,
+      "step": 12967
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4597352627506743e-06,
+      "loss": 0.0038,
+      "step": 12968
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.459693440120449e-06,
+      "loss": 1.3212,
+      "step": 12969
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4596516174902238e-06,
+      "loss": 0.0032,
+      "step": 12970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4596097948599986e-06,
+      "loss": 1.6548,
+      "step": 12971
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4595679722297736e-06,
+      "loss": 0.8442,
+      "step": 12972
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4595261495995484e-06,
+      "loss": 0.0008,
+      "step": 12973
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4594843269693232e-06,
+      "loss": 0.3521,
+      "step": 12974
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.459442504339098e-06,
+      "loss": 0.1253,
+      "step": 12975
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4594006817088728e-06,
+      "loss": 0.0038,
+      "step": 12976
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4593588590786473e-06,
+      "loss": 0.0041,
+      "step": 12977
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4593170364484221e-06,
+      "loss": 0.0013,
+      "step": 12978
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.459275213818197e-06,
+      "loss": 0.0007,
+      "step": 12979
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4592333911879717e-06,
+      "loss": 0.0043,
+      "step": 12980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4591915685577465e-06,
+      "loss": 0.0205,
+      "step": 12981
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4591497459275213e-06,
+      "loss": 2.1317,
+      "step": 12982
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.459107923297296e-06,
+      "loss": 0.0018,
+      "step": 12983
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4590661006670708e-06,
+      "loss": 0.0375,
+      "step": 12984
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4590242780368456e-06,
+      "loss": 0.0029,
+      "step": 12985
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4589824554066204e-06,
+      "loss": 0.4896,
+      "step": 12986
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4589406327763954e-06,
+      "loss": 0.001,
+      "step": 12987
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4588988101461702e-06,
+      "loss": 0.3982,
+      "step": 12988
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458856987515945e-06,
+      "loss": 0.1782,
+      "step": 12989
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4588151648857196e-06,
+      "loss": 0.0023,
+      "step": 12990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4587733422554943e-06,
+      "loss": 0.0027,
+      "step": 12991
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4587315196252691e-06,
+      "loss": 0.7233,
+      "step": 12992
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458689696995044e-06,
+      "loss": 0.0014,
+      "step": 12993
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4586478743648187e-06,
+      "loss": 0.0008,
+      "step": 12994
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4586060517345935e-06,
+      "loss": 0.6316,
+      "step": 12995
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4585642291043683e-06,
+      "loss": 0.048,
+      "step": 12996
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458522406474143e-06,
+      "loss": 0.7061,
+      "step": 12997
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4584805838439178e-06,
+      "loss": 0.0032,
+      "step": 12998
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4584387612136926e-06,
+      "loss": 0.0023,
+      "step": 12999
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4583969385834674e-06,
+      "loss": 0.0174,
+      "step": 13000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4583551159532424e-06,
+      "loss": 0.1155,
+      "step": 13001
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4583132933230172e-06,
+      "loss": 0.0098,
+      "step": 13002
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458271470692792e-06,
+      "loss": 0.0009,
+      "step": 13003
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4582296480625666e-06,
+      "loss": 0.0009,
+      "step": 13004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4581878254323413e-06,
+      "loss": 0.0006,
+      "step": 13005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4581460028021161e-06,
+      "loss": 0.1283,
+      "step": 13006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.458104180171891e-06,
+      "loss": 0.1499,
+      "step": 13007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4580623575416657e-06,
+      "loss": 1.2183,
+      "step": 13008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4580205349114405e-06,
+      "loss": 0.0313,
+      "step": 13009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4579787122812153e-06,
+      "loss": 0.0715,
+      "step": 13010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.45793688965099e-06,
+      "loss": 0.0006,
+      "step": 13011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4578950670207649e-06,
+      "loss": 0.0014,
+      "step": 13012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4578532443905396e-06,
+      "loss": 0.8585,
+      "step": 13013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4578114217603144e-06,
+      "loss": 0.0009,
+      "step": 13014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4577695991300892e-06,
+      "loss": 0.0018,
+      "step": 13015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4577277764998642e-06,
+      "loss": 0.0006,
+      "step": 13016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4576859538696388e-06,
+      "loss": 0.0055,
+      "step": 13017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4576441312394136e-06,
+      "loss": 0.0022,
+      "step": 13018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4576023086091884e-06,
+      "loss": 0.0008,
+      "step": 13019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4575604859789631e-06,
+      "loss": 0.0035,
+      "step": 13020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.457518663348738e-06,
+      "loss": 0.0013,
+      "step": 13021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4574768407185127e-06,
+      "loss": 0.0031,
+      "step": 13022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4574350180882875e-06,
+      "loss": 0.0233,
+      "step": 13023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4573931954580623e-06,
+      "loss": 3.3925,
+      "step": 13024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.457351372827837e-06,
+      "loss": 0.0338,
+      "step": 13025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4573095501976119e-06,
+      "loss": 0.0013,
+      "step": 13026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4572677275673866e-06,
+      "loss": 1.5987,
+      "step": 13027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4572259049371614e-06,
+      "loss": 0.0017,
+      "step": 13028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4571840823069362e-06,
+      "loss": 0.7472,
+      "step": 13029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4571422596767108e-06,
+      "loss": 0.0016,
+      "step": 13030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4571004370464858e-06,
+      "loss": 0.0048,
+      "step": 13031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4570586144162606e-06,
+      "loss": 0.0004,
+      "step": 13032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4570167917860354e-06,
+      "loss": 0.001,
+      "step": 13033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4569749691558102e-06,
+      "loss": 0.0011,
+      "step": 13034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.456933146525585e-06,
+      "loss": 0.4996,
+      "step": 13035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4568913238953597e-06,
+      "loss": 0.0008,
+      "step": 13036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4568495012651345e-06,
+      "loss": 0.001,
+      "step": 13037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4568076786349093e-06,
+      "loss": 0.001,
+      "step": 13038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.456765856004684e-06,
+      "loss": 0.0049,
+      "step": 13039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4567240333744589e-06,
+      "loss": 0.002,
+      "step": 13040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4566822107442337e-06,
+      "loss": 0.0011,
+      "step": 13041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4566403881140084e-06,
+      "loss": 0.0043,
+      "step": 13042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4565985654837832e-06,
+      "loss": 0.0008,
+      "step": 13043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4565567428535578e-06,
+      "loss": 0.001,
+      "step": 13044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4565149202233328e-06,
+      "loss": 0.0007,
+      "step": 13045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4564730975931076e-06,
+      "loss": 0.0005,
+      "step": 13046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4564312749628824e-06,
+      "loss": 0.0022,
+      "step": 13047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4563894523326572e-06,
+      "loss": 0.0018,
+      "step": 13048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.456347629702432e-06,
+      "loss": 0.0005,
+      "step": 13049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4563058070722067e-06,
+      "loss": 0.7014,
+      "step": 13050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4562639844419815e-06,
+      "loss": 0.0028,
+      "step": 13051
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4562221618117563e-06,
+      "loss": 0.0383,
+      "step": 13052
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.456180339181531e-06,
+      "loss": 0.1232,
+      "step": 13053
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4561385165513059e-06,
+      "loss": 0.0026,
+      "step": 13054
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4560966939210807e-06,
+      "loss": 0.0037,
+      "step": 13055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4560548712908555e-06,
+      "loss": 0.0007,
+      "step": 13056
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.45601304866063e-06,
+      "loss": 0.0005,
+      "step": 13057
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4559712260304048e-06,
+      "loss": 0.0053,
+      "step": 13058
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4559294034001796e-06,
+      "loss": 0.0021,
+      "step": 13059
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4558875807699546e-06,
+      "loss": 0.0044,
+      "step": 13060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4558457581397294e-06,
+      "loss": 0.0359,
+      "step": 13061
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4558039355095042e-06,
+      "loss": 0.0004,
+      "step": 13062
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455762112879279e-06,
+      "loss": 0.0017,
+      "step": 13063
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4557202902490537e-06,
+      "loss": 0.0007,
+      "step": 13064
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4556784676188285e-06,
+      "loss": 2.2652,
+      "step": 13065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4556366449886033e-06,
+      "loss": 0.0026,
+      "step": 13066
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455594822358378e-06,
+      "loss": 0.0006,
+      "step": 13067
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4555529997281529e-06,
+      "loss": 0.0003,
+      "step": 13068
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4555111770979277e-06,
+      "loss": 0.0018,
+      "step": 13069
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4554693544677025e-06,
+      "loss": 0.0017,
+      "step": 13070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455427531837477e-06,
+      "loss": 0.0041,
+      "step": 13071
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4553857092072518e-06,
+      "loss": 0.349,
+      "step": 13072
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4553438865770266e-06,
+      "loss": 0.0032,
+      "step": 13073
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4553020639468014e-06,
+      "loss": 0.0004,
+      "step": 13074
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4552602413165764e-06,
+      "loss": 0.1073,
+      "step": 13075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4552184186863512e-06,
+      "loss": 0.0035,
+      "step": 13076
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455176596056126e-06,
+      "loss": 0.3009,
+      "step": 13077
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4551347734259007e-06,
+      "loss": 2.3518,
+      "step": 13078
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4550929507956755e-06,
+      "loss": 0.0007,
+      "step": 13079
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4550511281654503e-06,
+      "loss": 0.0015,
+      "step": 13080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4550093055352251e-06,
+      "loss": 1.3422,
+      "step": 13081
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454967482905e-06,
+      "loss": 1.9215,
+      "step": 13082
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4549256602747747e-06,
+      "loss": 0.0057,
+      "step": 13083
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4548838376445493e-06,
+      "loss": 2.9215,
+      "step": 13084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454842015014324e-06,
+      "loss": 1.5291,
+      "step": 13085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4548001923840988e-06,
+      "loss": 0.0003,
+      "step": 13086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4547583697538736e-06,
+      "loss": 0.0023,
+      "step": 13087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4547165471236484e-06,
+      "loss": 0.0009,
+      "step": 13088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4546747244934234e-06,
+      "loss": 0.0021,
+      "step": 13089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4546329018631982e-06,
+      "loss": 0.0039,
+      "step": 13090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454591079232973e-06,
+      "loss": 0.0069,
+      "step": 13091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4545492566027478e-06,
+      "loss": 0.0049,
+      "step": 13092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4545074339725225e-06,
+      "loss": 0.0008,
+      "step": 13093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4544656113422973e-06,
+      "loss": 0.001,
+      "step": 13094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4544237887120721e-06,
+      "loss": 0.6137,
+      "step": 13095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454381966081847e-06,
+      "loss": 0.0122,
+      "step": 13096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4543401434516217e-06,
+      "loss": 0.0804,
+      "step": 13097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4542983208213963e-06,
+      "loss": 1.4079,
+      "step": 13098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454256498191171e-06,
+      "loss": 0.0028,
+      "step": 13099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4542146755609458e-06,
+      "loss": 2.7489,
+      "step": 13100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4541728529307206e-06,
+      "loss": 0.0068,
+      "step": 13101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4541310303004954e-06,
+      "loss": 0.0006,
+      "step": 13102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4540892076702702e-06,
+      "loss": 0.0015,
+      "step": 13103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4540473850400452e-06,
+      "loss": 0.0009,
+      "step": 13104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.45400556240982e-06,
+      "loss": 0.0031,
+      "step": 13105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4539637397795948e-06,
+      "loss": 0.0016,
+      "step": 13106
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4539219171493696e-06,
+      "loss": 0.005,
+      "step": 13107
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4538800945191443e-06,
+      "loss": 0.0061,
+      "step": 13108
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4538382718889191e-06,
+      "loss": 0.0005,
+      "step": 13109
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453796449258694e-06,
+      "loss": 0.003,
+      "step": 13110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4537546266284687e-06,
+      "loss": 1.1644,
+      "step": 13111
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4537128039982433e-06,
+      "loss": 0.3464,
+      "step": 13112
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453670981368018e-06,
+      "loss": 0.6064,
+      "step": 13113
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4536291587377928e-06,
+      "loss": 0.001,
+      "step": 13114
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4535873361075676e-06,
+      "loss": 0.0004,
+      "step": 13115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4535455134773424e-06,
+      "loss": 0.001,
+      "step": 13116
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4535036908471172e-06,
+      "loss": 0.2043,
+      "step": 13117
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453461868216892e-06,
+      "loss": 0.0003,
+      "step": 13118
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453420045586667e-06,
+      "loss": 0.0006,
+      "step": 13119
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4533782229564418e-06,
+      "loss": 0.0896,
+      "step": 13120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4533364003262166e-06,
+      "loss": 0.0046,
+      "step": 13121
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4532945776959913e-06,
+      "loss": 0.0053,
+      "step": 13122
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4532527550657661e-06,
+      "loss": 0.0006,
+      "step": 13123
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453210932435541e-06,
+      "loss": 0.0049,
+      "step": 13124
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4531691098053155e-06,
+      "loss": 0.0032,
+      "step": 13125
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4531272871750903e-06,
+      "loss": 3.4664,
+      "step": 13126
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453085464544865e-06,
+      "loss": 0.0116,
+      "step": 13127
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4530436419146399e-06,
+      "loss": 0.0009,
+      "step": 13128
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4530018192844146e-06,
+      "loss": 0.0011,
+      "step": 13129
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4529599966541894e-06,
+      "loss": 0.008,
+      "step": 13130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4529181740239642e-06,
+      "loss": 0.0249,
+      "step": 13131
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.452876351393739e-06,
+      "loss": 0.4376,
+      "step": 13132
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.452834528763514e-06,
+      "loss": 0.0024,
+      "step": 13133
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4527927061332888e-06,
+      "loss": 0.2229,
+      "step": 13134
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4527508835030636e-06,
+      "loss": 0.0038,
+      "step": 13135
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4527090608728384e-06,
+      "loss": 0.0041,
+      "step": 13136
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4526672382426131e-06,
+      "loss": 0.0005,
+      "step": 13137
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.452625415612388e-06,
+      "loss": 0.0033,
+      "step": 13138
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4525835929821625e-06,
+      "loss": 0.0437,
+      "step": 13139
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4525417703519373e-06,
+      "loss": 0.0013,
+      "step": 13140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.452499947721712e-06,
+      "loss": 0.007,
+      "step": 13141
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4524581250914869e-06,
+      "loss": 0.0011,
+      "step": 13142
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4524163024612616e-06,
+      "loss": 0.0007,
+      "step": 13143
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4523744798310364e-06,
+      "loss": 0.0008,
+      "step": 13144
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4523326572008112e-06,
+      "loss": 0.0017,
+      "step": 13145
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.452290834570586e-06,
+      "loss": 0.0003,
+      "step": 13146
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4522490119403608e-06,
+      "loss": 0.0008,
+      "step": 13147
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4522071893101358e-06,
+      "loss": 0.0006,
+      "step": 13148
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4521653666799106e-06,
+      "loss": 0.0008,
+      "step": 13149
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4521235440496854e-06,
+      "loss": 0.0031,
+      "step": 13150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4520817214194602e-06,
+      "loss": 0.0008,
+      "step": 13151
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4520398987892347e-06,
+      "loss": 1.6422,
+      "step": 13152
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4519980761590095e-06,
+      "loss": 0.0003,
+      "step": 13153
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4519562535287843e-06,
+      "loss": 0.0025,
+      "step": 13154
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.451914430898559e-06,
+      "loss": 0.0005,
+      "step": 13155
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4518726082683339e-06,
+      "loss": 0.0012,
+      "step": 13156
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4518307856381087e-06,
+      "loss": 1.3339,
+      "step": 13157
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4517889630078834e-06,
+      "loss": 2.7773,
+      "step": 13158
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4517471403776582e-06,
+      "loss": 0.0007,
+      "step": 13159
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.451705317747433e-06,
+      "loss": 0.0008,
+      "step": 13160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4516634951172078e-06,
+      "loss": 0.8608,
+      "step": 13161
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4516216724869828e-06,
+      "loss": 0.0005,
+      "step": 13162
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4515798498567576e-06,
+      "loss": 0.003,
+      "step": 13163
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4515380272265324e-06,
+      "loss": 0.0025,
+      "step": 13164
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4514962045963072e-06,
+      "loss": 0.0045,
+      "step": 13165
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4514543819660817e-06,
+      "loss": 0.0051,
+      "step": 13166
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4514125593358565e-06,
+      "loss": 0.0004,
+      "step": 13167
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4513707367056313e-06,
+      "loss": 0.0006,
+      "step": 13168
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.451328914075406e-06,
+      "loss": 0.0005,
+      "step": 13169
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4512870914451809e-06,
+      "loss": 0.0038,
+      "step": 13170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4512452688149557e-06,
+      "loss": 0.0006,
+      "step": 13171
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4512034461847305e-06,
+      "loss": 0.0013,
+      "step": 13172
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4511616235545052e-06,
+      "loss": 2.8283,
+      "step": 13173
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.45111980092428e-06,
+      "loss": 0.0005,
+      "step": 13174
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4510779782940548e-06,
+      "loss": 1.0853,
+      "step": 13175
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4510361556638296e-06,
+      "loss": 0.002,
+      "step": 13176
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4509943330336046e-06,
+      "loss": 0.001,
+      "step": 13177
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4509525104033794e-06,
+      "loss": 0.0007,
+      "step": 13178
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4509106877731542e-06,
+      "loss": 0.2413,
+      "step": 13179
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4508688651429287e-06,
+      "loss": 0.0021,
+      "step": 13180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4508270425127035e-06,
+      "loss": 0.0026,
+      "step": 13181
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4507852198824783e-06,
+      "loss": 0.6774,
+      "step": 13182
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.450743397252253e-06,
+      "loss": 0.0106,
+      "step": 13183
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4507015746220279e-06,
+      "loss": 0.0247,
+      "step": 13184
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4506597519918027e-06,
+      "loss": 0.0083,
+      "step": 13185
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4506179293615775e-06,
+      "loss": 0.0003,
+      "step": 13186
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4505761067313522e-06,
+      "loss": 0.0004,
+      "step": 13187
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.450534284101127e-06,
+      "loss": 0.001,
+      "step": 13188
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4504924614709018e-06,
+      "loss": 0.3823,
+      "step": 13189
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4504506388406766e-06,
+      "loss": 0.0019,
+      "step": 13190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4504088162104514e-06,
+      "loss": 0.0024,
+      "step": 13191
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4503669935802264e-06,
+      "loss": 1.3025,
+      "step": 13192
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.450325170950001e-06,
+      "loss": 0.0014,
+      "step": 13193
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4502833483197758e-06,
+      "loss": 0.0021,
+      "step": 13194
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4502415256895505e-06,
+      "loss": 2.5129,
+      "step": 13195
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4501997030593253e-06,
+      "loss": 0.0031,
+      "step": 13196
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4501578804291001e-06,
+      "loss": 2.1697,
+      "step": 13197
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.450116057798875e-06,
+      "loss": 0.0009,
+      "step": 13198
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4500742351686497e-06,
+      "loss": 0.0019,
+      "step": 13199
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4500324125384245e-06,
+      "loss": 0.0005,
+      "step": 13200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4499905899081993e-06,
+      "loss": 0.0022,
+      "step": 13201
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449948767277974e-06,
+      "loss": 0.0007,
+      "step": 13202
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4499069446477488e-06,
+      "loss": 0.0047,
+      "step": 13203
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4498651220175236e-06,
+      "loss": 0.0054,
+      "step": 13204
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4498232993872984e-06,
+      "loss": 0.0015,
+      "step": 13205
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4497814767570734e-06,
+      "loss": 0.0003,
+      "step": 13206
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449739654126848e-06,
+      "loss": 0.0042,
+      "step": 13207
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4496978314966228e-06,
+      "loss": 0.0005,
+      "step": 13208
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4496560088663975e-06,
+      "loss": 0.0003,
+      "step": 13209
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4496141862361723e-06,
+      "loss": 0.001,
+      "step": 13210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4495723636059471e-06,
+      "loss": 0.0005,
+      "step": 13211
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449530540975722e-06,
+      "loss": 0.2068,
+      "step": 13212
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4494887183454967e-06,
+      "loss": 0.0052,
+      "step": 13213
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4494468957152715e-06,
+      "loss": 0.0008,
+      "step": 13214
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4494050730850463e-06,
+      "loss": 0.0015,
+      "step": 13215
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449363250454821e-06,
+      "loss": 0.0679,
+      "step": 13216
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4493214278245958e-06,
+      "loss": 0.0006,
+      "step": 13217
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4492796051943706e-06,
+      "loss": 0.0003,
+      "step": 13218
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4492377825641454e-06,
+      "loss": 0.0024,
+      "step": 13219
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44919595993392e-06,
+      "loss": 0.0025,
+      "step": 13220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449154137303695e-06,
+      "loss": 0.0051,
+      "step": 13221
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4491123146734698e-06,
+      "loss": 0.0031,
+      "step": 13222
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4490704920432446e-06,
+      "loss": 0.0028,
+      "step": 13223
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4490286694130193e-06,
+      "loss": 0.003,
+      "step": 13224
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4489868467827941e-06,
+      "loss": 0.1161,
+      "step": 13225
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448945024152569e-06,
+      "loss": 1.3245,
+      "step": 13226
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4489032015223437e-06,
+      "loss": 0.0005,
+      "step": 13227
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4488613788921185e-06,
+      "loss": 0.0008,
+      "step": 13228
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4488195562618933e-06,
+      "loss": 0.0006,
+      "step": 13229
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448777733631668e-06,
+      "loss": 0.0015,
+      "step": 13230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4487359110014428e-06,
+      "loss": 0.7277,
+      "step": 13231
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4486940883712176e-06,
+      "loss": 0.0003,
+      "step": 13232
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4486522657409922e-06,
+      "loss": 0.9499,
+      "step": 13233
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448610443110767e-06,
+      "loss": 0.0003,
+      "step": 13234
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4485686204805418e-06,
+      "loss": 0.0004,
+      "step": 13235
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4485267978503168e-06,
+      "loss": 0.7068,
+      "step": 13236
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4484849752200916e-06,
+      "loss": 0.0083,
+      "step": 13237
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4484431525898664e-06,
+      "loss": 0.001,
+      "step": 13238
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4484013299596411e-06,
+      "loss": 0.3626,
+      "step": 13239
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448359507329416e-06,
+      "loss": 0.0003,
+      "step": 13240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4483176846991907e-06,
+      "loss": 0.0095,
+      "step": 13241
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4482758620689655e-06,
+      "loss": 0.0006,
+      "step": 13242
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4482340394387403e-06,
+      "loss": 0.0018,
+      "step": 13243
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448192216808515e-06,
+      "loss": 2.0367,
+      "step": 13244
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4481503941782899e-06,
+      "loss": 0.0058,
+      "step": 13245
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4481085715480646e-06,
+      "loss": 0.0027,
+      "step": 13246
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4480667489178392e-06,
+      "loss": 0.0024,
+      "step": 13247
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448024926287614e-06,
+      "loss": 0.0004,
+      "step": 13248
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4479831036573888e-06,
+      "loss": 0.0007,
+      "step": 13249
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4479412810271638e-06,
+      "loss": 0.0017,
+      "step": 13250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4478994583969386e-06,
+      "loss": 0.0004,
+      "step": 13251
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4478576357667134e-06,
+      "loss": 0.0003,
+      "step": 13252
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4478158131364881e-06,
+      "loss": 0.0012,
+      "step": 13253
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447773990506263e-06,
+      "loss": 0.0031,
+      "step": 13254
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4477321678760377e-06,
+      "loss": 0.6348,
+      "step": 13255
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4476903452458125e-06,
+      "loss": 0.0068,
+      "step": 13256
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4476485226155873e-06,
+      "loss": 2.353,
+      "step": 13257
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447606699985362e-06,
+      "loss": 0.0018,
+      "step": 13258
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4475648773551369e-06,
+      "loss": 0.5546,
+      "step": 13259
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4475230547249114e-06,
+      "loss": 0.0003,
+      "step": 13260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4474812320946862e-06,
+      "loss": 0.0023,
+      "step": 13261
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447439409464461e-06,
+      "loss": 0.0014,
+      "step": 13262
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4473975868342358e-06,
+      "loss": 0.0003,
+      "step": 13263
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4473557642040106e-06,
+      "loss": 0.0138,
+      "step": 13264
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4473139415737856e-06,
+      "loss": 0.0025,
+      "step": 13265
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4472721189435604e-06,
+      "loss": 0.0018,
+      "step": 13266
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4472302963133352e-06,
+      "loss": 2.7877,
+      "step": 13267
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44718847368311e-06,
+      "loss": 0.025,
+      "step": 13268
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4471466510528847e-06,
+      "loss": 0.0006,
+      "step": 13269
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4471048284226595e-06,
+      "loss": 0.0829,
+      "step": 13270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4470630057924343e-06,
+      "loss": 0.5625,
+      "step": 13271
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447021183162209e-06,
+      "loss": 0.0003,
+      "step": 13272
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4469793605319839e-06,
+      "loss": 0.0004,
+      "step": 13273
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4469375379017584e-06,
+      "loss": 0.0008,
+      "step": 13274
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4468957152715332e-06,
+      "loss": 0.0005,
+      "step": 13275
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446853892641308e-06,
+      "loss": 0.0003,
+      "step": 13276
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4468120700110828e-06,
+      "loss": 0.0011,
+      "step": 13277
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4467702473808576e-06,
+      "loss": 1.6512,
+      "step": 13278
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4467284247506324e-06,
+      "loss": 0.0026,
+      "step": 13279
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4466866021204074e-06,
+      "loss": 0.0062,
+      "step": 13280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4466447794901822e-06,
+      "loss": 0.0031,
+      "step": 13281
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446602956859957e-06,
+      "loss": 0.0003,
+      "step": 13282
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4465611342297317e-06,
+      "loss": 0.0004,
+      "step": 13283
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4465193115995065e-06,
+      "loss": 4.0449,
+      "step": 13284
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4464774889692813e-06,
+      "loss": 0.0004,
+      "step": 13285
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446435666339056e-06,
+      "loss": 3.4701,
+      "step": 13286
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4463938437088307e-06,
+      "loss": 3.0718,
+      "step": 13287
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4463520210786055e-06,
+      "loss": 0.0006,
+      "step": 13288
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4463101984483802e-06,
+      "loss": 0.0052,
+      "step": 13289
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446268375818155e-06,
+      "loss": 0.0004,
+      "step": 13290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4462265531879298e-06,
+      "loss": 0.0002,
+      "step": 13291
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4461847305577046e-06,
+      "loss": 0.0035,
+      "step": 13292
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4461429079274794e-06,
+      "loss": 0.0006,
+      "step": 13293
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4461010852972544e-06,
+      "loss": 2.4024,
+      "step": 13294
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4460592626670292e-06,
+      "loss": 0.0003,
+      "step": 13295
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446017440036804e-06,
+      "loss": 0.0006,
+      "step": 13296
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4459756174065787e-06,
+      "loss": 0.0086,
+      "step": 13297
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4459337947763535e-06,
+      "loss": 0.0081,
+      "step": 13298
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4458919721461283e-06,
+      "loss": 0.0003,
+      "step": 13299
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.445850149515903e-06,
+      "loss": 0.7127,
+      "step": 13300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4458083268856777e-06,
+      "loss": 0.0026,
+      "step": 13301
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4457665042554525e-06,
+      "loss": 0.0204,
+      "step": 13302
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4457246816252272e-06,
+      "loss": 0.0039,
+      "step": 13303
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.445682858995002e-06,
+      "loss": 0.0077,
+      "step": 13304
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4456410363647768e-06,
+      "loss": 2.9224,
+      "step": 13305
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4455992137345516e-06,
+      "loss": 0.0014,
+      "step": 13306
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4455573911043264e-06,
+      "loss": 0.0009,
+      "step": 13307
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4455155684741012e-06,
+      "loss": 0.0052,
+      "step": 13308
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4454737458438762e-06,
+      "loss": 0.0017,
+      "step": 13309
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.445431923213651e-06,
+      "loss": 0.0024,
+      "step": 13310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4453901005834258e-06,
+      "loss": 0.0101,
+      "step": 13311
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4453482779532005e-06,
+      "loss": 0.0031,
+      "step": 13312
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4453064553229753e-06,
+      "loss": 1.4329,
+      "step": 13313
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4452646326927501e-06,
+      "loss": 0.0021,
+      "step": 13314
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4452228100625247e-06,
+      "loss": 2.1182,
+      "step": 13315
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4451809874322995e-06,
+      "loss": 0.0004,
+      "step": 13316
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4451391648020743e-06,
+      "loss": 0.0068,
+      "step": 13317
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.445097342171849e-06,
+      "loss": 0.0006,
+      "step": 13318
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4450555195416238e-06,
+      "loss": 0.0006,
+      "step": 13319
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4450136969113986e-06,
+      "loss": 0.0077,
+      "step": 13320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4449718742811734e-06,
+      "loss": 0.0006,
+      "step": 13321
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4449300516509482e-06,
+      "loss": 0.0027,
+      "step": 13322
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444888229020723e-06,
+      "loss": 0.0013,
+      "step": 13323
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444846406390498e-06,
+      "loss": 0.0018,
+      "step": 13324
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4448045837602728e-06,
+      "loss": 1.1944,
+      "step": 13325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4447627611300475e-06,
+      "loss": 0.0014,
+      "step": 13326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4447209384998223e-06,
+      "loss": 0.0012,
+      "step": 13327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.444679115869597e-06,
+      "loss": 0.0008,
+      "step": 13328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4446372932393717e-06,
+      "loss": 0.0011,
+      "step": 13329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4445954706091465e-06,
+      "loss": 0.0447,
+      "step": 13330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4445536479789213e-06,
+      "loss": 0.2344,
+      "step": 13331
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.444511825348696e-06,
+      "loss": 0.0014,
+      "step": 13332
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4444700027184708e-06,
+      "loss": 0.0016,
+      "step": 13333
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4444281800882456e-06,
+      "loss": 0.0005,
+      "step": 13334
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4443863574580204e-06,
+      "loss": 0.0976,
+      "step": 13335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4443445348277952e-06,
+      "loss": 0.0087,
+      "step": 13336
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.44430271219757e-06,
+      "loss": 0.0008,
+      "step": 13337
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.444260889567345e-06,
+      "loss": 0.0014,
+      "step": 13338
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4442190669371198e-06,
+      "loss": 0.0028,
+      "step": 13339
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4441772443068946e-06,
+      "loss": 0.0003,
+      "step": 13340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4441354216766693e-06,
+      "loss": 0.002,
+      "step": 13341
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.444093599046444e-06,
+      "loss": 0.004,
+      "step": 13342
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4440517764162187e-06,
+      "loss": 0.007,
+      "step": 13343
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4440099537859935e-06,
+      "loss": 0.005,
+      "step": 13344
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4439681311557683e-06,
+      "loss": 0.0048,
+      "step": 13345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.443926308525543e-06,
+      "loss": 0.0011,
+      "step": 13346
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4438844858953178e-06,
+      "loss": 1.0623,
+      "step": 13347
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4438426632650926e-06,
+      "loss": 0.0013,
+      "step": 13348
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4438008406348674e-06,
+      "loss": 0.0018,
+      "step": 13349
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4437590180046422e-06,
+      "loss": 0.0007,
+      "step": 13350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.443717195374417e-06,
+      "loss": 0.0004,
+      "step": 13351
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4436753727441918e-06,
+      "loss": 0.0889,
+      "step": 13352
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4436335501139668e-06,
+      "loss": 0.0006,
+      "step": 13353
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4435917274837416e-06,
+      "loss": 0.0008,
+      "step": 13354
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4435499048535161e-06,
+      "loss": 4.5865,
+      "step": 13355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.443508082223291e-06,
+      "loss": 0.0004,
+      "step": 13356
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4434662595930657e-06,
+      "loss": 2.2587,
+      "step": 13357
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4434244369628405e-06,
+      "loss": 0.0004,
+      "step": 13358
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4433826143326153e-06,
+      "loss": 0.0013,
+      "step": 13359
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.44334079170239e-06,
+      "loss": 0.0014,
+      "step": 13360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4432989690721649e-06,
+      "loss": 0.0013,
+      "step": 13361
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4432571464419396e-06,
+      "loss": 0.0023,
+      "step": 13362
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4432153238117144e-06,
+      "loss": 0.0005,
+      "step": 13363
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4431735011814892e-06,
+      "loss": 0.001,
+      "step": 13364
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.443131678551264e-06,
+      "loss": 0.0004,
+      "step": 13365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4430898559210388e-06,
+      "loss": 0.0022,
+      "step": 13366
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4430480332908138e-06,
+      "loss": 0.002,
+      "step": 13367
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4430062106605886e-06,
+      "loss": 0.0033,
+      "step": 13368
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4429643880303631e-06,
+      "loss": 0.0004,
+      "step": 13369
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442922565400138e-06,
+      "loss": 0.5735,
+      "step": 13370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4428807427699127e-06,
+      "loss": 0.001,
+      "step": 13371
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4428389201396875e-06,
+      "loss": 0.0007,
+      "step": 13372
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4427970975094623e-06,
+      "loss": 0.0003,
+      "step": 13373
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442755274879237e-06,
+      "loss": 0.0017,
+      "step": 13374
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4427134522490119e-06,
+      "loss": 0.0034,
+      "step": 13375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4426716296187867e-06,
+      "loss": 0.0055,
+      "step": 13376
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4426298069885614e-06,
+      "loss": 0.0012,
+      "step": 13377
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4425879843583362e-06,
+      "loss": 0.0009,
+      "step": 13378
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442546161728111e-06,
+      "loss": 0.0013,
+      "step": 13379
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4425043390978858e-06,
+      "loss": 0.0297,
+      "step": 13380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4424625164676606e-06,
+      "loss": 0.0017,
+      "step": 13381
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4424206938374356e-06,
+      "loss": 0.0005,
+      "step": 13382
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4423788712072102e-06,
+      "loss": 0.0012,
+      "step": 13383
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442337048576985e-06,
+      "loss": 0.0017,
+      "step": 13384
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4422952259467597e-06,
+      "loss": 0.0054,
+      "step": 13385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4422534033165345e-06,
+      "loss": 0.0008,
+      "step": 13386
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4422115806863093e-06,
+      "loss": 0.0007,
+      "step": 13387
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442169758056084e-06,
+      "loss": 0.0009,
+      "step": 13388
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4421279354258589e-06,
+      "loss": 0.0008,
+      "step": 13389
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4420861127956337e-06,
+      "loss": 0.0098,
+      "step": 13390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4420442901654084e-06,
+      "loss": 0.0084,
+      "step": 13391
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4420024675351832e-06,
+      "loss": 0.2156,
+      "step": 13392
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441960644904958e-06,
+      "loss": 0.4258,
+      "step": 13393
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4419188222747328e-06,
+      "loss": 0.0053,
+      "step": 13394
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4418769996445074e-06,
+      "loss": 0.0004,
+      "step": 13395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4418351770142822e-06,
+      "loss": 0.0008,
+      "step": 13396
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4417933543840572e-06,
+      "loss": 0.0006,
+      "step": 13397
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441751531753832e-06,
+      "loss": 0.0038,
+      "step": 13398
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4417097091236067e-06,
+      "loss": 0.001,
+      "step": 13399
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4416678864933815e-06,
+      "loss": 0.002,
+      "step": 13400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4416260638631563e-06,
+      "loss": 2.1472,
+      "step": 13401
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441584241232931e-06,
+      "loss": 0.0019,
+      "step": 13402
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4415424186027059e-06,
+      "loss": 3.3961,
+      "step": 13403
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4415005959724807e-06,
+      "loss": 0.1293,
+      "step": 13404
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4414587733422555e-06,
+      "loss": 0.0016,
+      "step": 13405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4414169507120302e-06,
+      "loss": 0.0008,
+      "step": 13406
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441375128081805e-06,
+      "loss": 0.0008,
+      "step": 13407
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4413333054515798e-06,
+      "loss": 0.0022,
+      "step": 13408
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4412914828213544e-06,
+      "loss": 0.0035,
+      "step": 13409
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4412496601911292e-06,
+      "loss": 2.8963,
+      "step": 13410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4412078375609042e-06,
+      "loss": 0.004,
+      "step": 13411
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441166014930679e-06,
+      "loss": 0.0006,
+      "step": 13412
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4411241923004537e-06,
+      "loss": 0.0044,
+      "step": 13413
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4410823696702285e-06,
+      "loss": 1.4694,
+      "step": 13414
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4410405470400033e-06,
+      "loss": 0.0033,
+      "step": 13415
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440998724409778e-06,
+      "loss": 0.4427,
+      "step": 13416
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4409569017795529e-06,
+      "loss": 0.0007,
+      "step": 13417
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4409150791493277e-06,
+      "loss": 0.019,
+      "step": 13418
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4408732565191025e-06,
+      "loss": 0.0016,
+      "step": 13419
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4408314338888773e-06,
+      "loss": 1.718,
+      "step": 13420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440789611258652e-06,
+      "loss": 0.0008,
+      "step": 13421
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4407477886284268e-06,
+      "loss": 0.0002,
+      "step": 13422
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4407059659982014e-06,
+      "loss": 0.0079,
+      "step": 13423
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4406641433679762e-06,
+      "loss": 1.4261,
+      "step": 13424
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440622320737751e-06,
+      "loss": 0.4742,
+      "step": 13425
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440580498107526e-06,
+      "loss": 0.0023,
+      "step": 13426
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4405386754773008e-06,
+      "loss": 0.7664,
+      "step": 13427
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4404968528470755e-06,
+      "loss": 0.0073,
+      "step": 13428
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4404550302168503e-06,
+      "loss": 0.0028,
+      "step": 13429
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4404132075866251e-06,
+      "loss": 0.4298,
+      "step": 13430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4403713849564e-06,
+      "loss": 0.0009,
+      "step": 13431
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4403295623261747e-06,
+      "loss": 0.0386,
+      "step": 13432
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4402877396959495e-06,
+      "loss": 0.0008,
+      "step": 13433
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4402459170657243e-06,
+      "loss": 0.5852,
+      "step": 13434
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440204094435499e-06,
+      "loss": 0.0071,
+      "step": 13435
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4401622718052736e-06,
+      "loss": 0.0415,
+      "step": 13436
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4401204491750484e-06,
+      "loss": 0.0003,
+      "step": 13437
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4400786265448232e-06,
+      "loss": 1.9904,
+      "step": 13438
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440036803914598e-06,
+      "loss": 0.0007,
+      "step": 13439
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4399949812843728e-06,
+      "loss": 1.8961,
+      "step": 13440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4399531586541478e-06,
+      "loss": 0.0236,
+      "step": 13441
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4399113360239225e-06,
+      "loss": 0.001,
+      "step": 13442
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4398695133936973e-06,
+      "loss": 1.5481,
+      "step": 13443
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4398276907634721e-06,
+      "loss": 0.0003,
+      "step": 13444
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.439785868133247e-06,
+      "loss": 0.0069,
+      "step": 13445
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4397440455030217e-06,
+      "loss": 0.0028,
+      "step": 13446
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4397022228727965e-06,
+      "loss": 0.0011,
+      "step": 13447
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4396604002425713e-06,
+      "loss": 3.076,
+      "step": 13448
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.439618577612346e-06,
+      "loss": 0.0008,
+      "step": 13449
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4395767549821206e-06,
+      "loss": 0.0005,
+      "step": 13450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4395349323518954e-06,
+      "loss": 0.0003,
+      "step": 13451
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4394931097216702e-06,
+      "loss": 0.0006,
+      "step": 13452
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.439451287091445e-06,
+      "loss": 0.0012,
+      "step": 13453
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4394094644612198e-06,
+      "loss": 0.6684,
+      "step": 13454
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4393676418309948e-06,
+      "loss": 1.2555,
+      "step": 13455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4393258192007696e-06,
+      "loss": 0.0012,
+      "step": 13456
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4392839965705443e-06,
+      "loss": 0.0018,
+      "step": 13457
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4392421739403191e-06,
+      "loss": 0.0046,
+      "step": 13458
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.439200351310094e-06,
+      "loss": 0.006,
+      "step": 13459
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4391585286798687e-06,
+      "loss": 0.0025,
+      "step": 13460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4391167060496435e-06,
+      "loss": 0.0009,
+      "step": 13461
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4390748834194183e-06,
+      "loss": 0.0016,
+      "step": 13462
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4390330607891929e-06,
+      "loss": 0.006,
+      "step": 13463
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4389912381589676e-06,
+      "loss": 0.5543,
+      "step": 13464
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4389494155287424e-06,
+      "loss": 0.012,
+      "step": 13465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4389075928985172e-06,
+      "loss": 0.0012,
+      "step": 13466
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.438865770268292e-06,
+      "loss": 0.0255,
+      "step": 13467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4388239476380668e-06,
+      "loss": 0.0051,
+      "step": 13468
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4387821250078416e-06,
+      "loss": 0.0022,
+      "step": 13469
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4387403023776166e-06,
+      "loss": 0.0004,
+      "step": 13470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4386984797473914e-06,
+      "loss": 2.2078,
+      "step": 13471
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4386566571171661e-06,
+      "loss": 0.0102,
+      "step": 13472
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.438614834486941e-06,
+      "loss": 0.7213,
+      "step": 13473
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4385730118567157e-06,
+      "loss": 0.0718,
+      "step": 13474
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4385311892264905e-06,
+      "loss": 0.002,
+      "step": 13475
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4384893665962653e-06,
+      "loss": 0.0004,
+      "step": 13476
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4384475439660399e-06,
+      "loss": 1.6627,
+      "step": 13477
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4384057213358146e-06,
+      "loss": 0.0006,
+      "step": 13478
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4383638987055894e-06,
+      "loss": 1.322,
+      "step": 13479
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4383220760753642e-06,
+      "loss": 0.0014,
+      "step": 13480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.438280253445139e-06,
+      "loss": 0.0011,
+      "step": 13481
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4382384308149138e-06,
+      "loss": 0.0029,
+      "step": 13482
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4381966081846886e-06,
+      "loss": 0.0031,
+      "step": 13483
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4381547855544634e-06,
+      "loss": 0.0277,
+      "step": 13484
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4381129629242384e-06,
+      "loss": 0.6021,
+      "step": 13485
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4380711402940131e-06,
+      "loss": 0.1428,
+      "step": 13486
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.438029317663788e-06,
+      "loss": 0.0005,
+      "step": 13487
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4379874950335627e-06,
+      "loss": 0.357,
+      "step": 13488
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4379456724033375e-06,
+      "loss": 0.0451,
+      "step": 13489
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.437903849773112e-06,
+      "loss": 0.0016,
+      "step": 13490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4378620271428869e-06,
+      "loss": 0.0003,
+      "step": 13491
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4378202045126617e-06,
+      "loss": 0.0012,
+      "step": 13492
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4377783818824364e-06,
+      "loss": 1.9186,
+      "step": 13493
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4377365592522112e-06,
+      "loss": 0.4291,
+      "step": 13494
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.437694736621986e-06,
+      "loss": 0.0006,
+      "step": 13495
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4376529139917608e-06,
+      "loss": 4.3335,
+      "step": 13496
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4376110913615356e-06,
+      "loss": 0.0037,
+      "step": 13497
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4375692687313104e-06,
+      "loss": 0.0014,
+      "step": 13498
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4375274461010854e-06,
+      "loss": 0.0003,
+      "step": 13499
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4374856234708602e-06,
+      "loss": 0.0025,
+      "step": 13500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.437443800840635e-06,
+      "loss": 0.001,
+      "step": 13501
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4374019782104097e-06,
+      "loss": 0.3543,
+      "step": 13502
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4373601555801845e-06,
+      "loss": 0.0015,
+      "step": 13503
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.437318332949959e-06,
+      "loss": 0.0004,
+      "step": 13504
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4372765103197339e-06,
+      "loss": 0.6733,
+      "step": 13505
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4372346876895087e-06,
+      "loss": 0.0005,
+      "step": 13506
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4371928650592834e-06,
+      "loss": 0.3247,
+      "step": 13507
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4371510424290582e-06,
+      "loss": 0.0008,
+      "step": 13508
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.437109219798833e-06,
+      "loss": 0.0025,
+      "step": 13509
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4370673971686078e-06,
+      "loss": 0.003,
+      "step": 13510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4370255745383826e-06,
+      "loss": 0.0076,
+      "step": 13511
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4369837519081574e-06,
+      "loss": 0.0035,
+      "step": 13512
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4369419292779322e-06,
+      "loss": 0.0258,
+      "step": 13513
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4369001066477072e-06,
+      "loss": 0.0018,
+      "step": 13514
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.436858284017482e-06,
+      "loss": 1.1698,
+      "step": 13515
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4368164613872567e-06,
+      "loss": 0.0015,
+      "step": 13516
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4367746387570315e-06,
+      "loss": 0.002,
+      "step": 13517
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.436732816126806e-06,
+      "loss": 0.0008,
+      "step": 13518
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4366909934965809e-06,
+      "loss": 4.4917,
+      "step": 13519
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4366491708663557e-06,
+      "loss": 0.0006,
+      "step": 13520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4366073482361305e-06,
+      "loss": 0.8574,
+      "step": 13521
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4365655256059052e-06,
+      "loss": 0.0003,
+      "step": 13522
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.43652370297568e-06,
+      "loss": 0.0007,
+      "step": 13523
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4364818803454548e-06,
+      "loss": 0.0014,
+      "step": 13524
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4364400577152296e-06,
+      "loss": 0.3947,
+      "step": 13525
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4363982350850044e-06,
+      "loss": 0.0039,
+      "step": 13526
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4363564124547792e-06,
+      "loss": 0.018,
+      "step": 13527
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.436314589824554e-06,
+      "loss": 0.0006,
+      "step": 13528
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.436272767194329e-06,
+      "loss": 0.0453,
+      "step": 13529
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4362309445641037e-06,
+      "loss": 0.0228,
+      "step": 13530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4361891219338783e-06,
+      "loss": 0.0009,
+      "step": 13531
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4361472993036531e-06,
+      "loss": 0.0039,
+      "step": 13532
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4361054766734279e-06,
+      "loss": 0.6739,
+      "step": 13533
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4360636540432027e-06,
+      "loss": 0.0062,
+      "step": 13534
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4360218314129775e-06,
+      "loss": 0.0045,
+      "step": 13535
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4359800087827523e-06,
+      "loss": 0.0117,
+      "step": 13536
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435938186152527e-06,
+      "loss": 0.706,
+      "step": 13537
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4358963635223018e-06,
+      "loss": 0.0006,
+      "step": 13538
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4358545408920766e-06,
+      "loss": 0.0016,
+      "step": 13539
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4358127182618514e-06,
+      "loss": 0.6066,
+      "step": 13540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4357708956316262e-06,
+      "loss": 2.3417,
+      "step": 13541
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435729073001401e-06,
+      "loss": 1.1401,
+      "step": 13542
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435687250371176e-06,
+      "loss": 0.1776,
+      "step": 13543
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4356454277409508e-06,
+      "loss": 0.0012,
+      "step": 13544
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4356036051107253e-06,
+      "loss": 2.9556,
+      "step": 13545
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4355617824805001e-06,
+      "loss": 0.9396,
+      "step": 13546
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435519959850275e-06,
+      "loss": 0.0107,
+      "step": 13547
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4354781372200497e-06,
+      "loss": 0.4144,
+      "step": 13548
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4354363145898245e-06,
+      "loss": 0.0025,
+      "step": 13549
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4353944919595993e-06,
+      "loss": 0.0039,
+      "step": 13550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435352669329374e-06,
+      "loss": 1.4774,
+      "step": 13551
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4353108466991488e-06,
+      "loss": 0.0166,
+      "step": 13552
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4352690240689236e-06,
+      "loss": 0.005,
+      "step": 13553
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4352272014386984e-06,
+      "loss": 0.0013,
+      "step": 13554
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4351853788084732e-06,
+      "loss": 0.0014,
+      "step": 13555
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435143556178248e-06,
+      "loss": 1.296,
+      "step": 13556
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4351017335480228e-06,
+      "loss": 0.0037,
+      "step": 13557
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4350599109177976e-06,
+      "loss": 0.005,
+      "step": 13558
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4350180882875723e-06,
+      "loss": 2.2663,
+      "step": 13559
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4349762656573471e-06,
+      "loss": 0.0025,
+      "step": 13560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434934443027122e-06,
+      "loss": 0.001,
+      "step": 13561
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4348926203968967e-06,
+      "loss": 0.0013,
+      "step": 13562
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4348507977666715e-06,
+      "loss": 0.0079,
+      "step": 13563
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4348089751364463e-06,
+      "loss": 0.0074,
+      "step": 13564
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434767152506221e-06,
+      "loss": 0.0013,
+      "step": 13565
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4347253298759958e-06,
+      "loss": 0.1753,
+      "step": 13566
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4346835072457706e-06,
+      "loss": 2.501,
+      "step": 13567
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4346416846155454e-06,
+      "loss": 0.0012,
+      "step": 13568
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4345998619853202e-06,
+      "loss": 0.2533,
+      "step": 13569
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434558039355095e-06,
+      "loss": 0.0017,
+      "step": 13570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4345162167248696e-06,
+      "loss": 0.0004,
+      "step": 13571
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4344743940946446e-06,
+      "loss": 0.1032,
+      "step": 13572
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4344325714644193e-06,
+      "loss": 0.0024,
+      "step": 13573
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4343907488341941e-06,
+      "loss": 0.0006,
+      "step": 13574
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434348926203969e-06,
+      "loss": 0.0043,
+      "step": 13575
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4343071035737437e-06,
+      "loss": 0.0008,
+      "step": 13576
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4342652809435185e-06,
+      "loss": 0.0056,
+      "step": 13577
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4342234583132933e-06,
+      "loss": 0.0016,
+      "step": 13578
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434181635683068e-06,
+      "loss": 3.5921,
+      "step": 13579
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4341398130528429e-06,
+      "loss": 0.0005,
+      "step": 13580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4340979904226176e-06,
+      "loss": 0.0006,
+      "step": 13581
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4340561677923924e-06,
+      "loss": 0.0008,
+      "step": 13582
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4340143451621672e-06,
+      "loss": 0.0012,
+      "step": 13583
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433972522531942e-06,
+      "loss": 0.0135,
+      "step": 13584
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4339306999017166e-06,
+      "loss": 0.0027,
+      "step": 13585
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4338888772714914e-06,
+      "loss": 0.0008,
+      "step": 13586
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4338470546412664e-06,
+      "loss": 2.7354,
+      "step": 13587
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4338052320110411e-06,
+      "loss": 0.0038,
+      "step": 13588
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433763409380816e-06,
+      "loss": 0.0019,
+      "step": 13589
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4337215867505907e-06,
+      "loss": 0.0005,
+      "step": 13590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4336797641203655e-06,
+      "loss": 0.004,
+      "step": 13591
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4336379414901403e-06,
+      "loss": 0.0007,
+      "step": 13592
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433596118859915e-06,
+      "loss": 0.0006,
+      "step": 13593
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4335542962296899e-06,
+      "loss": 0.0008,
+      "step": 13594
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4335124735994646e-06,
+      "loss": 1.7421,
+      "step": 13595
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4334706509692394e-06,
+      "loss": 0.0007,
+      "step": 13596
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4334288283390142e-06,
+      "loss": 0.0003,
+      "step": 13597
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4333870057087888e-06,
+      "loss": 0.0036,
+      "step": 13598
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4333451830785636e-06,
+      "loss": 0.0109,
+      "step": 13599
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4333033604483384e-06,
+      "loss": 0.0013,
+      "step": 13600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4332615378181132e-06,
+      "loss": 0.0006,
+      "step": 13601
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4332197151878882e-06,
+      "loss": 0.0058,
+      "step": 13602
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433177892557663e-06,
+      "loss": 0.4842,
+      "step": 13603
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4331360699274377e-06,
+      "loss": 0.0016,
+      "step": 13604
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4330942472972125e-06,
+      "loss": 0.4797,
+      "step": 13605
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4330524246669873e-06,
+      "loss": 0.0011,
+      "step": 13606
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433010602036762e-06,
+      "loss": 0.0012,
+      "step": 13607
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4329687794065369e-06,
+      "loss": 0.0006,
+      "step": 13608
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4329269567763117e-06,
+      "loss": 0.0006,
+      "step": 13609
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4328851341460864e-06,
+      "loss": 0.0018,
+      "step": 13610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4328433115158612e-06,
+      "loss": 0.8653,
+      "step": 13611
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4328014888856358e-06,
+      "loss": 0.0007,
+      "step": 13612
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4327596662554106e-06,
+      "loss": 0.0016,
+      "step": 13613
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4327178436251854e-06,
+      "loss": 0.0014,
+      "step": 13614
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4326760209949602e-06,
+      "loss": 0.0095,
+      "step": 13615
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4326341983647352e-06,
+      "loss": 3.8626,
+      "step": 13616
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.43259237573451e-06,
+      "loss": 0.0126,
+      "step": 13617
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4325505531042847e-06,
+      "loss": 0.0005,
+      "step": 13618
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4325087304740595e-06,
+      "loss": 0.0006,
+      "step": 13619
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4324669078438343e-06,
+      "loss": 0.0007,
+      "step": 13620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.432425085213609e-06,
+      "loss": 0.0016,
+      "step": 13621
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4323832625833839e-06,
+      "loss": 0.0033,
+      "step": 13622
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4323414399531587e-06,
+      "loss": 0.0055,
+      "step": 13623
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4322996173229334e-06,
+      "loss": 0.0022,
+      "step": 13624
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4322577946927082e-06,
+      "loss": 2.6721,
+      "step": 13625
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4322159720624828e-06,
+      "loss": 0.0762,
+      "step": 13626
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4321741494322576e-06,
+      "loss": 0.0033,
+      "step": 13627
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4321323268020324e-06,
+      "loss": 0.0015,
+      "step": 13628
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4320905041718072e-06,
+      "loss": 0.0064,
+      "step": 13629
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.432048681541582e-06,
+      "loss": 0.2981,
+      "step": 13630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.432006858911357e-06,
+      "loss": 1.7231,
+      "step": 13631
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4319650362811317e-06,
+      "loss": 3.2226,
+      "step": 13632
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4319232136509065e-06,
+      "loss": 0.0008,
+      "step": 13633
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4318813910206813e-06,
+      "loss": 0.5065,
+      "step": 13634
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.431839568390456e-06,
+      "loss": 1.9349,
+      "step": 13635
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4317977457602309e-06,
+      "loss": 0.0006,
+      "step": 13636
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4317559231300057e-06,
+      "loss": 4.4298,
+      "step": 13637
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4317141004997805e-06,
+      "loss": 0.006,
+      "step": 13638
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.431672277869555e-06,
+      "loss": 0.0015,
+      "step": 13639
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4316304552393298e-06,
+      "loss": 0.0049,
+      "step": 13640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4315886326091046e-06,
+      "loss": 0.0007,
+      "step": 13641
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4315468099788794e-06,
+      "loss": 0.0016,
+      "step": 13642
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4315049873486542e-06,
+      "loss": 0.7594,
+      "step": 13643
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.431463164718429e-06,
+      "loss": 1.7828,
+      "step": 13644
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4314213420882038e-06,
+      "loss": 0.0007,
+      "step": 13645
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4313795194579787e-06,
+      "loss": 0.0026,
+      "step": 13646
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4313376968277535e-06,
+      "loss": 0.0004,
+      "step": 13647
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4312958741975283e-06,
+      "loss": 0.0062,
+      "step": 13648
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4312540515673031e-06,
+      "loss": 0.0023,
+      "step": 13649
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.431212228937078e-06,
+      "loss": 0.0051,
+      "step": 13650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4311704063068527e-06,
+      "loss": 0.0218,
+      "step": 13651
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4311285836766275e-06,
+      "loss": 0.0018,
+      "step": 13652
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.431086761046402e-06,
+      "loss": 0.0023,
+      "step": 13653
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4310449384161768e-06,
+      "loss": 0.0082,
+      "step": 13654
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4310031157859516e-06,
+      "loss": 0.0005,
+      "step": 13655
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4309612931557264e-06,
+      "loss": 0.0016,
+      "step": 13656
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4309194705255012e-06,
+      "loss": 0.112,
+      "step": 13657
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.430877647895276e-06,
+      "loss": 0.0006,
+      "step": 13658
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4308358252650508e-06,
+      "loss": 0.0004,
+      "step": 13659
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4307940026348258e-06,
+      "loss": 0.0038,
+      "step": 13660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4307521800046005e-06,
+      "loss": 0.0007,
+      "step": 13661
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4307103573743753e-06,
+      "loss": 1.5362,
+      "step": 13662
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4306685347441501e-06,
+      "loss": 0.0007,
+      "step": 13663
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.430626712113925e-06,
+      "loss": 0.0034,
+      "step": 13664
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4305848894836997e-06,
+      "loss": 0.0173,
+      "step": 13665
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4305430668534743e-06,
+      "loss": 0.7958,
+      "step": 13666
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.430501244223249e-06,
+      "loss": 0.0005,
+      "step": 13667
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4304594215930238e-06,
+      "loss": 0.0024,
+      "step": 13668
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4304175989627986e-06,
+      "loss": 0.0035,
+      "step": 13669
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4303757763325734e-06,
+      "loss": 0.0009,
+      "step": 13670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4303339537023482e-06,
+      "loss": 0.002,
+      "step": 13671
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.430292131072123e-06,
+      "loss": 0.0013,
+      "step": 13672
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4302503084418978e-06,
+      "loss": 0.0012,
+      "step": 13673
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4302084858116726e-06,
+      "loss": 0.0085,
+      "step": 13674
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4301666631814476e-06,
+      "loss": 0.0015,
+      "step": 13675
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4301248405512223e-06,
+      "loss": 0.0006,
+      "step": 13676
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4300830179209971e-06,
+      "loss": 0.0006,
+      "step": 13677
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.430041195290772e-06,
+      "loss": 0.0006,
+      "step": 13678
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4299993726605467e-06,
+      "loss": 0.0031,
+      "step": 13679
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4299575500303213e-06,
+      "loss": 0.387,
+      "step": 13680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429915727400096e-06,
+      "loss": 0.003,
+      "step": 13681
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4298739047698708e-06,
+      "loss": 0.69,
+      "step": 13682
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4298320821396456e-06,
+      "loss": 0.0107,
+      "step": 13683
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4297902595094204e-06,
+      "loss": 0.0012,
+      "step": 13684
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4297484368791952e-06,
+      "loss": 0.004,
+      "step": 13685
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.42970661424897e-06,
+      "loss": 0.8117,
+      "step": 13686
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4296647916187448e-06,
+      "loss": 1.8203,
+      "step": 13687
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4296229689885196e-06,
+      "loss": 1.4368,
+      "step": 13688
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4295811463582943e-06,
+      "loss": 0.0004,
+      "step": 13689
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4295393237280693e-06,
+      "loss": 0.0015,
+      "step": 13690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4294975010978441e-06,
+      "loss": 0.0007,
+      "step": 13691
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429455678467619e-06,
+      "loss": 0.0009,
+      "step": 13692
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4294138558373937e-06,
+      "loss": 0.0063,
+      "step": 13693
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4293720332071683e-06,
+      "loss": 0.0023,
+      "step": 13694
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429330210576943e-06,
+      "loss": 0.0012,
+      "step": 13695
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4292883879467179e-06,
+      "loss": 0.0006,
+      "step": 13696
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4292465653164926e-06,
+      "loss": 0.0021,
+      "step": 13697
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4292047426862674e-06,
+      "loss": 0.2517,
+      "step": 13698
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4291629200560422e-06,
+      "loss": 0.0005,
+      "step": 13699
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429121097425817e-06,
+      "loss": 0.0011,
+      "step": 13700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4290792747955918e-06,
+      "loss": 0.302,
+      "step": 13701
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4290374521653666e-06,
+      "loss": 0.441,
+      "step": 13702
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4289956295351414e-06,
+      "loss": 2.9681,
+      "step": 13703
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4289538069049164e-06,
+      "loss": 0.0035,
+      "step": 13704
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4289119842746911e-06,
+      "loss": 0.0007,
+      "step": 13705
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428870161644466e-06,
+      "loss": 0.0006,
+      "step": 13706
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4288283390142405e-06,
+      "loss": 0.0065,
+      "step": 13707
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4287865163840153e-06,
+      "loss": 0.0006,
+      "step": 13708
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.42874469375379e-06,
+      "loss": 0.0011,
+      "step": 13709
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4287028711235649e-06,
+      "loss": 0.0006,
+      "step": 13710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4286610484933396e-06,
+      "loss": 0.0049,
+      "step": 13711
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4286192258631144e-06,
+      "loss": 1.2933,
+      "step": 13712
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4285774032328892e-06,
+      "loss": 0.0009,
+      "step": 13713
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428535580602664e-06,
+      "loss": 0.0031,
+      "step": 13714
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4284937579724388e-06,
+      "loss": 0.0008,
+      "step": 13715
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4284519353422136e-06,
+      "loss": 0.0035,
+      "step": 13716
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4284101127119884e-06,
+      "loss": 0.0021,
+      "step": 13717
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4283682900817632e-06,
+      "loss": 0.0058,
+      "step": 13718
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4283264674515382e-06,
+      "loss": 0.0006,
+      "step": 13719
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428284644821313e-06,
+      "loss": 0.0004,
+      "step": 13720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4282428221910875e-06,
+      "loss": 0.0008,
+      "step": 13721
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4282009995608623e-06,
+      "loss": 1.7107,
+      "step": 13722
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428159176930637e-06,
+      "loss": 2.2346,
+      "step": 13723
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4281173543004119e-06,
+      "loss": 0.001,
+      "step": 13724
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4280755316701867e-06,
+      "loss": 0.0009,
+      "step": 13725
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4280337090399614e-06,
+      "loss": 0.0015,
+      "step": 13726
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4279918864097362e-06,
+      "loss": 0.543,
+      "step": 13727
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427950063779511e-06,
+      "loss": 0.0005,
+      "step": 13728
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4279082411492858e-06,
+      "loss": 0.7197,
+      "step": 13729
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4278664185190606e-06,
+      "loss": 0.0022,
+      "step": 13730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4278245958888354e-06,
+      "loss": 2.2182,
+      "step": 13731
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4277827732586102e-06,
+      "loss": 0.0007,
+      "step": 13732
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427740950628385e-06,
+      "loss": 0.1858,
+      "step": 13733
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4276991279981597e-06,
+      "loss": 0.0006,
+      "step": 13734
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4276573053679345e-06,
+      "loss": 0.0009,
+      "step": 13735
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4276154827377093e-06,
+      "loss": 0.1,
+      "step": 13736
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427573660107484e-06,
+      "loss": 2.4401,
+      "step": 13737
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4275318374772589e-06,
+      "loss": 0.0053,
+      "step": 13738
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4274900148470337e-06,
+      "loss": 0.0008,
+      "step": 13739
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4274481922168085e-06,
+      "loss": 0.0098,
+      "step": 13740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4274063695865832e-06,
+      "loss": 0.0568,
+      "step": 13741
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427364546956358e-06,
+      "loss": 0.0047,
+      "step": 13742
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4273227243261328e-06,
+      "loss": 0.0013,
+      "step": 13743
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4272809016959076e-06,
+      "loss": 1.2767,
+      "step": 13744
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4272390790656824e-06,
+      "loss": 0.0009,
+      "step": 13745
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4271972564354572e-06,
+      "loss": 0.0015,
+      "step": 13746
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4271554338052317e-06,
+      "loss": 0.0027,
+      "step": 13747
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4271136111750067e-06,
+      "loss": 0.0021,
+      "step": 13748
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4270717885447815e-06,
+      "loss": 0.0161,
+      "step": 13749
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4270299659145563e-06,
+      "loss": 0.2786,
+      "step": 13750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426988143284331e-06,
+      "loss": 1.6432,
+      "step": 13751
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4269463206541059e-06,
+      "loss": 0.4916,
+      "step": 13752
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4269044980238807e-06,
+      "loss": 0.0067,
+      "step": 13753
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4268626753936555e-06,
+      "loss": 0.0033,
+      "step": 13754
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4268208527634302e-06,
+      "loss": 0.002,
+      "step": 13755
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426779030133205e-06,
+      "loss": 0.0024,
+      "step": 13756
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4267372075029798e-06,
+      "loss": 0.0007,
+      "step": 13757
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4266953848727546e-06,
+      "loss": 0.0009,
+      "step": 13758
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4266535622425294e-06,
+      "loss": 0.0029,
+      "step": 13759
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4266117396123042e-06,
+      "loss": 3.2933,
+      "step": 13760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4265699169820788e-06,
+      "loss": 1.8889,
+      "step": 13761
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4265280943518535e-06,
+      "loss": 0.0041,
+      "step": 13762
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4264862717216285e-06,
+      "loss": 0.0033,
+      "step": 13763
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4264444490914033e-06,
+      "loss": 1.9709,
+      "step": 13764
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4264026264611781e-06,
+      "loss": 0.0006,
+      "step": 13765
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426360803830953e-06,
+      "loss": 0.0007,
+      "step": 13766
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4263189812007277e-06,
+      "loss": 0.0009,
+      "step": 13767
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4262771585705025e-06,
+      "loss": 3.006,
+      "step": 13768
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4262353359402773e-06,
+      "loss": 0.0007,
+      "step": 13769
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426193513310052e-06,
+      "loss": 0.0098,
+      "step": 13770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4261516906798268e-06,
+      "loss": 0.0008,
+      "step": 13771
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4261098680496016e-06,
+      "loss": 0.0007,
+      "step": 13772
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4260680454193764e-06,
+      "loss": 0.0222,
+      "step": 13773
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426026222789151e-06,
+      "loss": 0.0033,
+      "step": 13774
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4259844001589258e-06,
+      "loss": 0.0034,
+      "step": 13775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4259425775287005e-06,
+      "loss": 0.0021,
+      "step": 13776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4259007548984755e-06,
+      "loss": 0.0051,
+      "step": 13777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4258589322682503e-06,
+      "loss": 0.0035,
+      "step": 13778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4258171096380251e-06,
+      "loss": 0.0012,
+      "step": 13779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4257752870078e-06,
+      "loss": 0.0015,
+      "step": 13780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4257334643775747e-06,
+      "loss": 0.5065,
+      "step": 13781
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4256916417473495e-06,
+      "loss": 0.0012,
+      "step": 13782
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4256498191171243e-06,
+      "loss": 0.0013,
+      "step": 13783
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.425607996486899e-06,
+      "loss": 0.0007,
+      "step": 13784
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4255661738566738e-06,
+      "loss": 0.0018,
+      "step": 13785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4255243512264486e-06,
+      "loss": 0.0057,
+      "step": 13786
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4254825285962234e-06,
+      "loss": 0.0022,
+      "step": 13787
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.425440705965998e-06,
+      "loss": 1.8095,
+      "step": 13788
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4253988833357728e-06,
+      "loss": 1.7795,
+      "step": 13789
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4253570607055476e-06,
+      "loss": 0.0007,
+      "step": 13790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4253152380753223e-06,
+      "loss": 1.2845,
+      "step": 13791
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4252734154450973e-06,
+      "loss": 0.0014,
+      "step": 13792
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4252315928148721e-06,
+      "loss": 1.011,
+      "step": 13793
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.425189770184647e-06,
+      "loss": 0.0015,
+      "step": 13794
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4251479475544217e-06,
+      "loss": 0.0008,
+      "step": 13795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4251061249241965e-06,
+      "loss": 0.0006,
+      "step": 13796
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4250643022939713e-06,
+      "loss": 2.2585,
+      "step": 13797
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.425022479663746e-06,
+      "loss": 0.0018,
+      "step": 13798
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4249806570335208e-06,
+      "loss": 0.002,
+      "step": 13799
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4249388344032956e-06,
+      "loss": 0.0111,
+      "step": 13800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4248970117730702e-06,
+      "loss": 0.0042,
+      "step": 13801
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.424855189142845e-06,
+      "loss": 0.0024,
+      "step": 13802
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4248133665126198e-06,
+      "loss": 0.0069,
+      "step": 13803
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4247715438823946e-06,
+      "loss": 0.0011,
+      "step": 13804
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4247297212521694e-06,
+      "loss": 0.0011,
+      "step": 13805
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4246878986219441e-06,
+      "loss": 0.0239,
+      "step": 13806
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4246460759917191e-06,
+      "loss": 0.004,
+      "step": 13807
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.424604253361494e-06,
+      "loss": 0.0187,
+      "step": 13808
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4245624307312687e-06,
+      "loss": 1.6775,
+      "step": 13809
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4245206081010435e-06,
+      "loss": 0.0505,
+      "step": 13810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4244787854708183e-06,
+      "loss": 0.0012,
+      "step": 13811
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.424436962840593e-06,
+      "loss": 0.0037,
+      "step": 13812
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4243951402103679e-06,
+      "loss": 0.0046,
+      "step": 13813
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4243533175801426e-06,
+      "loss": 0.902,
+      "step": 13814
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4243114949499172e-06,
+      "loss": 0.0005,
+      "step": 13815
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.424269672319692e-06,
+      "loss": 0.003,
+      "step": 13816
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4242278496894668e-06,
+      "loss": 0.0026,
+      "step": 13817
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4241860270592416e-06,
+      "loss": 0.0088,
+      "step": 13818
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4241442044290164e-06,
+      "loss": 0.0014,
+      "step": 13819
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4241023817987911e-06,
+      "loss": 0.003,
+      "step": 13820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4240605591685661e-06,
+      "loss": 0.001,
+      "step": 13821
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.424018736538341e-06,
+      "loss": 0.0008,
+      "step": 13822
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4239769139081157e-06,
+      "loss": 0.0011,
+      "step": 13823
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4239350912778905e-06,
+      "loss": 0.011,
+      "step": 13824
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4238932686476653e-06,
+      "loss": 0.0009,
+      "step": 13825
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.42385144601744e-06,
+      "loss": 0.003,
+      "step": 13826
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4238096233872149e-06,
+      "loss": 0.0059,
+      "step": 13827
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4237678007569896e-06,
+      "loss": 1.3815,
+      "step": 13828
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4237259781267642e-06,
+      "loss": 0.6181,
+      "step": 13829
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.423684155496539e-06,
+      "loss": 0.0012,
+      "step": 13830
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4236423328663138e-06,
+      "loss": 0.0025,
+      "step": 13831
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4236005102360886e-06,
+      "loss": 1.4506,
+      "step": 13832
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4235586876058634e-06,
+      "loss": 0.0061,
+      "step": 13833
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4235168649756382e-06,
+      "loss": 0.0026,
+      "step": 13834
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.423475042345413e-06,
+      "loss": 2.4381,
+      "step": 13835
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.423433219715188e-06,
+      "loss": 0.0021,
+      "step": 13836
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4233913970849627e-06,
+      "loss": 0.002,
+      "step": 13837
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4233495744547375e-06,
+      "loss": 0.0022,
+      "step": 13838
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4233077518245123e-06,
+      "loss": 0.0023,
+      "step": 13839
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.423265929194287e-06,
+      "loss": 0.575,
+      "step": 13840
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4232241065640619e-06,
+      "loss": 0.0008,
+      "step": 13841
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4231822839338364e-06,
+      "loss": 3.1494,
+      "step": 13842
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4231404613036112e-06,
+      "loss": 0.0032,
+      "step": 13843
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.423098638673386e-06,
+      "loss": 0.1885,
+      "step": 13844
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4230568160431608e-06,
+      "loss": 0.0024,
+      "step": 13845
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4230149934129356e-06,
+      "loss": 0.0009,
+      "step": 13846
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4229731707827104e-06,
+      "loss": 0.0009,
+      "step": 13847
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4229313481524852e-06,
+      "loss": 0.1271,
+      "step": 13848
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.42288952552226e-06,
+      "loss": 0.0217,
+      "step": 13849
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4228477028920347e-06,
+      "loss": 0.001,
+      "step": 13850
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4228058802618097e-06,
+      "loss": 1.358,
+      "step": 13851
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4227640576315845e-06,
+      "loss": 0.0071,
+      "step": 13852
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4227222350013593e-06,
+      "loss": 2.0438,
+      "step": 13853
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422680412371134e-06,
+      "loss": 0.0019,
+      "step": 13854
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4226385897409089e-06,
+      "loss": 0.0017,
+      "step": 13855
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4225967671106835e-06,
+      "loss": 0.0015,
+      "step": 13856
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4225549444804582e-06,
+      "loss": 0.0016,
+      "step": 13857
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422513121850233e-06,
+      "loss": 0.0063,
+      "step": 13858
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4224712992200078e-06,
+      "loss": 0.0009,
+      "step": 13859
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4224294765897826e-06,
+      "loss": 0.004,
+      "step": 13860
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4223876539595574e-06,
+      "loss": 0.0037,
+      "step": 13861
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4223458313293322e-06,
+      "loss": 0.0015,
+      "step": 13862
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422304008699107e-06,
+      "loss": 0.0006,
+      "step": 13863
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4222621860688817e-06,
+      "loss": 0.0005,
+      "step": 13864
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4222203634386567e-06,
+      "loss": 0.0019,
+      "step": 13865
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4221785408084315e-06,
+      "loss": 0.0377,
+      "step": 13866
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4221367181782063e-06,
+      "loss": 0.1774,
+      "step": 13867
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422094895547981e-06,
+      "loss": 0.0004,
+      "step": 13868
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4220530729177557e-06,
+      "loss": 0.0004,
+      "step": 13869
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4220112502875305e-06,
+      "loss": 0.0406,
+      "step": 13870
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4219694276573052e-06,
+      "loss": 0.7232,
+      "step": 13871
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.42192760502708e-06,
+      "loss": 0.0006,
+      "step": 13872
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4218857823968548e-06,
+      "loss": 0.0006,
+      "step": 13873
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4218439597666296e-06,
+      "loss": 0.0061,
+      "step": 13874
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4218021371364044e-06,
+      "loss": 0.002,
+      "step": 13875
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4217603145061792e-06,
+      "loss": 0.3353,
+      "step": 13876
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421718491875954e-06,
+      "loss": 0.0029,
+      "step": 13877
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4216766692457288e-06,
+      "loss": 0.5524,
+      "step": 13878
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4216348466155035e-06,
+      "loss": 2.8152,
+      "step": 13879
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4215930239852785e-06,
+      "loss": 0.2651,
+      "step": 13880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4215512013550533e-06,
+      "loss": 2.0467,
+      "step": 13881
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4215093787248281e-06,
+      "loss": 0.0008,
+      "step": 13882
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4214675560946027e-06,
+      "loss": 0.2856,
+      "step": 13883
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4214257334643775e-06,
+      "loss": 0.001,
+      "step": 13884
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4213839108341523e-06,
+      "loss": 0.0033,
+      "step": 13885
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421342088203927e-06,
+      "loss": 0.001,
+      "step": 13886
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4213002655737018e-06,
+      "loss": 0.0048,
+      "step": 13887
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4212584429434766e-06,
+      "loss": 0.0311,
+      "step": 13888
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4212166203132514e-06,
+      "loss": 0.0132,
+      "step": 13889
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4211747976830262e-06,
+      "loss": 3.4429,
+      "step": 13890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421132975052801e-06,
+      "loss": 2.8804,
+      "step": 13891
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4210911524225758e-06,
+      "loss": 0.0017,
+      "step": 13892
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4210493297923505e-06,
+      "loss": 0.0005,
+      "step": 13893
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4210075071621253e-06,
+      "loss": 0.0017,
+      "step": 13894
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4209656845319003e-06,
+      "loss": 0.0005,
+      "step": 13895
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4209238619016751e-06,
+      "loss": 0.0022,
+      "step": 13896
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4208820392714497e-06,
+      "loss": 0.0023,
+      "step": 13897
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4208402166412245e-06,
+      "loss": 0.0471,
+      "step": 13898
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4207983940109993e-06,
+      "loss": 0.9561,
+      "step": 13899
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420756571380774e-06,
+      "loss": 0.0005,
+      "step": 13900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4207147487505488e-06,
+      "loss": 0.0028,
+      "step": 13901
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4206729261203236e-06,
+      "loss": 0.0179,
+      "step": 13902
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4206311034900984e-06,
+      "loss": 0.5751,
+      "step": 13903
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4205892808598732e-06,
+      "loss": 0.0012,
+      "step": 13904
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420547458229648e-06,
+      "loss": 0.0025,
+      "step": 13905
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4205056355994228e-06,
+      "loss": 0.0026,
+      "step": 13906
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4204638129691976e-06,
+      "loss": 0.0006,
+      "step": 13907
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4204219903389723e-06,
+      "loss": 0.0053,
+      "step": 13908
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4203801677087473e-06,
+      "loss": 0.0064,
+      "step": 13909
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420338345078522e-06,
+      "loss": 0.0045,
+      "step": 13910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4202965224482967e-06,
+      "loss": 0.1608,
+      "step": 13911
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4202546998180715e-06,
+      "loss": 0.0041,
+      "step": 13912
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4202128771878463e-06,
+      "loss": 0.0031,
+      "step": 13913
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420171054557621e-06,
+      "loss": 0.0004,
+      "step": 13914
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4201292319273958e-06,
+      "loss": 0.0034,
+      "step": 13915
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4200874092971706e-06,
+      "loss": 2.2713,
+      "step": 13916
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4200455866669454e-06,
+      "loss": 1.6538,
+      "step": 13917
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4200037640367202e-06,
+      "loss": 2.5228,
+      "step": 13918
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419961941406495e-06,
+      "loss": 0.3319,
+      "step": 13919
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4199201187762698e-06,
+      "loss": 0.0005,
+      "step": 13920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4198782961460446e-06,
+      "loss": 3.5867,
+      "step": 13921
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4198364735158194e-06,
+      "loss": 0.7185,
+      "step": 13922
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419794650885594e-06,
+      "loss": 0.0057,
+      "step": 13923
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419752828255369e-06,
+      "loss": 0.0073,
+      "step": 13924
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4197110056251437e-06,
+      "loss": 0.0012,
+      "step": 13925
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4196691829949185e-06,
+      "loss": 0.0014,
+      "step": 13926
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4196273603646933e-06,
+      "loss": 0.0006,
+      "step": 13927
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419585537734468e-06,
+      "loss": 0.001,
+      "step": 13928
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4195437151042429e-06,
+      "loss": 0.0005,
+      "step": 13929
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4195018924740176e-06,
+      "loss": 0.9701,
+      "step": 13930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4194600698437924e-06,
+      "loss": 0.0009,
+      "step": 13931
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4194182472135672e-06,
+      "loss": 2.3671,
+      "step": 13932
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419376424583342e-06,
+      "loss": 0.1852,
+      "step": 13933
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4193346019531168e-06,
+      "loss": 0.0019,
+      "step": 13934
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4192927793228916e-06,
+      "loss": 0.0009,
+      "step": 13935
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4192509566926664e-06,
+      "loss": 0.0034,
+      "step": 13936
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419209134062441e-06,
+      "loss": 2.2802,
+      "step": 13937
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4191673114322157e-06,
+      "loss": 0.0016,
+      "step": 13938
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4191254888019907e-06,
+      "loss": 0.252,
+      "step": 13939
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4190836661717655e-06,
+      "loss": 0.0024,
+      "step": 13940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4190418435415403e-06,
+      "loss": 0.0012,
+      "step": 13941
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419000020911315e-06,
+      "loss": 0.0025,
+      "step": 13942
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4189581982810899e-06,
+      "loss": 2.0103,
+      "step": 13943
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4189163756508647e-06,
+      "loss": 0.0033,
+      "step": 13944
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4188745530206394e-06,
+      "loss": 0.5484,
+      "step": 13945
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4188327303904142e-06,
+      "loss": 0.7455,
+      "step": 13946
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.418790907760189e-06,
+      "loss": 0.0008,
+      "step": 13947
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4187490851299638e-06,
+      "loss": 0.001,
+      "step": 13948
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4187072624997386e-06,
+      "loss": 0.0011,
+      "step": 13949
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4186654398695132e-06,
+      "loss": 0.2055,
+      "step": 13950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.418623617239288e-06,
+      "loss": 0.0049,
+      "step": 13951
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4185817946090627e-06,
+      "loss": 0.0133,
+      "step": 13952
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4185399719788377e-06,
+      "loss": 2.0176,
+      "step": 13953
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4184981493486125e-06,
+      "loss": 0.003,
+      "step": 13954
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4184563267183873e-06,
+      "loss": 0.306,
+      "step": 13955
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.418414504088162e-06,
+      "loss": 0.002,
+      "step": 13956
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4183726814579369e-06,
+      "loss": 0.0013,
+      "step": 13957
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4183308588277117e-06,
+      "loss": 0.0047,
+      "step": 13958
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4182890361974864e-06,
+      "loss": 0.102,
+      "step": 13959
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4182472135672612e-06,
+      "loss": 0.0009,
+      "step": 13960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.418205390937036e-06,
+      "loss": 0.0022,
+      "step": 13961
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4181635683068108e-06,
+      "loss": 0.2869,
+      "step": 13962
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4181217456765856e-06,
+      "loss": 0.5031,
+      "step": 13963
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4180799230463602e-06,
+      "loss": 0.0039,
+      "step": 13964
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.418038100416135e-06,
+      "loss": 0.001,
+      "step": 13965
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4179962777859097e-06,
+      "loss": 0.0026,
+      "step": 13966
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4179544551556845e-06,
+      "loss": 0.0012,
+      "step": 13967
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4179126325254595e-06,
+      "loss": 0.0035,
+      "step": 13968
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4178708098952343e-06,
+      "loss": 0.0008,
+      "step": 13969
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.417828987265009e-06,
+      "loss": 0.0012,
+      "step": 13970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4177871646347839e-06,
+      "loss": 0.0005,
+      "step": 13971
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4177453420045587e-06,
+      "loss": 0.0129,
+      "step": 13972
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4177035193743335e-06,
+      "loss": 0.1535,
+      "step": 13973
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4176616967441082e-06,
+      "loss": 0.0061,
+      "step": 13974
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.417619874113883e-06,
+      "loss": 0.0007,
+      "step": 13975
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4175780514836578e-06,
+      "loss": 0.0018,
+      "step": 13976
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4175362288534324e-06,
+      "loss": 0.0009,
+      "step": 13977
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4174944062232072e-06,
+      "loss": 0.003,
+      "step": 13978
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.417452583592982e-06,
+      "loss": 0.0055,
+      "step": 13979
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4174107609627567e-06,
+      "loss": 0.0009,
+      "step": 13980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4173689383325315e-06,
+      "loss": 0.0168,
+      "step": 13981
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4173271157023065e-06,
+      "loss": 0.0009,
+      "step": 13982
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4172852930720813e-06,
+      "loss": 0.0005,
+      "step": 13983
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.417243470441856e-06,
+      "loss": 0.0013,
+      "step": 13984
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4172016478116309e-06,
+      "loss": 0.001,
+      "step": 13985
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4171598251814057e-06,
+      "loss": 0.0024,
+      "step": 13986
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4171180025511805e-06,
+      "loss": 0.195,
+      "step": 13987
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4170761799209553e-06,
+      "loss": 0.0012,
+      "step": 13988
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.41703435729073e-06,
+      "loss": 0.0004,
+      "step": 13989
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4169925346605048e-06,
+      "loss": 0.1268,
+      "step": 13990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4169507120302794e-06,
+      "loss": 0.0008,
+      "step": 13991
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4169088894000542e-06,
+      "loss": 0.0009,
+      "step": 13992
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.416867066769829e-06,
+      "loss": 0.0011,
+      "step": 13993
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4168252441396038e-06,
+      "loss": 0.0008,
+      "step": 13994
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4167834215093785e-06,
+      "loss": 2.8862,
+      "step": 13995
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4167415988791533e-06,
+      "loss": 0.0063,
+      "step": 13996
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4166997762489283e-06,
+      "loss": 0.0007,
+      "step": 13997
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4166579536187031e-06,
+      "loss": 0.5251,
+      "step": 13998
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.416616130988478e-06,
+      "loss": 0.337,
+      "step": 13999
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4165743083582527e-06,
+      "loss": 0.0012,
+      "step": 14000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4165324857280275e-06,
+      "loss": 0.0016,
+      "step": 14001
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4164906630978023e-06,
+      "loss": 0.0008,
+      "step": 14002
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.416448840467577e-06,
+      "loss": 0.0014,
+      "step": 14003
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4164070178373516e-06,
+      "loss": 0.0006,
+      "step": 14004
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4163651952071264e-06,
+      "loss": 0.0006,
+      "step": 14005
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4163233725769012e-06,
+      "loss": 0.0175,
+      "step": 14006
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.416281549946676e-06,
+      "loss": 0.0007,
+      "step": 14007
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4162397273164508e-06,
+      "loss": 0.0012,
+      "step": 14008
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4161979046862256e-06,
+      "loss": 0.0008,
+      "step": 14009
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4161560820560003e-06,
+      "loss": 0.0036,
+      "step": 14010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4161142594257751e-06,
+      "loss": 0.0004,
+      "step": 14011
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4160724367955501e-06,
+      "loss": 0.0022,
+      "step": 14012
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.416030614165325e-06,
+      "loss": 0.2044,
+      "step": 14013
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4159887915350997e-06,
+      "loss": 0.0005,
+      "step": 14014
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4159469689048745e-06,
+      "loss": 1.4388,
+      "step": 14015
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4159051462746493e-06,
+      "loss": 0.0057,
+      "step": 14016
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415863323644424e-06,
+      "loss": 0.0032,
+      "step": 14017
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4158215010141986e-06,
+      "loss": 0.0024,
+      "step": 14018
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4157796783839734e-06,
+      "loss": 0.0047,
+      "step": 14019
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4157378557537482e-06,
+      "loss": 0.0007,
+      "step": 14020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415696033123523e-06,
+      "loss": 4.6753,
+      "step": 14021
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4156542104932978e-06,
+      "loss": 0.0018,
+      "step": 14022
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4156123878630726e-06,
+      "loss": 0.0004,
+      "step": 14023
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4155705652328473e-06,
+      "loss": 0.0006,
+      "step": 14024
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4155287426026221e-06,
+      "loss": 0.0048,
+      "step": 14025
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4154869199723971e-06,
+      "loss": 0.0005,
+      "step": 14026
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415445097342172e-06,
+      "loss": 0.0033,
+      "step": 14027
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4154032747119467e-06,
+      "loss": 0.0014,
+      "step": 14028
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4153614520817215e-06,
+      "loss": 0.0024,
+      "step": 14029
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4153196294514963e-06,
+      "loss": 0.0014,
+      "step": 14030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415277806821271e-06,
+      "loss": 0.6787,
+      "step": 14031
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4152359841910456e-06,
+      "loss": 0.0371,
+      "step": 14032
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4151941615608204e-06,
+      "loss": 1.0297,
+      "step": 14033
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4151523389305952e-06,
+      "loss": 1.3481,
+      "step": 14034
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.41511051630037e-06,
+      "loss": 0.0012,
+      "step": 14035
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4150686936701448e-06,
+      "loss": 0.0085,
+      "step": 14036
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4150268710399196e-06,
+      "loss": 0.0009,
+      "step": 14037
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4149850484096944e-06,
+      "loss": 0.0029,
+      "step": 14038
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4149432257794691e-06,
+      "loss": 0.0013,
+      "step": 14039
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414901403149244e-06,
+      "loss": 2.6888,
+      "step": 14040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414859580519019e-06,
+      "loss": 0.0009,
+      "step": 14041
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4148177578887937e-06,
+      "loss": 0.0004,
+      "step": 14042
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4147759352585685e-06,
+      "loss": 0.0031,
+      "step": 14043
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4147341126283433e-06,
+      "loss": 0.0006,
+      "step": 14044
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4146922899981179e-06,
+      "loss": 0.0013,
+      "step": 14045
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4146504673678926e-06,
+      "loss": 0.0014,
+      "step": 14046
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4146086447376674e-06,
+      "loss": 0.003,
+      "step": 14047
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4145668221074422e-06,
+      "loss": 1.6965,
+      "step": 14048
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414524999477217e-06,
+      "loss": 2.2509,
+      "step": 14049
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4144831768469918e-06,
+      "loss": 0.0006,
+      "step": 14050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4144413542167666e-06,
+      "loss": 0.2597,
+      "step": 14051
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4143995315865414e-06,
+      "loss": 1.1997,
+      "step": 14052
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4143577089563161e-06,
+      "loss": 0.0013,
+      "step": 14053
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414315886326091e-06,
+      "loss": 0.5296,
+      "step": 14054
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4142740636958657e-06,
+      "loss": 0.0031,
+      "step": 14055
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4142322410656407e-06,
+      "loss": 0.0033,
+      "step": 14056
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4141904184354155e-06,
+      "loss": 0.0032,
+      "step": 14057
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4141485958051903e-06,
+      "loss": 1.4955,
+      "step": 14058
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4141067731749649e-06,
+      "loss": 0.0015,
+      "step": 14059
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4140649505447397e-06,
+      "loss": 0.0021,
+      "step": 14060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4140231279145144e-06,
+      "loss": 0.0006,
+      "step": 14061
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4139813052842892e-06,
+      "loss": 2.1894,
+      "step": 14062
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413939482654064e-06,
+      "loss": 0.0019,
+      "step": 14063
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4138976600238388e-06,
+      "loss": 1.2401,
+      "step": 14064
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4138558373936136e-06,
+      "loss": 0.001,
+      "step": 14065
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4138140147633884e-06,
+      "loss": 0.0013,
+      "step": 14066
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4137721921331632e-06,
+      "loss": 0.0031,
+      "step": 14067
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413730369502938e-06,
+      "loss": 0.0121,
+      "step": 14068
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4136885468727127e-06,
+      "loss": 0.0019,
+      "step": 14069
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4136467242424877e-06,
+      "loss": 0.005,
+      "step": 14070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4136049016122625e-06,
+      "loss": 0.0039,
+      "step": 14071
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413563078982037e-06,
+      "loss": 0.0024,
+      "step": 14072
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4135212563518119e-06,
+      "loss": 0.0082,
+      "step": 14073
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4134794337215867e-06,
+      "loss": 0.4748,
+      "step": 14074
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4134376110913614e-06,
+      "loss": 0.0028,
+      "step": 14075
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4133957884611362e-06,
+      "loss": 0.0928,
+      "step": 14076
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413353965830911e-06,
+      "loss": 0.3945,
+      "step": 14077
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4133121432006858e-06,
+      "loss": 0.0022,
+      "step": 14078
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4132703205704606e-06,
+      "loss": 0.0005,
+      "step": 14079
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4132284979402354e-06,
+      "loss": 0.0025,
+      "step": 14080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4131866753100102e-06,
+      "loss": 0.0029,
+      "step": 14081
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413144852679785e-06,
+      "loss": 0.0012,
+      "step": 14082
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4131030300495597e-06,
+      "loss": 0.001,
+      "step": 14083
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4130612074193345e-06,
+      "loss": 0.0006,
+      "step": 14084
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4130193847891095e-06,
+      "loss": 0.0009,
+      "step": 14085
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.412977562158884e-06,
+      "loss": 0.7338,
+      "step": 14086
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4129357395286589e-06,
+      "loss": 0.002,
+      "step": 14087
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4128939168984337e-06,
+      "loss": 2.891,
+      "step": 14088
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4128520942682085e-06,
+      "loss": 0.0006,
+      "step": 14089
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4128102716379832e-06,
+      "loss": 0.9968,
+      "step": 14090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.412768449007758e-06,
+      "loss": 0.0009,
+      "step": 14091
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4127266263775328e-06,
+      "loss": 1.5569,
+      "step": 14092
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4126848037473076e-06,
+      "loss": 1.7643,
+      "step": 14093
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4126429811170824e-06,
+      "loss": 0.0007,
+      "step": 14094
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4126011584868572e-06,
+      "loss": 0.0007,
+      "step": 14095
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.412559335856632e-06,
+      "loss": 0.0073,
+      "step": 14096
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4125175132264067e-06,
+      "loss": 2.9325,
+      "step": 14097
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4124756905961815e-06,
+      "loss": 0.0876,
+      "step": 14098
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4124338679659561e-06,
+      "loss": 0.0014,
+      "step": 14099
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4123920453357311e-06,
+      "loss": 0.3036,
+      "step": 14100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4123502227055059e-06,
+      "loss": 0.0039,
+      "step": 14101
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4123084000752807e-06,
+      "loss": 0.4599,
+      "step": 14102
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4122665774450555e-06,
+      "loss": 0.0006,
+      "step": 14103
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4122247548148303e-06,
+      "loss": 0.367,
+      "step": 14104
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.412182932184605e-06,
+      "loss": 2.0558,
+      "step": 14105
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4121411095543798e-06,
+      "loss": 0.0008,
+      "step": 14106
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4120992869241546e-06,
+      "loss": 0.0007,
+      "step": 14107
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4120574642939294e-06,
+      "loss": 0.0076,
+      "step": 14108
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4120156416637042e-06,
+      "loss": 0.5939,
+      "step": 14109
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.411973819033479e-06,
+      "loss": 0.0014,
+      "step": 14110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4119319964032538e-06,
+      "loss": 0.0006,
+      "step": 14111
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4118901737730283e-06,
+      "loss": 0.0014,
+      "step": 14112
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4118483511428031e-06,
+      "loss": 0.0035,
+      "step": 14113
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4118065285125781e-06,
+      "loss": 0.0012,
+      "step": 14114
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.411764705882353e-06,
+      "loss": 0.0172,
+      "step": 14115
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4117228832521277e-06,
+      "loss": 0.1789,
+      "step": 14116
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4116810606219025e-06,
+      "loss": 0.0031,
+      "step": 14117
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4116392379916773e-06,
+      "loss": 2.2699,
+      "step": 14118
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.411597415361452e-06,
+      "loss": 0.0022,
+      "step": 14119
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4115555927312268e-06,
+      "loss": 0.0141,
+      "step": 14120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4115137701010016e-06,
+      "loss": 0.1373,
+      "step": 14121
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4114719474707764e-06,
+      "loss": 0.0019,
+      "step": 14122
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4114301248405512e-06,
+      "loss": 0.0052,
+      "step": 14123
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.411388302210326e-06,
+      "loss": 0.0012,
+      "step": 14124
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4113464795801008e-06,
+      "loss": 0.0004,
+      "step": 14125
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4113046569498753e-06,
+      "loss": 0.0004,
+      "step": 14126
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4112628343196501e-06,
+      "loss": 0.0015,
+      "step": 14127
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.411221011689425e-06,
+      "loss": 0.0026,
+      "step": 14128
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4111791890592e-06,
+      "loss": 0.0018,
+      "step": 14129
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4111373664289747e-06,
+      "loss": 0.0014,
+      "step": 14130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4110955437987495e-06,
+      "loss": 0.0032,
+      "step": 14131
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4110537211685243e-06,
+      "loss": 0.0065,
+      "step": 14132
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.411011898538299e-06,
+      "loss": 0.0009,
+      "step": 14133
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4109700759080738e-06,
+      "loss": 0.0019,
+      "step": 14134
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4109282532778486e-06,
+      "loss": 0.5541,
+      "step": 14135
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4108864306476234e-06,
+      "loss": 0.0031,
+      "step": 14136
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4108446080173982e-06,
+      "loss": 0.0511,
+      "step": 14137
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.410802785387173e-06,
+      "loss": 2.5668,
+      "step": 14138
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4107609627569478e-06,
+      "loss": 0.0009,
+      "step": 14139
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4107191401267223e-06,
+      "loss": 0.0026,
+      "step": 14140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4106773174964971e-06,
+      "loss": 0.0017,
+      "step": 14141
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.410635494866272e-06,
+      "loss": 0.0026,
+      "step": 14142
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.410593672236047e-06,
+      "loss": 0.0015,
+      "step": 14143
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4105518496058217e-06,
+      "loss": 0.0063,
+      "step": 14144
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4105100269755965e-06,
+      "loss": 0.0025,
+      "step": 14145
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4104682043453713e-06,
+      "loss": 0.0071,
+      "step": 14146
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.410426381715146e-06,
+      "loss": 0.0034,
+      "step": 14147
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4103845590849209e-06,
+      "loss": 0.0032,
+      "step": 14148
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4103427364546956e-06,
+      "loss": 2.3133,
+      "step": 14149
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4103009138244704e-06,
+      "loss": 0.0006,
+      "step": 14150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4102590911942452e-06,
+      "loss": 1.816,
+      "step": 14151
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.41021726856402e-06,
+      "loss": 1.8275,
+      "step": 14152
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4101754459337946e-06,
+      "loss": 0.0017,
+      "step": 14153
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4101336233035694e-06,
+      "loss": 0.0007,
+      "step": 14154
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4100918006733441e-06,
+      "loss": 0.0003,
+      "step": 14155
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.410049978043119e-06,
+      "loss": 0.0008,
+      "step": 14156
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4100081554128937e-06,
+      "loss": 0.0009,
+      "step": 14157
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4099663327826687e-06,
+      "loss": 0.5204,
+      "step": 14158
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4099245101524435e-06,
+      "loss": 0.0007,
+      "step": 14159
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4098826875222183e-06,
+      "loss": 0.0054,
+      "step": 14160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409840864891993e-06,
+      "loss": 0.0006,
+      "step": 14161
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4097990422617679e-06,
+      "loss": 0.0004,
+      "step": 14162
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4097572196315426e-06,
+      "loss": 0.355,
+      "step": 14163
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4097153970013174e-06,
+      "loss": 0.0007,
+      "step": 14164
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4096735743710922e-06,
+      "loss": 0.0006,
+      "step": 14165
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409631751740867e-06,
+      "loss": 0.0003,
+      "step": 14166
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4095899291106416e-06,
+      "loss": 0.0123,
+      "step": 14167
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4095481064804164e-06,
+      "loss": 0.0005,
+      "step": 14168
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4095062838501912e-06,
+      "loss": 0.0048,
+      "step": 14169
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409464461219966e-06,
+      "loss": 0.001,
+      "step": 14170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4094226385897407e-06,
+      "loss": 0.0006,
+      "step": 14171
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4093808159595155e-06,
+      "loss": 0.0004,
+      "step": 14172
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4093389933292905e-06,
+      "loss": 0.011,
+      "step": 14173
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4092971706990653e-06,
+      "loss": 0.0077,
+      "step": 14174
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.40925534806884e-06,
+      "loss": 0.0024,
+      "step": 14175
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4092135254386149e-06,
+      "loss": 0.0024,
+      "step": 14176
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4091717028083897e-06,
+      "loss": 0.0006,
+      "step": 14177
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4091298801781644e-06,
+      "loss": 0.0018,
+      "step": 14178
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4090880575479392e-06,
+      "loss": 0.002,
+      "step": 14179
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4090462349177138e-06,
+      "loss": 1.3998,
+      "step": 14180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4090044122874886e-06,
+      "loss": 0.014,
+      "step": 14181
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4089625896572634e-06,
+      "loss": 1.7503,
+      "step": 14182
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4089207670270382e-06,
+      "loss": 0.0009,
+      "step": 14183
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.408878944396813e-06,
+      "loss": 2.9126,
+      "step": 14184
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4088371217665877e-06,
+      "loss": 0.0038,
+      "step": 14185
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4087952991363625e-06,
+      "loss": 0.0003,
+      "step": 14186
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4087534765061375e-06,
+      "loss": 0.0002,
+      "step": 14187
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4087116538759123e-06,
+      "loss": 1.543,
+      "step": 14188
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.408669831245687e-06,
+      "loss": 0.3235,
+      "step": 14189
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4086280086154619e-06,
+      "loss": 1.4421,
+      "step": 14190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4085861859852367e-06,
+      "loss": 0.0004,
+      "step": 14191
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4085443633550114e-06,
+      "loss": 0.0005,
+      "step": 14192
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4085025407247862e-06,
+      "loss": 0.0012,
+      "step": 14193
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4084607180945608e-06,
+      "loss": 0.0021,
+      "step": 14194
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4084188954643356e-06,
+      "loss": 1.0189,
+      "step": 14195
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4083770728341104e-06,
+      "loss": 0.0033,
+      "step": 14196
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4083352502038852e-06,
+      "loss": 0.0009,
+      "step": 14197
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.40829342757366e-06,
+      "loss": 0.0015,
+      "step": 14198
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4082516049434347e-06,
+      "loss": 0.587,
+      "step": 14199
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4082097823132095e-06,
+      "loss": 0.0026,
+      "step": 14200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4081679596829843e-06,
+      "loss": 0.0023,
+      "step": 14201
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4081261370527593e-06,
+      "loss": 0.0022,
+      "step": 14202
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.408084314422534e-06,
+      "loss": 0.0005,
+      "step": 14203
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4080424917923089e-06,
+      "loss": 0.0003,
+      "step": 14204
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4080006691620837e-06,
+      "loss": 0.0933,
+      "step": 14205
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4079588465318585e-06,
+      "loss": 0.0005,
+      "step": 14206
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.407917023901633e-06,
+      "loss": 0.0004,
+      "step": 14207
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4078752012714078e-06,
+      "loss": 0.0003,
+      "step": 14208
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4078333786411826e-06,
+      "loss": 0.0004,
+      "step": 14209
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4077915560109574e-06,
+      "loss": 0.0307,
+      "step": 14210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4077497333807322e-06,
+      "loss": 0.0105,
+      "step": 14211
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.407707910750507e-06,
+      "loss": 0.0005,
+      "step": 14212
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4076660881202818e-06,
+      "loss": 0.1384,
+      "step": 14213
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4076242654900565e-06,
+      "loss": 0.2819,
+      "step": 14214
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4075824428598313e-06,
+      "loss": 0.0009,
+      "step": 14215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4075406202296061e-06,
+      "loss": 0.0009,
+      "step": 14216
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4074987975993811e-06,
+      "loss": 0.0011,
+      "step": 14217
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.407456974969156e-06,
+      "loss": 0.002,
+      "step": 14218
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4074151523389307e-06,
+      "loss": 3.2829,
+      "step": 14219
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4073733297087055e-06,
+      "loss": 0.7276,
+      "step": 14220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.40733150707848e-06,
+      "loss": 0.0021,
+      "step": 14221
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4072896844482548e-06,
+      "loss": 0.0029,
+      "step": 14222
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4072478618180296e-06,
+      "loss": 0.0022,
+      "step": 14223
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4072060391878044e-06,
+      "loss": 0.0007,
+      "step": 14224
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4071642165575792e-06,
+      "loss": 0.0023,
+      "step": 14225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.407122393927354e-06,
+      "loss": 0.0005,
+      "step": 14226
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4070805712971288e-06,
+      "loss": 0.0004,
+      "step": 14227
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4070387486669035e-06,
+      "loss": 0.0048,
+      "step": 14228
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4069969260366783e-06,
+      "loss": 2.7879,
+      "step": 14229
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4069551034064531e-06,
+      "loss": 0.0005,
+      "step": 14230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4069132807762281e-06,
+      "loss": 0.0152,
+      "step": 14231
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406871458146003e-06,
+      "loss": 0.001,
+      "step": 14232
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4068296355157777e-06,
+      "loss": 0.0005,
+      "step": 14233
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4067878128855525e-06,
+      "loss": 1.5865,
+      "step": 14234
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406745990255327e-06,
+      "loss": 0.0006,
+      "step": 14235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4067041676251018e-06,
+      "loss": 3.2805,
+      "step": 14236
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4066623449948766e-06,
+      "loss": 0.0016,
+      "step": 14237
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4066205223646514e-06,
+      "loss": 0.001,
+      "step": 14238
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4065786997344262e-06,
+      "loss": 0.0004,
+      "step": 14239
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406536877104201e-06,
+      "loss": 0.0032,
+      "step": 14240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4064950544739758e-06,
+      "loss": 0.6645,
+      "step": 14241
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4064532318437506e-06,
+      "loss": 0.0017,
+      "step": 14242
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4064114092135253e-06,
+      "loss": 1.015,
+      "step": 14243
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4063695865833001e-06,
+      "loss": 0.0033,
+      "step": 14244
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406327763953075e-06,
+      "loss": 0.0071,
+      "step": 14245
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.40628594132285e-06,
+      "loss": 0.004,
+      "step": 14246
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4062441186926247e-06,
+      "loss": 0.0011,
+      "step": 14247
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4062022960623993e-06,
+      "loss": 2.114,
+      "step": 14248
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406160473432174e-06,
+      "loss": 0.0005,
+      "step": 14249
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4061186508019488e-06,
+      "loss": 1.0189,
+      "step": 14250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4060768281717236e-06,
+      "loss": 0.0011,
+      "step": 14251
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4060350055414984e-06,
+      "loss": 0.0011,
+      "step": 14252
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4059931829112732e-06,
+      "loss": 0.0008,
+      "step": 14253
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405951360281048e-06,
+      "loss": 0.0059,
+      "step": 14254
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4059095376508228e-06,
+      "loss": 0.0005,
+      "step": 14255
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4058677150205976e-06,
+      "loss": 0.0045,
+      "step": 14256
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4058258923903723e-06,
+      "loss": 0.65,
+      "step": 14257
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4057840697601471e-06,
+      "loss": 0.0033,
+      "step": 14258
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405742247129922e-06,
+      "loss": 0.0011,
+      "step": 14259
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4057004244996967e-06,
+      "loss": 0.0008,
+      "step": 14260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4056586018694717e-06,
+      "loss": 1.4817,
+      "step": 14261
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4056167792392463e-06,
+      "loss": 0.0005,
+      "step": 14262
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405574956609021e-06,
+      "loss": 2.9787,
+      "step": 14263
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4055331339787959e-06,
+      "loss": 0.1299,
+      "step": 14264
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4054913113485706e-06,
+      "loss": 0.003,
+      "step": 14265
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4054494887183454e-06,
+      "loss": 0.0028,
+      "step": 14266
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4054076660881202e-06,
+      "loss": 2.3925,
+      "step": 14267
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405365843457895e-06,
+      "loss": 0.004,
+      "step": 14268
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4053240208276698e-06,
+      "loss": 0.0009,
+      "step": 14269
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4052821981974446e-06,
+      "loss": 0.001,
+      "step": 14270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4052403755672194e-06,
+      "loss": 0.0631,
+      "step": 14271
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4051985529369941e-06,
+      "loss": 0.0027,
+      "step": 14272
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405156730306769e-06,
+      "loss": 0.0022,
+      "step": 14273
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4051149076765437e-06,
+      "loss": 2.0081,
+      "step": 14274
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4050730850463185e-06,
+      "loss": 0.0007,
+      "step": 14275
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4050312624160933e-06,
+      "loss": 0.0081,
+      "step": 14276
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.404989439785868e-06,
+      "loss": 0.287,
+      "step": 14277
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4049476171556429e-06,
+      "loss": 0.0043,
+      "step": 14278
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4049057945254176e-06,
+      "loss": 0.0008,
+      "step": 14279
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4048639718951924e-06,
+      "loss": 0.0036,
+      "step": 14280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4048221492649672e-06,
+      "loss": 0.0006,
+      "step": 14281
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.404780326634742e-06,
+      "loss": 0.0008,
+      "step": 14282
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4047385040045168e-06,
+      "loss": 0.0062,
+      "step": 14283
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4046966813742916e-06,
+      "loss": 3.6848,
+      "step": 14284
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4046548587440664e-06,
+      "loss": 0.67,
+      "step": 14285
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4046130361138412e-06,
+      "loss": 0.0017,
+      "step": 14286
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.404571213483616e-06,
+      "loss": 0.0253,
+      "step": 14287
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4045293908533905e-06,
+      "loss": 0.1512,
+      "step": 14288
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4044875682231653e-06,
+      "loss": 0.4675,
+      "step": 14289
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4044457455929403e-06,
+      "loss": 0.0005,
+      "step": 14290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.404403922962715e-06,
+      "loss": 0.0024,
+      "step": 14291
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4043621003324899e-06,
+      "loss": 0.0018,
+      "step": 14292
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4043202777022647e-06,
+      "loss": 0.0033,
+      "step": 14293
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4042784550720394e-06,
+      "loss": 0.0011,
+      "step": 14294
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4042366324418142e-06,
+      "loss": 0.0008,
+      "step": 14295
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.404194809811589e-06,
+      "loss": 0.0794,
+      "step": 14296
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4041529871813638e-06,
+      "loss": 0.0052,
+      "step": 14297
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4041111645511386e-06,
+      "loss": 1.1534,
+      "step": 14298
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4040693419209134e-06,
+      "loss": 0.0019,
+      "step": 14299
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4040275192906882e-06,
+      "loss": 0.0009,
+      "step": 14300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.403985696660463e-06,
+      "loss": 0.0032,
+      "step": 14301
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4039438740302375e-06,
+      "loss": 1.7092,
+      "step": 14302
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4039020514000123e-06,
+      "loss": 0.0004,
+      "step": 14303
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.403860228769787e-06,
+      "loss": 0.0042,
+      "step": 14304
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.403818406139562e-06,
+      "loss": 0.0052,
+      "step": 14305
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4037765835093369e-06,
+      "loss": 0.0014,
+      "step": 14306
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4037347608791117e-06,
+      "loss": 0.0035,
+      "step": 14307
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4036929382488865e-06,
+      "loss": 0.001,
+      "step": 14308
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4036511156186612e-06,
+      "loss": 1.7709,
+      "step": 14309
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.403609292988436e-06,
+      "loss": 0.0008,
+      "step": 14310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4035674703582108e-06,
+      "loss": 0.0009,
+      "step": 14311
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4035256477279856e-06,
+      "loss": 0.0014,
+      "step": 14312
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4034838250977604e-06,
+      "loss": 0.0029,
+      "step": 14313
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4034420024675352e-06,
+      "loss": 0.0009,
+      "step": 14314
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4034001798373097e-06,
+      "loss": 0.0017,
+      "step": 14315
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4033583572070845e-06,
+      "loss": 0.1073,
+      "step": 14316
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4033165345768593e-06,
+      "loss": 0.0004,
+      "step": 14317
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.403274711946634e-06,
+      "loss": 0.0382,
+      "step": 14318
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.403232889316409e-06,
+      "loss": 3.1491,
+      "step": 14319
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4031910666861839e-06,
+      "loss": 0.0005,
+      "step": 14320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4031492440559587e-06,
+      "loss": 0.0023,
+      "step": 14321
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4031074214257335e-06,
+      "loss": 0.0018,
+      "step": 14322
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4030655987955082e-06,
+      "loss": 0.0009,
+      "step": 14323
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.403023776165283e-06,
+      "loss": 0.0075,
+      "step": 14324
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4029819535350578e-06,
+      "loss": 0.0004,
+      "step": 14325
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4029401309048326e-06,
+      "loss": 0.8961,
+      "step": 14326
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4028983082746074e-06,
+      "loss": 0.0009,
+      "step": 14327
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4028564856443822e-06,
+      "loss": 0.0658,
+      "step": 14328
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4028146630141568e-06,
+      "loss": 0.002,
+      "step": 14329
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4027728403839315e-06,
+      "loss": 0.0039,
+      "step": 14330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4027310177537063e-06,
+      "loss": 0.0011,
+      "step": 14331
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4026891951234811e-06,
+      "loss": 1.2404,
+      "step": 14332
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402647372493256e-06,
+      "loss": 1.0462,
+      "step": 14333
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402605549863031e-06,
+      "loss": 0.0016,
+      "step": 14334
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4025637272328057e-06,
+      "loss": 1.4337,
+      "step": 14335
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4025219046025805e-06,
+      "loss": 0.0019,
+      "step": 14336
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4024800819723553e-06,
+      "loss": 0.002,
+      "step": 14337
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.40243825934213e-06,
+      "loss": 0.0006,
+      "step": 14338
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4023964367119048e-06,
+      "loss": 0.0017,
+      "step": 14339
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4023546140816796e-06,
+      "loss": 0.003,
+      "step": 14340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4023127914514544e-06,
+      "loss": 0.003,
+      "step": 14341
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4022709688212292e-06,
+      "loss": 0.0008,
+      "step": 14342
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4022291461910038e-06,
+      "loss": 0.4435,
+      "step": 14343
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4021873235607785e-06,
+      "loss": 2.7828,
+      "step": 14344
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4021455009305533e-06,
+      "loss": 1.7664,
+      "step": 14345
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4021036783003281e-06,
+      "loss": 0.0009,
+      "step": 14346
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402061855670103e-06,
+      "loss": 0.0007,
+      "step": 14347
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402020033039878e-06,
+      "loss": 0.0004,
+      "step": 14348
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4019782104096527e-06,
+      "loss": 0.0191,
+      "step": 14349
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4019363877794275e-06,
+      "loss": 0.0159,
+      "step": 14350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4018945651492023e-06,
+      "loss": 0.0008,
+      "step": 14351
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.401852742518977e-06,
+      "loss": 0.0009,
+      "step": 14352
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4018109198887518e-06,
+      "loss": 0.0356,
+      "step": 14353
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4017690972585266e-06,
+      "loss": 3.9197,
+      "step": 14354
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4017272746283014e-06,
+      "loss": 0.0035,
+      "step": 14355
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.401685451998076e-06,
+      "loss": 0.0094,
+      "step": 14356
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4016436293678508e-06,
+      "loss": 0.0015,
+      "step": 14357
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4016018067376256e-06,
+      "loss": 0.003,
+      "step": 14358
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4015599841074003e-06,
+      "loss": 0.0022,
+      "step": 14359
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4015181614771751e-06,
+      "loss": 2.5237,
+      "step": 14360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.40147633884695e-06,
+      "loss": 2.278,
+      "step": 14361
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4014345162167247e-06,
+      "loss": 0.0028,
+      "step": 14362
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4013926935864997e-06,
+      "loss": 0.0014,
+      "step": 14363
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4013508709562745e-06,
+      "loss": 0.0948,
+      "step": 14364
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4013090483260493e-06,
+      "loss": 0.0018,
+      "step": 14365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.401267225695824e-06,
+      "loss": 2.5245,
+      "step": 14366
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4012254030655988e-06,
+      "loss": 0.8581,
+      "step": 14367
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4011835804353736e-06,
+      "loss": 0.0023,
+      "step": 14368
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4011417578051484e-06,
+      "loss": 0.003,
+      "step": 14369
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.401099935174923e-06,
+      "loss": 1.6588,
+      "step": 14370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4010581125446978e-06,
+      "loss": 0.0053,
+      "step": 14371
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4010162899144726e-06,
+      "loss": 0.001,
+      "step": 14372
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4009744672842474e-06,
+      "loss": 0.0015,
+      "step": 14373
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4009326446540221e-06,
+      "loss": 0.0006,
+      "step": 14374
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400890822023797e-06,
+      "loss": 0.0005,
+      "step": 14375
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4008489993935717e-06,
+      "loss": 0.0006,
+      "step": 14376
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4008071767633465e-06,
+      "loss": 0.0013,
+      "step": 14377
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4007653541331215e-06,
+      "loss": 0.0385,
+      "step": 14378
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4007235315028963e-06,
+      "loss": 0.0017,
+      "step": 14379
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400681708872671e-06,
+      "loss": 0.0015,
+      "step": 14380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4006398862424459e-06,
+      "loss": 1.632,
+      "step": 14381
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4005980636122206e-06,
+      "loss": 0.0008,
+      "step": 14382
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4005562409819952e-06,
+      "loss": 0.0017,
+      "step": 14383
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.40051441835177e-06,
+      "loss": 0.0025,
+      "step": 14384
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4004725957215448e-06,
+      "loss": 0.005,
+      "step": 14385
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4004307730913196e-06,
+      "loss": 3.5,
+      "step": 14386
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4003889504610944e-06,
+      "loss": 0.0038,
+      "step": 14387
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4003471278308691e-06,
+      "loss": 0.0042,
+      "step": 14388
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400305305200644e-06,
+      "loss": 0.0025,
+      "step": 14389
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4002634825704187e-06,
+      "loss": 0.0012,
+      "step": 14390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4002216599401935e-06,
+      "loss": 0.4204,
+      "step": 14391
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4001798373099685e-06,
+      "loss": 0.0008,
+      "step": 14392
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4001380146797433e-06,
+      "loss": 0.0012,
+      "step": 14393
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400096192049518e-06,
+      "loss": 1.8976,
+      "step": 14394
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4000543694192929e-06,
+      "loss": 0.0061,
+      "step": 14395
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4000125467890676e-06,
+      "loss": 0.0052,
+      "step": 14396
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3999707241588422e-06,
+      "loss": 2.744,
+      "step": 14397
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399928901528617e-06,
+      "loss": 0.0055,
+      "step": 14398
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3998870788983918e-06,
+      "loss": 0.0023,
+      "step": 14399
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3998452562681666e-06,
+      "loss": 0.0046,
+      "step": 14400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3998034336379414e-06,
+      "loss": 0.0065,
+      "step": 14401
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3997616110077162e-06,
+      "loss": 0.0007,
+      "step": 14402
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399719788377491e-06,
+      "loss": 0.0004,
+      "step": 14403
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3996779657472657e-06,
+      "loss": 0.0006,
+      "step": 14404
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3996361431170405e-06,
+      "loss": 0.0009,
+      "step": 14405
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3995943204868153e-06,
+      "loss": 0.0015,
+      "step": 14406
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3995524978565903e-06,
+      "loss": 0.006,
+      "step": 14407
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399510675226365e-06,
+      "loss": 0.0188,
+      "step": 14408
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3994688525961399e-06,
+      "loss": 0.0013,
+      "step": 14409
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3994270299659147e-06,
+      "loss": 0.0047,
+      "step": 14410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3993852073356892e-06,
+      "loss": 0.0023,
+      "step": 14411
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399343384705464e-06,
+      "loss": 0.0018,
+      "step": 14412
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3993015620752388e-06,
+      "loss": 0.0046,
+      "step": 14413
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3992597394450136e-06,
+      "loss": 0.0011,
+      "step": 14414
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3992179168147884e-06,
+      "loss": 0.0008,
+      "step": 14415
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3991760941845632e-06,
+      "loss": 0.0041,
+      "step": 14416
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399134271554338e-06,
+      "loss": 2.4089,
+      "step": 14417
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3990924489241127e-06,
+      "loss": 0.0004,
+      "step": 14418
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3990506262938875e-06,
+      "loss": 2.1539,
+      "step": 14419
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3990088036636623e-06,
+      "loss": 0.001,
+      "step": 14420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398966981033437e-06,
+      "loss": 0.1025,
+      "step": 14421
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398925158403212e-06,
+      "loss": 0.0049,
+      "step": 14422
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3988833357729869e-06,
+      "loss": 0.0026,
+      "step": 14423
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3988415131427615e-06,
+      "loss": 0.0007,
+      "step": 14424
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3987996905125362e-06,
+      "loss": 0.0113,
+      "step": 14425
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398757867882311e-06,
+      "loss": 0.2207,
+      "step": 14426
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3987160452520858e-06,
+      "loss": 0.0686,
+      "step": 14427
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3986742226218606e-06,
+      "loss": 0.0014,
+      "step": 14428
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3986323999916354e-06,
+      "loss": 0.0031,
+      "step": 14429
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3985905773614102e-06,
+      "loss": 0.003,
+      "step": 14430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398548754731185e-06,
+      "loss": 0.0019,
+      "step": 14431
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3985069321009597e-06,
+      "loss": 0.0008,
+      "step": 14432
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3984651094707345e-06,
+      "loss": 0.0035,
+      "step": 14433
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3984232868405093e-06,
+      "loss": 1.4529,
+      "step": 14434
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398381464210284e-06,
+      "loss": 0.0067,
+      "step": 14435
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398339641580059e-06,
+      "loss": 0.001,
+      "step": 14436
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3982978189498339e-06,
+      "loss": 0.0011,
+      "step": 14437
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3982559963196085e-06,
+      "loss": 0.395,
+      "step": 14438
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3982141736893832e-06,
+      "loss": 0.0008,
+      "step": 14439
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398172351059158e-06,
+      "loss": 0.0014,
+      "step": 14440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3981305284289328e-06,
+      "loss": 0.0029,
+      "step": 14441
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3980887057987076e-06,
+      "loss": 0.0007,
+      "step": 14442
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3980468831684824e-06,
+      "loss": 0.0005,
+      "step": 14443
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3980050605382572e-06,
+      "loss": 0.0006,
+      "step": 14444
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.397963237908032e-06,
+      "loss": 0.0043,
+      "step": 14445
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3979214152778068e-06,
+      "loss": 0.0118,
+      "step": 14446
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3978795926475815e-06,
+      "loss": 0.3829,
+      "step": 14447
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3978377700173563e-06,
+      "loss": 0.0034,
+      "step": 14448
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3977959473871311e-06,
+      "loss": 0.0013,
+      "step": 14449
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.397754124756906e-06,
+      "loss": 0.6082,
+      "step": 14450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3977123021266807e-06,
+      "loss": 0.0005,
+      "step": 14451
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3976704794964555e-06,
+      "loss": 0.0005,
+      "step": 14452
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3976286568662303e-06,
+      "loss": 0.0033,
+      "step": 14453
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.397586834236005e-06,
+      "loss": 0.0012,
+      "step": 14454
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3975450116057798e-06,
+      "loss": 0.0005,
+      "step": 14455
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3975031889755546e-06,
+      "loss": 0.0004,
+      "step": 14456
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3974613663453294e-06,
+      "loss": 0.0011,
+      "step": 14457
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3974195437151042e-06,
+      "loss": 0.0005,
+      "step": 14458
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.397377721084879e-06,
+      "loss": 2.5131,
+      "step": 14459
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3973358984546538e-06,
+      "loss": 0.0016,
+      "step": 14460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3972940758244285e-06,
+      "loss": 0.0006,
+      "step": 14461
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3972522531942033e-06,
+      "loss": 0.8248,
+      "step": 14462
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3972104305639781e-06,
+      "loss": 0.0198,
+      "step": 14463
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3971686079337527e-06,
+      "loss": 0.0004,
+      "step": 14464
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3971267853035275e-06,
+      "loss": 1.9833,
+      "step": 14465
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3970849626733025e-06,
+      "loss": 0.003,
+      "step": 14466
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3970431400430773e-06,
+      "loss": 0.001,
+      "step": 14467
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.397001317412852e-06,
+      "loss": 0.0044,
+      "step": 14468
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3969594947826268e-06,
+      "loss": 0.0011,
+      "step": 14469
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3969176721524016e-06,
+      "loss": 1.517,
+      "step": 14470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3968758495221764e-06,
+      "loss": 0.0044,
+      "step": 14471
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3968340268919512e-06,
+      "loss": 0.0089,
+      "step": 14472
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.396792204261726e-06,
+      "loss": 0.0017,
+      "step": 14473
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3967503816315008e-06,
+      "loss": 0.0006,
+      "step": 14474
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3967085590012756e-06,
+      "loss": 3.4032,
+      "step": 14475
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3966667363710503e-06,
+      "loss": 0.0007,
+      "step": 14476
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3966249137408251e-06,
+      "loss": 0.0149,
+      "step": 14477
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3965830911105997e-06,
+      "loss": 0.0036,
+      "step": 14478
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3965412684803745e-06,
+      "loss": 0.5225,
+      "step": 14479
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3964994458501495e-06,
+      "loss": 0.0007,
+      "step": 14480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3964576232199243e-06,
+      "loss": 0.0003,
+      "step": 14481
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.396415800589699e-06,
+      "loss": 0.0045,
+      "step": 14482
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3963739779594738e-06,
+      "loss": 0.0033,
+      "step": 14483
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3963321553292486e-06,
+      "loss": 2.1333,
+      "step": 14484
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3962903326990234e-06,
+      "loss": 0.0022,
+      "step": 14485
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3962485100687982e-06,
+      "loss": 0.0094,
+      "step": 14486
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.396206687438573e-06,
+      "loss": 0.0015,
+      "step": 14487
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3961648648083478e-06,
+      "loss": 0.0011,
+      "step": 14488
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3961230421781226e-06,
+      "loss": 0.0069,
+      "step": 14489
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3960812195478974e-06,
+      "loss": 2.018,
+      "step": 14490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.396039396917672e-06,
+      "loss": 0.0016,
+      "step": 14491
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3959975742874467e-06,
+      "loss": 0.0014,
+      "step": 14492
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3959557516572215e-06,
+      "loss": 0.0016,
+      "step": 14493
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3959139290269963e-06,
+      "loss": 0.0951,
+      "step": 14494
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3958721063967713e-06,
+      "loss": 0.0044,
+      "step": 14495
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395830283766546e-06,
+      "loss": 1.5821,
+      "step": 14496
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3957884611363209e-06,
+      "loss": 0.001,
+      "step": 14497
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3957466385060956e-06,
+      "loss": 2.2163,
+      "step": 14498
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3957048158758704e-06,
+      "loss": 0.0179,
+      "step": 14499
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3956629932456452e-06,
+      "loss": 0.006,
+      "step": 14500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.39562117061542e-06,
+      "loss": 0.0008,
+      "step": 14501
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3955793479851948e-06,
+      "loss": 0.953,
+      "step": 14502
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3955375253549696e-06,
+      "loss": 0.0061,
+      "step": 14503
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3954957027247444e-06,
+      "loss": 1.0587,
+      "step": 14504
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395453880094519e-06,
+      "loss": 0.0011,
+      "step": 14505
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3954120574642937e-06,
+      "loss": 0.001,
+      "step": 14506
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3953702348340685e-06,
+      "loss": 0.0042,
+      "step": 14507
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3953284122038433e-06,
+      "loss": 0.0005,
+      "step": 14508
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395286589573618e-06,
+      "loss": 0.0021,
+      "step": 14509
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395244766943393e-06,
+      "loss": 0.0017,
+      "step": 14510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3952029443131679e-06,
+      "loss": 0.0004,
+      "step": 14511
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3951611216829427e-06,
+      "loss": 0.0006,
+      "step": 14512
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3951192990527174e-06,
+      "loss": 0.0006,
+      "step": 14513
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3950774764224922e-06,
+      "loss": 0.0006,
+      "step": 14514
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395035653792267e-06,
+      "loss": 0.0007,
+      "step": 14515
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3949938311620418e-06,
+      "loss": 0.001,
+      "step": 14516
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3949520085318166e-06,
+      "loss": 1.5716,
+      "step": 14517
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3949101859015912e-06,
+      "loss": 1.2095,
+      "step": 14518
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394868363271366e-06,
+      "loss": 0.0015,
+      "step": 14519
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3948265406411407e-06,
+      "loss": 0.0404,
+      "step": 14520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3947847180109155e-06,
+      "loss": 0.0013,
+      "step": 14521
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3947428953806903e-06,
+      "loss": 0.0105,
+      "step": 14522
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394701072750465e-06,
+      "loss": 0.0003,
+      "step": 14523
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.39465925012024e-06,
+      "loss": 0.0013,
+      "step": 14524
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3946174274900149e-06,
+      "loss": 0.0022,
+      "step": 14525
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3945756048597897e-06,
+      "loss": 0.0014,
+      "step": 14526
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3945337822295644e-06,
+      "loss": 0.0006,
+      "step": 14527
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3944919595993392e-06,
+      "loss": 0.0006,
+      "step": 14528
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394450136969114e-06,
+      "loss": 0.0034,
+      "step": 14529
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3944083143388888e-06,
+      "loss": 0.0003,
+      "step": 14530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3943664917086636e-06,
+      "loss": 0.0004,
+      "step": 14531
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3943246690784382e-06,
+      "loss": 0.3367,
+      "step": 14532
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394282846448213e-06,
+      "loss": 0.0007,
+      "step": 14533
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3942410238179877e-06,
+      "loss": 4.131,
+      "step": 14534
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3941992011877625e-06,
+      "loss": 0.0114,
+      "step": 14535
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3941573785575373e-06,
+      "loss": 0.0012,
+      "step": 14536
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394115555927312e-06,
+      "loss": 0.0323,
+      "step": 14537
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3940737332970869e-06,
+      "loss": 0.0011,
+      "step": 14538
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3940319106668619e-06,
+      "loss": 0.0068,
+      "step": 14539
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3939900880366367e-06,
+      "loss": 0.0033,
+      "step": 14540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3939482654064115e-06,
+      "loss": 1.4362,
+      "step": 14541
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3939064427761862e-06,
+      "loss": 0.001,
+      "step": 14542
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.393864620145961e-06,
+      "loss": 1.8384,
+      "step": 14543
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3938227975157358e-06,
+      "loss": 0.1566,
+      "step": 14544
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3937809748855106e-06,
+      "loss": 0.0016,
+      "step": 14545
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3937391522552852e-06,
+      "loss": 0.0019,
+      "step": 14546
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.39369732962506e-06,
+      "loss": 0.0004,
+      "step": 14547
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3936555069948347e-06,
+      "loss": 0.0006,
+      "step": 14548
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3936136843646095e-06,
+      "loss": 0.0423,
+      "step": 14549
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3935718617343843e-06,
+      "loss": 1.5136,
+      "step": 14550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3935300391041591e-06,
+      "loss": 0.0011,
+      "step": 14551
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3934882164739339e-06,
+      "loss": 0.0006,
+      "step": 14552
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3934463938437089e-06,
+      "loss": 0.0031,
+      "step": 14553
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3934045712134837e-06,
+      "loss": 0.0003,
+      "step": 14554
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3933627485832585e-06,
+      "loss": 0.0015,
+      "step": 14555
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3933209259530332e-06,
+      "loss": 0.0155,
+      "step": 14556
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.393279103322808e-06,
+      "loss": 0.0014,
+      "step": 14557
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3932372806925828e-06,
+      "loss": 0.0037,
+      "step": 14558
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3931954580623574e-06,
+      "loss": 0.0015,
+      "step": 14559
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3931536354321322e-06,
+      "loss": 0.2874,
+      "step": 14560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.393111812801907e-06,
+      "loss": 0.0008,
+      "step": 14561
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3930699901716818e-06,
+      "loss": 0.0008,
+      "step": 14562
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3930281675414565e-06,
+      "loss": 0.0008,
+      "step": 14563
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3929863449112313e-06,
+      "loss": 0.0007,
+      "step": 14564
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3929445222810061e-06,
+      "loss": 0.0042,
+      "step": 14565
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392902699650781e-06,
+      "loss": 0.001,
+      "step": 14566
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3928608770205557e-06,
+      "loss": 0.0029,
+      "step": 14567
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3928190543903307e-06,
+      "loss": 0.0015,
+      "step": 14568
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3927772317601055e-06,
+      "loss": 0.4166,
+      "step": 14569
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3927354091298803e-06,
+      "loss": 0.708,
+      "step": 14570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392693586499655e-06,
+      "loss": 0.4003,
+      "step": 14571
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3926517638694298e-06,
+      "loss": 0.3734,
+      "step": 14572
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3926099412392044e-06,
+      "loss": 0.1113,
+      "step": 14573
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3925681186089792e-06,
+      "loss": 0.0004,
+      "step": 14574
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392526295978754e-06,
+      "loss": 0.0005,
+      "step": 14575
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3924844733485288e-06,
+      "loss": 0.0185,
+      "step": 14576
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3924426507183036e-06,
+      "loss": 0.0012,
+      "step": 14577
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3924008280880783e-06,
+      "loss": 0.0006,
+      "step": 14578
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3923590054578531e-06,
+      "loss": 0.0015,
+      "step": 14579
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392317182827628e-06,
+      "loss": 0.0039,
+      "step": 14580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3922753601974027e-06,
+      "loss": 0.0005,
+      "step": 14581
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3922335375671775e-06,
+      "loss": 0.005,
+      "step": 14582
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3921917149369525e-06,
+      "loss": 0.0006,
+      "step": 14583
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3921498923067273e-06,
+      "loss": 0.0011,
+      "step": 14584
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392108069676502e-06,
+      "loss": 0.2195,
+      "step": 14585
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3920662470462766e-06,
+      "loss": 0.4295,
+      "step": 14586
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3920244244160514e-06,
+      "loss": 0.005,
+      "step": 14587
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3919826017858262e-06,
+      "loss": 0.0085,
+      "step": 14588
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.391940779155601e-06,
+      "loss": 0.0008,
+      "step": 14589
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3918989565253758e-06,
+      "loss": 0.0005,
+      "step": 14590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3918571338951506e-06,
+      "loss": 0.219,
+      "step": 14591
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3918153112649253e-06,
+      "loss": 2.7157,
+      "step": 14592
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3917734886347001e-06,
+      "loss": 0.142,
+      "step": 14593
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.391731666004475e-06,
+      "loss": 0.0004,
+      "step": 14594
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3916898433742497e-06,
+      "loss": 0.0018,
+      "step": 14595
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3916480207440245e-06,
+      "loss": 0.0033,
+      "step": 14596
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3916061981137995e-06,
+      "loss": 0.0005,
+      "step": 14597
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3915643754835743e-06,
+      "loss": 0.0008,
+      "step": 14598
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.391522552853349e-06,
+      "loss": 0.0011,
+      "step": 14599
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3914807302231236e-06,
+      "loss": 0.0009,
+      "step": 14600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3914389075928984e-06,
+      "loss": 3.4426,
+      "step": 14601
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3913970849626732e-06,
+      "loss": 0.0019,
+      "step": 14602
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.391355262332448e-06,
+      "loss": 0.9373,
+      "step": 14603
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3913134397022228e-06,
+      "loss": 0.0007,
+      "step": 14604
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3912716170719976e-06,
+      "loss": 0.001,
+      "step": 14605
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3912297944417724e-06,
+      "loss": 0.0021,
+      "step": 14606
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3911879718115471e-06,
+      "loss": 0.0012,
+      "step": 14607
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.391146149181322e-06,
+      "loss": 0.0075,
+      "step": 14608
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3911043265510967e-06,
+      "loss": 0.0006,
+      "step": 14609
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3910625039208715e-06,
+      "loss": 0.0018,
+      "step": 14610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3910206812906463e-06,
+      "loss": 0.0003,
+      "step": 14611
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3909788586604213e-06,
+      "loss": 0.7126,
+      "step": 14612
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.390937036030196e-06,
+      "loss": 0.0005,
+      "step": 14613
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3908952133999706e-06,
+      "loss": 3.3824,
+      "step": 14614
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3908533907697454e-06,
+      "loss": 0.001,
+      "step": 14615
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3908115681395202e-06,
+      "loss": 0.0017,
+      "step": 14616
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.390769745509295e-06,
+      "loss": 2.4744,
+      "step": 14617
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3907279228790698e-06,
+      "loss": 0.0022,
+      "step": 14618
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3906861002488446e-06,
+      "loss": 1.7224,
+      "step": 14619
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3906442776186194e-06,
+      "loss": 0.0016,
+      "step": 14620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3906024549883941e-06,
+      "loss": 0.0019,
+      "step": 14621
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.390560632358169e-06,
+      "loss": 0.0126,
+      "step": 14622
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3905188097279437e-06,
+      "loss": 0.0012,
+      "step": 14623
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3904769870977185e-06,
+      "loss": 0.0032,
+      "step": 14624
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3904351644674933e-06,
+      "loss": 1.1407,
+      "step": 14625
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3903933418372679e-06,
+      "loss": 0.0005,
+      "step": 14626
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3903515192070429e-06,
+      "loss": 0.0755,
+      "step": 14627
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3903096965768177e-06,
+      "loss": 0.0013,
+      "step": 14628
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3902678739465924e-06,
+      "loss": 0.3086,
+      "step": 14629
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3902260513163672e-06,
+      "loss": 0.0022,
+      "step": 14630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.390184228686142e-06,
+      "loss": 0.0055,
+      "step": 14631
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3901424060559168e-06,
+      "loss": 0.001,
+      "step": 14632
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3901005834256916e-06,
+      "loss": 0.0005,
+      "step": 14633
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3900587607954664e-06,
+      "loss": 0.007,
+      "step": 14634
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3900169381652412e-06,
+      "loss": 0.0034,
+      "step": 14635
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389975115535016e-06,
+      "loss": 0.0034,
+      "step": 14636
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3899332929047907e-06,
+      "loss": 0.206,
+      "step": 14637
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3898914702745655e-06,
+      "loss": 0.0056,
+      "step": 14638
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3898496476443403e-06,
+      "loss": 0.0008,
+      "step": 14639
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3898078250141149e-06,
+      "loss": 0.0008,
+      "step": 14640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3897660023838899e-06,
+      "loss": 0.0009,
+      "step": 14641
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3897241797536647e-06,
+      "loss": 0.0005,
+      "step": 14642
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3896823571234394e-06,
+      "loss": 1.0434,
+      "step": 14643
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3896405344932142e-06,
+      "loss": 0.5663,
+      "step": 14644
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389598711862989e-06,
+      "loss": 0.0006,
+      "step": 14645
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3895568892327638e-06,
+      "loss": 0.0012,
+      "step": 14646
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3895150666025386e-06,
+      "loss": 0.0005,
+      "step": 14647
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3894732439723134e-06,
+      "loss": 0.001,
+      "step": 14648
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3894314213420882e-06,
+      "loss": 0.0023,
+      "step": 14649
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389389598711863e-06,
+      "loss": 0.1995,
+      "step": 14650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3893477760816377e-06,
+      "loss": 0.109,
+      "step": 14651
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3893059534514125e-06,
+      "loss": 0.001,
+      "step": 14652
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3892641308211873e-06,
+      "loss": 0.0006,
+      "step": 14653
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3892223081909619e-06,
+      "loss": 0.0004,
+      "step": 14654
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3891804855607367e-06,
+      "loss": 0.4003,
+      "step": 14655
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3891386629305117e-06,
+      "loss": 0.0109,
+      "step": 14656
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3890968403002865e-06,
+      "loss": 0.0005,
+      "step": 14657
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3890550176700612e-06,
+      "loss": 0.0038,
+      "step": 14658
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389013195039836e-06,
+      "loss": 0.0006,
+      "step": 14659
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3889713724096108e-06,
+      "loss": 0.0323,
+      "step": 14660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3889295497793856e-06,
+      "loss": 0.0007,
+      "step": 14661
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3888877271491604e-06,
+      "loss": 2.0782,
+      "step": 14662
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3888459045189352e-06,
+      "loss": 0.0043,
+      "step": 14663
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.38880408188871e-06,
+      "loss": 0.0059,
+      "step": 14664
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3887622592584847e-06,
+      "loss": 0.0009,
+      "step": 14665
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3887204366282595e-06,
+      "loss": 1.6619,
+      "step": 14666
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3886786139980341e-06,
+      "loss": 0.0006,
+      "step": 14667
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388636791367809e-06,
+      "loss": 0.0005,
+      "step": 14668
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3885949687375837e-06,
+      "loss": 1.7747,
+      "step": 14669
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3885531461073585e-06,
+      "loss": 2.1055,
+      "step": 14670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3885113234771335e-06,
+      "loss": 0.0045,
+      "step": 14671
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3884695008469083e-06,
+      "loss": 0.0165,
+      "step": 14672
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388427678216683e-06,
+      "loss": 1.8531,
+      "step": 14673
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3883858555864578e-06,
+      "loss": 1.5171,
+      "step": 14674
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3883440329562326e-06,
+      "loss": 0.0476,
+      "step": 14675
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3883022103260074e-06,
+      "loss": 0.6521,
+      "step": 14676
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3882603876957822e-06,
+      "loss": 0.5433,
+      "step": 14677
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388218565065557e-06,
+      "loss": 0.0037,
+      "step": 14678
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3881767424353318e-06,
+      "loss": 0.5852,
+      "step": 14679
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3881349198051065e-06,
+      "loss": 0.0034,
+      "step": 14680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3880930971748811e-06,
+      "loss": 0.3616,
+      "step": 14681
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388051274544656e-06,
+      "loss": 0.0008,
+      "step": 14682
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3880094519144307e-06,
+      "loss": 0.0005,
+      "step": 14683
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3879676292842055e-06,
+      "loss": 0.0014,
+      "step": 14684
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3879258066539805e-06,
+      "loss": 0.0189,
+      "step": 14685
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3878839840237553e-06,
+      "loss": 0.0005,
+      "step": 14686
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.38784216139353e-06,
+      "loss": 0.012,
+      "step": 14687
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3878003387633048e-06,
+      "loss": 0.0011,
+      "step": 14688
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3877585161330796e-06,
+      "loss": 0.6976,
+      "step": 14689
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3877166935028544e-06,
+      "loss": 0.0039,
+      "step": 14690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3876748708726292e-06,
+      "loss": 0.0005,
+      "step": 14691
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.387633048242404e-06,
+      "loss": 0.0031,
+      "step": 14692
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3875912256121788e-06,
+      "loss": 0.0012,
+      "step": 14693
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3875494029819533e-06,
+      "loss": 0.0005,
+      "step": 14694
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3875075803517281e-06,
+      "loss": 0.0027,
+      "step": 14695
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.387465757721503e-06,
+      "loss": 0.0016,
+      "step": 14696
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3874239350912777e-06,
+      "loss": 0.0038,
+      "step": 14697
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3873821124610525e-06,
+      "loss": 0.0017,
+      "step": 14698
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3873402898308273e-06,
+      "loss": 0.0024,
+      "step": 14699
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3872984672006023e-06,
+      "loss": 0.0004,
+      "step": 14700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.387256644570377e-06,
+      "loss": 0.0015,
+      "step": 14701
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3872148219401518e-06,
+      "loss": 0.0013,
+      "step": 14702
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3871729993099266e-06,
+      "loss": 0.6566,
+      "step": 14703
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3871311766797014e-06,
+      "loss": 0.003,
+      "step": 14704
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3870893540494762e-06,
+      "loss": 0.0022,
+      "step": 14705
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.387047531419251e-06,
+      "loss": 0.1306,
+      "step": 14706
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3870057087890258e-06,
+      "loss": 0.0006,
+      "step": 14707
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3869638861588003e-06,
+      "loss": 0.0029,
+      "step": 14708
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3869220635285751e-06,
+      "loss": 0.0015,
+      "step": 14709
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.38688024089835e-06,
+      "loss": 0.0014,
+      "step": 14710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3868384182681247e-06,
+      "loss": 0.0114,
+      "step": 14711
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3867965956378995e-06,
+      "loss": 0.0009,
+      "step": 14712
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3867547730076743e-06,
+      "loss": 0.0014,
+      "step": 14713
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386712950377449e-06,
+      "loss": 0.0016,
+      "step": 14714
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386671127747224e-06,
+      "loss": 0.0017,
+      "step": 14715
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3866293051169989e-06,
+      "loss": 2.1123,
+      "step": 14716
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3865874824867736e-06,
+      "loss": 0.0106,
+      "step": 14717
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3865456598565484e-06,
+      "loss": 0.0008,
+      "step": 14718
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3865038372263232e-06,
+      "loss": 0.001,
+      "step": 14719
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386462014596098e-06,
+      "loss": 0.0014,
+      "step": 14720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3864201919658726e-06,
+      "loss": 0.82,
+      "step": 14721
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3863783693356474e-06,
+      "loss": 0.0012,
+      "step": 14722
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3863365467054221e-06,
+      "loss": 0.0087,
+      "step": 14723
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386294724075197e-06,
+      "loss": 0.0007,
+      "step": 14724
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3862529014449717e-06,
+      "loss": 0.0007,
+      "step": 14725
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3862110788147465e-06,
+      "loss": 0.0004,
+      "step": 14726
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3861692561845213e-06,
+      "loss": 0.0002,
+      "step": 14727
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386127433554296e-06,
+      "loss": 0.2415,
+      "step": 14728
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386085610924071e-06,
+      "loss": 0.0008,
+      "step": 14729
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3860437882938459e-06,
+      "loss": 0.0032,
+      "step": 14730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3860019656636206e-06,
+      "loss": 0.0003,
+      "step": 14731
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3859601430333954e-06,
+      "loss": 0.0009,
+      "step": 14732
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3859183204031702e-06,
+      "loss": 2.3227,
+      "step": 14733
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385876497772945e-06,
+      "loss": 0.0015,
+      "step": 14734
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3858346751427196e-06,
+      "loss": 0.0007,
+      "step": 14735
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3857928525124944e-06,
+      "loss": 1.8276,
+      "step": 14736
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3857510298822692e-06,
+      "loss": 0.0017,
+      "step": 14737
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385709207252044e-06,
+      "loss": 1.2517,
+      "step": 14738
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3856673846218187e-06,
+      "loss": 0.0065,
+      "step": 14739
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3856255619915935e-06,
+      "loss": 0.0013,
+      "step": 14740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3855837393613683e-06,
+      "loss": 0.0027,
+      "step": 14741
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385541916731143e-06,
+      "loss": 0.6161,
+      "step": 14742
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3855000941009179e-06,
+      "loss": 0.0005,
+      "step": 14743
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3854582714706929e-06,
+      "loss": 0.0016,
+      "step": 14744
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3854164488404677e-06,
+      "loss": 0.0667,
+      "step": 14745
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3853746262102424e-06,
+      "loss": 0.2088,
+      "step": 14746
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3853328035800172e-06,
+      "loss": 3.6975,
+      "step": 14747
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385290980949792e-06,
+      "loss": 0.002,
+      "step": 14748
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3852491583195666e-06,
+      "loss": 0.1398,
+      "step": 14749
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3852073356893414e-06,
+      "loss": 0.0014,
+      "step": 14750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3851655130591162e-06,
+      "loss": 0.0012,
+      "step": 14751
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385123690428891e-06,
+      "loss": 0.0022,
+      "step": 14752
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3850818677986657e-06,
+      "loss": 0.0753,
+      "step": 14753
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3850400451684405e-06,
+      "loss": 0.0118,
+      "step": 14754
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3849982225382153e-06,
+      "loss": 0.0003,
+      "step": 14755
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.38495639990799e-06,
+      "loss": 1.0377,
+      "step": 14756
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3849145772777649e-06,
+      "loss": 0.0029,
+      "step": 14757
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3848727546475399e-06,
+      "loss": 1.3673,
+      "step": 14758
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3848309320173147e-06,
+      "loss": 0.0003,
+      "step": 14759
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3847891093870894e-06,
+      "loss": 0.0038,
+      "step": 14760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3847472867568642e-06,
+      "loss": 0.002,
+      "step": 14761
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3847054641266388e-06,
+      "loss": 0.0016,
+      "step": 14762
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3846636414964136e-06,
+      "loss": 3.468,
+      "step": 14763
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3846218188661884e-06,
+      "loss": 0.0018,
+      "step": 14764
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3845799962359632e-06,
+      "loss": 1.5962,
+      "step": 14765
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.384538173605738e-06,
+      "loss": 0.0005,
+      "step": 14766
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3844963509755127e-06,
+      "loss": 0.0007,
+      "step": 14767
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3844545283452875e-06,
+      "loss": 0.0009,
+      "step": 14768
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3844127057150623e-06,
+      "loss": 0.0006,
+      "step": 14769
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.384370883084837e-06,
+      "loss": 0.0032,
+      "step": 14770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3843290604546119e-06,
+      "loss": 0.4112,
+      "step": 14771
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3842872378243867e-06,
+      "loss": 0.0027,
+      "step": 14772
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3842454151941617e-06,
+      "loss": 0.0026,
+      "step": 14773
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3842035925639365e-06,
+      "loss": 0.0058,
+      "step": 14774
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3841617699337112e-06,
+      "loss": 0.7697,
+      "step": 14775
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3841199473034858e-06,
+      "loss": 0.0006,
+      "step": 14776
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3840781246732606e-06,
+      "loss": 0.0004,
+      "step": 14777
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3840363020430354e-06,
+      "loss": 0.0011,
+      "step": 14778
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3839944794128102e-06,
+      "loss": 0.008,
+      "step": 14779
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.383952656782585e-06,
+      "loss": 0.6661,
+      "step": 14780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3839108341523598e-06,
+      "loss": 1.0086,
+      "step": 14781
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3838690115221345e-06,
+      "loss": 0.0007,
+      "step": 14782
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3838271888919093e-06,
+      "loss": 0.0047,
+      "step": 14783
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3837853662616841e-06,
+      "loss": 0.0005,
+      "step": 14784
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.383743543631459e-06,
+      "loss": 0.0043,
+      "step": 14785
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3837017210012337e-06,
+      "loss": 0.0003,
+      "step": 14786
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3836598983710085e-06,
+      "loss": 0.0006,
+      "step": 14787
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3836180757407835e-06,
+      "loss": 0.001,
+      "step": 14788
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.383576253110558e-06,
+      "loss": 0.0036,
+      "step": 14789
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3835344304803328e-06,
+      "loss": 0.003,
+      "step": 14790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3834926078501076e-06,
+      "loss": 0.0046,
+      "step": 14791
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3834507852198824e-06,
+      "loss": 0.0005,
+      "step": 14792
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3834089625896572e-06,
+      "loss": 0.0023,
+      "step": 14793
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.383367139959432e-06,
+      "loss": 0.0003,
+      "step": 14794
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3833253173292068e-06,
+      "loss": 0.0004,
+      "step": 14795
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3832834946989815e-06,
+      "loss": 2.8974,
+      "step": 14796
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3832416720687563e-06,
+      "loss": 0.3069,
+      "step": 14797
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3831998494385311e-06,
+      "loss": 0.0012,
+      "step": 14798
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.383158026808306e-06,
+      "loss": 0.0028,
+      "step": 14799
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3831162041780807e-06,
+      "loss": 0.0005,
+      "step": 14800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3830743815478555e-06,
+      "loss": 0.0023,
+      "step": 14801
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3830325589176305e-06,
+      "loss": 0.0042,
+      "step": 14802
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382990736287405e-06,
+      "loss": 0.0008,
+      "step": 14803
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3829489136571798e-06,
+      "loss": 0.0157,
+      "step": 14804
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3829070910269546e-06,
+      "loss": 0.0009,
+      "step": 14805
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3828652683967294e-06,
+      "loss": 0.0007,
+      "step": 14806
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3828234457665042e-06,
+      "loss": 0.0003,
+      "step": 14807
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382781623136279e-06,
+      "loss": 0.0002,
+      "step": 14808
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3827398005060538e-06,
+      "loss": 0.0013,
+      "step": 14809
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3826979778758286e-06,
+      "loss": 0.4767,
+      "step": 14810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3826561552456033e-06,
+      "loss": 0.0026,
+      "step": 14811
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3826143326153781e-06,
+      "loss": 0.0026,
+      "step": 14812
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382572509985153e-06,
+      "loss": 0.0033,
+      "step": 14813
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3825306873549277e-06,
+      "loss": 0.0014,
+      "step": 14814
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3824888647247025e-06,
+      "loss": 0.1946,
+      "step": 14815
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382447042094477e-06,
+      "loss": 0.0014,
+      "step": 14816
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382405219464252e-06,
+      "loss": 0.0018,
+      "step": 14817
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3823633968340268e-06,
+      "loss": 0.0003,
+      "step": 14818
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3823215742038016e-06,
+      "loss": 0.001,
+      "step": 14819
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3822797515735764e-06,
+      "loss": 0.0003,
+      "step": 14820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3822379289433512e-06,
+      "loss": 1.7874,
+      "step": 14821
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382196106313126e-06,
+      "loss": 0.0022,
+      "step": 14822
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3821542836829008e-06,
+      "loss": 0.0024,
+      "step": 14823
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3821124610526756e-06,
+      "loss": 0.003,
+      "step": 14824
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3820706384224503e-06,
+      "loss": 0.0033,
+      "step": 14825
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3820288157922251e-06,
+      "loss": 0.0015,
+      "step": 14826
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381986993162e-06,
+      "loss": 3.9032,
+      "step": 14827
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3819451705317747e-06,
+      "loss": 0.0006,
+      "step": 14828
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3819033479015493e-06,
+      "loss": 0.0055,
+      "step": 14829
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381861525271324e-06,
+      "loss": 0.0005,
+      "step": 14830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3818197026410989e-06,
+      "loss": 0.0008,
+      "step": 14831
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3817778800108739e-06,
+      "loss": 0.0004,
+      "step": 14832
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3817360573806486e-06,
+      "loss": 0.0006,
+      "step": 14833
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3816942347504234e-06,
+      "loss": 0.0006,
+      "step": 14834
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3816524121201982e-06,
+      "loss": 0.0194,
+      "step": 14835
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381610589489973e-06,
+      "loss": 0.0005,
+      "step": 14836
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3815687668597478e-06,
+      "loss": 0.6955,
+      "step": 14837
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3815269442295226e-06,
+      "loss": 0.0003,
+      "step": 14838
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3814851215992974e-06,
+      "loss": 0.0005,
+      "step": 14839
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3814432989690721e-06,
+      "loss": 0.0022,
+      "step": 14840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381401476338847e-06,
+      "loss": 1.5838,
+      "step": 14841
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3813596537086217e-06,
+      "loss": 0.0012,
+      "step": 14842
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3813178310783963e-06,
+      "loss": 0.0005,
+      "step": 14843
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381276008448171e-06,
+      "loss": 0.0013,
+      "step": 14844
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3812341858179459e-06,
+      "loss": 0.0066,
+      "step": 14845
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3811923631877209e-06,
+      "loss": 0.0004,
+      "step": 14846
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3811505405574956e-06,
+      "loss": 0.0003,
+      "step": 14847
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3811087179272704e-06,
+      "loss": 0.0755,
+      "step": 14848
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3810668952970452e-06,
+      "loss": 0.0016,
+      "step": 14849
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.38102507266682e-06,
+      "loss": 0.0003,
+      "step": 14850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3809832500365948e-06,
+      "loss": 0.1513,
+      "step": 14851
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3809414274063696e-06,
+      "loss": 0.0008,
+      "step": 14852
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3808996047761444e-06,
+      "loss": 0.003,
+      "step": 14853
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3808577821459192e-06,
+      "loss": 0.0237,
+      "step": 14854
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380815959515694e-06,
+      "loss": 0.0009,
+      "step": 14855
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3807741368854687e-06,
+      "loss": 0.0008,
+      "step": 14856
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3807323142552433e-06,
+      "loss": 0.0009,
+      "step": 14857
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380690491625018e-06,
+      "loss": 0.007,
+      "step": 14858
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3806486689947929e-06,
+      "loss": 0.356,
+      "step": 14859
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3806068463645677e-06,
+      "loss": 0.0006,
+      "step": 14860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3805650237343427e-06,
+      "loss": 0.0006,
+      "step": 14861
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3805232011041174e-06,
+      "loss": 0.0008,
+      "step": 14862
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3804813784738922e-06,
+      "loss": 0.0005,
+      "step": 14863
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380439555843667e-06,
+      "loss": 0.0784,
+      "step": 14864
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3803977332134418e-06,
+      "loss": 0.0006,
+      "step": 14865
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3803559105832166e-06,
+      "loss": 2.0362,
+      "step": 14866
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3803140879529914e-06,
+      "loss": 0.0003,
+      "step": 14867
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3802722653227662e-06,
+      "loss": 0.0016,
+      "step": 14868
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380230442692541e-06,
+      "loss": 0.0003,
+      "step": 14869
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3801886200623155e-06,
+      "loss": 0.0006,
+      "step": 14870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3801467974320903e-06,
+      "loss": 0.8686,
+      "step": 14871
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380104974801865e-06,
+      "loss": 0.2923,
+      "step": 14872
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3800631521716399e-06,
+      "loss": 0.3674,
+      "step": 14873
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3800213295414147e-06,
+      "loss": 0.0007,
+      "step": 14874
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3799795069111895e-06,
+      "loss": 3.0997,
+      "step": 14875
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3799376842809645e-06,
+      "loss": 0.0015,
+      "step": 14876
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3798958616507392e-06,
+      "loss": 0.0559,
+      "step": 14877
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379854039020514e-06,
+      "loss": 0.0031,
+      "step": 14878
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3798122163902888e-06,
+      "loss": 0.0025,
+      "step": 14879
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3797703937600636e-06,
+      "loss": 0.0017,
+      "step": 14880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3797285711298384e-06,
+      "loss": 0.0028,
+      "step": 14881
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3796867484996132e-06,
+      "loss": 0.0003,
+      "step": 14882
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379644925869388e-06,
+      "loss": 0.0097,
+      "step": 14883
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3796031032391625e-06,
+      "loss": 2.2608,
+      "step": 14884
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3795612806089373e-06,
+      "loss": 0.0017,
+      "step": 14885
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379519457978712e-06,
+      "loss": 0.0013,
+      "step": 14886
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3794776353484869e-06,
+      "loss": 0.0045,
+      "step": 14887
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3794358127182617e-06,
+      "loss": 0.0011,
+      "step": 14888
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3793939900880365e-06,
+      "loss": 0.0011,
+      "step": 14889
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3793521674578115e-06,
+      "loss": 0.0612,
+      "step": 14890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3793103448275862e-06,
+      "loss": 0.266,
+      "step": 14891
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379268522197361e-06,
+      "loss": 0.0015,
+      "step": 14892
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3792266995671358e-06,
+      "loss": 0.7912,
+      "step": 14893
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3791848769369106e-06,
+      "loss": 0.7313,
+      "step": 14894
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3791430543066854e-06,
+      "loss": 0.0055,
+      "step": 14895
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3791012316764602e-06,
+      "loss": 0.0003,
+      "step": 14896
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3790594090462348e-06,
+      "loss": 0.0003,
+      "step": 14897
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3790175864160095e-06,
+      "loss": 0.0014,
+      "step": 14898
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3789757637857843e-06,
+      "loss": 0.0005,
+      "step": 14899
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3789339411555591e-06,
+      "loss": 0.0004,
+      "step": 14900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.378892118525334e-06,
+      "loss": 0.0018,
+      "step": 14901
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3788502958951087e-06,
+      "loss": 3.0198,
+      "step": 14902
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3788084732648835e-06,
+      "loss": 0.0048,
+      "step": 14903
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3787666506346583e-06,
+      "loss": 0.0004,
+      "step": 14904
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3787248280044333e-06,
+      "loss": 0.0006,
+      "step": 14905
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.378683005374208e-06,
+      "loss": 0.0013,
+      "step": 14906
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3786411827439828e-06,
+      "loss": 0.3523,
+      "step": 14907
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3785993601137576e-06,
+      "loss": 0.0046,
+      "step": 14908
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3785575374835324e-06,
+      "loss": 0.0002,
+      "step": 14909
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3785157148533072e-06,
+      "loss": 0.0402,
+      "step": 14910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3784738922230818e-06,
+      "loss": 0.0003,
+      "step": 14911
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3784320695928565e-06,
+      "loss": 0.226,
+      "step": 14912
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3783902469626313e-06,
+      "loss": 0.0027,
+      "step": 14913
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3783484243324061e-06,
+      "loss": 0.0026,
+      "step": 14914
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.378306601702181e-06,
+      "loss": 0.0013,
+      "step": 14915
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3782647790719557e-06,
+      "loss": 0.0034,
+      "step": 14916
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3782229564417305e-06,
+      "loss": 0.0018,
+      "step": 14917
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3781811338115053e-06,
+      "loss": 0.3176,
+      "step": 14918
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.37813931118128e-06,
+      "loss": 0.0032,
+      "step": 14919
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.378097488551055e-06,
+      "loss": 0.0003,
+      "step": 14920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3780556659208298e-06,
+      "loss": 0.0002,
+      "step": 14921
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3780138432906046e-06,
+      "loss": 0.0165,
+      "step": 14922
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3779720206603794e-06,
+      "loss": 0.11,
+      "step": 14923
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.377930198030154e-06,
+      "loss": 3.3092,
+      "step": 14924
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3778883753999288e-06,
+      "loss": 0.0815,
+      "step": 14925
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3778465527697036e-06,
+      "loss": 0.0033,
+      "step": 14926
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3778047301394783e-06,
+      "loss": 3.3837,
+      "step": 14927
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3777629075092531e-06,
+      "loss": 0.0003,
+      "step": 14928
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.377721084879028e-06,
+      "loss": 0.0008,
+      "step": 14929
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3776792622488027e-06,
+      "loss": 0.0005,
+      "step": 14930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3776374396185775e-06,
+      "loss": 0.0021,
+      "step": 14931
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3775956169883523e-06,
+      "loss": 0.0004,
+      "step": 14932
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.377553794358127e-06,
+      "loss": 0.0034,
+      "step": 14933
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.377511971727902e-06,
+      "loss": 0.0002,
+      "step": 14934
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3774701490976768e-06,
+      "loss": 0.0006,
+      "step": 14935
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3774283264674516e-06,
+      "loss": 0.0008,
+      "step": 14936
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3773865038372264e-06,
+      "loss": 0.1641,
+      "step": 14937
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.377344681207001e-06,
+      "loss": 1.8442,
+      "step": 14938
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3773028585767758e-06,
+      "loss": 0.3145,
+      "step": 14939
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3772610359465506e-06,
+      "loss": 0.0008,
+      "step": 14940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3772192133163254e-06,
+      "loss": 1.1522,
+      "step": 14941
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3771773906861001e-06,
+      "loss": 0.0008,
+      "step": 14942
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.377135568055875e-06,
+      "loss": 0.0017,
+      "step": 14943
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3770937454256497e-06,
+      "loss": 0.0003,
+      "step": 14944
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3770519227954245e-06,
+      "loss": 0.0002,
+      "step": 14945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3770101001651993e-06,
+      "loss": 0.0023,
+      "step": 14946
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.376968277534974e-06,
+      "loss": 0.2055,
+      "step": 14947
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3769264549047489e-06,
+      "loss": 0.0005,
+      "step": 14948
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3768846322745239e-06,
+      "loss": 0.0368,
+      "step": 14949
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3768428096442986e-06,
+      "loss": 0.0009,
+      "step": 14950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3768009870140734e-06,
+      "loss": 2.4617,
+      "step": 14951
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.376759164383848e-06,
+      "loss": 0.0026,
+      "step": 14952
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3767173417536228e-06,
+      "loss": 0.0003,
+      "step": 14953
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3766755191233976e-06,
+      "loss": 0.0008,
+      "step": 14954
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3766336964931724e-06,
+      "loss": 2.5173,
+      "step": 14955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3765918738629471e-06,
+      "loss": 0.0049,
+      "step": 14956
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.376550051232722e-06,
+      "loss": 0.0189,
+      "step": 14957
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3765082286024967e-06,
+      "loss": 1.748,
+      "step": 14958
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3764664059722715e-06,
+      "loss": 0.0014,
+      "step": 14959
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3764245833420463e-06,
+      "loss": 0.2256,
+      "step": 14960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.376382760711821e-06,
+      "loss": 0.0015,
+      "step": 14961
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3763409380815959e-06,
+      "loss": 0.0023,
+      "step": 14962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3762991154513709e-06,
+      "loss": 0.0004,
+      "step": 14963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3762572928211456e-06,
+      "loss": 0.0012,
+      "step": 14964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3762154701909202e-06,
+      "loss": 0.0044,
+      "step": 14965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.376173647560695e-06,
+      "loss": 0.0015,
+      "step": 14966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3761318249304698e-06,
+      "loss": 2.5034,
+      "step": 14967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3760900023002446e-06,
+      "loss": 0.0031,
+      "step": 14968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3760481796700194e-06,
+      "loss": 0.0017,
+      "step": 14969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3760063570397942e-06,
+      "loss": 0.0031,
+      "step": 14970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375964534409569e-06,
+      "loss": 0.002,
+      "step": 14971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3759227117793437e-06,
+      "loss": 2.5578,
+      "step": 14972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3758808891491185e-06,
+      "loss": 0.0009,
+      "step": 14973
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3758390665188933e-06,
+      "loss": 0.0011,
+      "step": 14974
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375797243888668e-06,
+      "loss": 0.001,
+      "step": 14975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3757554212584429e-06,
+      "loss": 0.0171,
+      "step": 14976
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3757135986282177e-06,
+      "loss": 0.0013,
+      "step": 14977
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3756717759979927e-06,
+      "loss": 0.0023,
+      "step": 14978
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3756299533677672e-06,
+      "loss": 0.0033,
+      "step": 14979
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375588130737542e-06,
+      "loss": 0.0018,
+      "step": 14980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3755463081073168e-06,
+      "loss": 0.6116,
+      "step": 14981
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3755044854770916e-06,
+      "loss": 0.0003,
+      "step": 14982
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3754626628468664e-06,
+      "loss": 2.0713,
+      "step": 14983
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3754208402166412e-06,
+      "loss": 0.9119,
+      "step": 14984
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375379017586416e-06,
+      "loss": 0.0036,
+      "step": 14985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3753371949561907e-06,
+      "loss": 0.0268,
+      "step": 14986
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3752953723259655e-06,
+      "loss": 0.0009,
+      "step": 14987
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3752535496957403e-06,
+      "loss": 0.0006,
+      "step": 14988
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375211727065515e-06,
+      "loss": 0.0031,
+      "step": 14989
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3751699044352899e-06,
+      "loss": 0.0016,
+      "step": 14990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3751280818050647e-06,
+      "loss": 0.0013,
+      "step": 14991
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3750862591748392e-06,
+      "loss": 0.0003,
+      "step": 14992
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3750444365446142e-06,
+      "loss": 3.2013,
+      "step": 14993
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375002613914389e-06,
+      "loss": 2.222,
+      "step": 14994
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3749607912841638e-06,
+      "loss": 0.0004,
+      "step": 14995
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3749189686539386e-06,
+      "loss": 0.0015,
+      "step": 14996
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3748771460237134e-06,
+      "loss": 0.0007,
+      "step": 14997
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3748353233934882e-06,
+      "loss": 0.0072,
+      "step": 14998
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374793500763263e-06,
+      "loss": 1.6858,
+      "step": 14999
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3747516781330377e-06,
+      "loss": 0.0019,
+      "step": 15000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3747098555028125e-06,
+      "loss": 2.7683,
+      "step": 15001
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3746680328725873e-06,
+      "loss": 0.0005,
+      "step": 15002
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374626210242362e-06,
+      "loss": 2.2765,
+      "step": 15003
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3745843876121369e-06,
+      "loss": 0.0004,
+      "step": 15004
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3745425649819115e-06,
+      "loss": 0.0014,
+      "step": 15005
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3745007423516863e-06,
+      "loss": 0.0022,
+      "step": 15006
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3744589197214612e-06,
+      "loss": 0.0005,
+      "step": 15007
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374417097091236e-06,
+      "loss": 0.0014,
+      "step": 15008
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3743752744610108e-06,
+      "loss": 0.0049,
+      "step": 15009
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3743334518307856e-06,
+      "loss": 0.5148,
+      "step": 15010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3742916292005604e-06,
+      "loss": 0.0045,
+      "step": 15011
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3742498065703352e-06,
+      "loss": 0.0097,
+      "step": 15012
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.37420798394011e-06,
+      "loss": 1.5141,
+      "step": 15013
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3741661613098848e-06,
+      "loss": 0.0013,
+      "step": 15014
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3741243386796595e-06,
+      "loss": 0.0003,
+      "step": 15015
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3740825160494343e-06,
+      "loss": 0.001,
+      "step": 15016
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3740406934192091e-06,
+      "loss": 0.0007,
+      "step": 15017
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373998870788984e-06,
+      "loss": 0.0042,
+      "step": 15018
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3739570481587585e-06,
+      "loss": 0.003,
+      "step": 15019
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3739152255285333e-06,
+      "loss": 0.2845,
+      "step": 15020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373873402898308e-06,
+      "loss": 0.0019,
+      "step": 15021
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373831580268083e-06,
+      "loss": 0.0008,
+      "step": 15022
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3737897576378578e-06,
+      "loss": 0.0042,
+      "step": 15023
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3737479350076326e-06,
+      "loss": 0.4919,
+      "step": 15024
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3737061123774074e-06,
+      "loss": 0.0005,
+      "step": 15025
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3736642897471822e-06,
+      "loss": 0.4302,
+      "step": 15026
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373622467116957e-06,
+      "loss": 0.0037,
+      "step": 15027
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3735806444867318e-06,
+      "loss": 0.0022,
+      "step": 15028
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3735388218565065e-06,
+      "loss": 0.0022,
+      "step": 15029
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3734969992262813e-06,
+      "loss": 1.7061,
+      "step": 15030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3734551765960561e-06,
+      "loss": 0.0042,
+      "step": 15031
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3734133539658307e-06,
+      "loss": 0.0004,
+      "step": 15032
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3733715313356055e-06,
+      "loss": 0.0021,
+      "step": 15033
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3733297087053803e-06,
+      "loss": 0.0004,
+      "step": 15034
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373287886075155e-06,
+      "loss": 0.651,
+      "step": 15035
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3732460634449298e-06,
+      "loss": 0.0004,
+      "step": 15036
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3732042408147048e-06,
+      "loss": 0.2776,
+      "step": 15037
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3731624181844796e-06,
+      "loss": 0.0035,
+      "step": 15038
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3731205955542544e-06,
+      "loss": 0.0016,
+      "step": 15039
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3730787729240292e-06,
+      "loss": 0.0024,
+      "step": 15040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373036950293804e-06,
+      "loss": 0.0004,
+      "step": 15041
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3729951276635788e-06,
+      "loss": 0.0015,
+      "step": 15042
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3729533050333536e-06,
+      "loss": 2.223,
+      "step": 15043
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3729114824031283e-06,
+      "loss": 0.0021,
+      "step": 15044
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3728696597729031e-06,
+      "loss": 0.0369,
+      "step": 15045
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3728278371426777e-06,
+      "loss": 0.0219,
+      "step": 15046
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3727860145124525e-06,
+      "loss": 0.0005,
+      "step": 15047
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3727441918822273e-06,
+      "loss": 0.0004,
+      "step": 15048
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.372702369252002e-06,
+      "loss": 0.0007,
+      "step": 15049
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3726605466217768e-06,
+      "loss": 0.0094,
+      "step": 15050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3726187239915518e-06,
+      "loss": 0.0008,
+      "step": 15051
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3725769013613266e-06,
+      "loss": 0.3323,
+      "step": 15052
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3725350787311014e-06,
+      "loss": 0.0004,
+      "step": 15053
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3724932561008762e-06,
+      "loss": 2.0126,
+      "step": 15054
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.372451433470651e-06,
+      "loss": 0.0026,
+      "step": 15055
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3724096108404258e-06,
+      "loss": 0.0026,
+      "step": 15056
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3723677882102006e-06,
+      "loss": 2.409,
+      "step": 15057
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3723259655799754e-06,
+      "loss": 0.0005,
+      "step": 15058
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3722841429497501e-06,
+      "loss": 2.3179,
+      "step": 15059
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3722423203195247e-06,
+      "loss": 0.0028,
+      "step": 15060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3722004976892995e-06,
+      "loss": 2.1769,
+      "step": 15061
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3721586750590743e-06,
+      "loss": 2.3128,
+      "step": 15062
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.372116852428849e-06,
+      "loss": 0.0004,
+      "step": 15063
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3720750297986239e-06,
+      "loss": 0.007,
+      "step": 15064
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3720332071683986e-06,
+      "loss": 2.5171,
+      "step": 15065
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3719913845381736e-06,
+      "loss": 0.0032,
+      "step": 15066
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3719495619079484e-06,
+      "loss": 1.108,
+      "step": 15067
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3719077392777232e-06,
+      "loss": 0.0005,
+      "step": 15068
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.371865916647498e-06,
+      "loss": 0.0015,
+      "step": 15069
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3718240940172728e-06,
+      "loss": 0.4052,
+      "step": 15070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3717822713870476e-06,
+      "loss": 0.0007,
+      "step": 15071
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3717404487568224e-06,
+      "loss": 1.992,
+      "step": 15072
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.371698626126597e-06,
+      "loss": 0.0021,
+      "step": 15073
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3716568034963717e-06,
+      "loss": 0.0011,
+      "step": 15074
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3716149808661465e-06,
+      "loss": 0.0065,
+      "step": 15075
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3715731582359213e-06,
+      "loss": 0.0037,
+      "step": 15076
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.371531335605696e-06,
+      "loss": 0.0028,
+      "step": 15077
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3714895129754709e-06,
+      "loss": 0.1585,
+      "step": 15078
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3714476903452457e-06,
+      "loss": 0.0018,
+      "step": 15079
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3714058677150204e-06,
+      "loss": 0.0011,
+      "step": 15080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3713640450847954e-06,
+      "loss": 0.0025,
+      "step": 15081
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3713222224545702e-06,
+      "loss": 0.0015,
+      "step": 15082
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.371280399824345e-06,
+      "loss": 0.0004,
+      "step": 15083
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3712385771941198e-06,
+      "loss": 0.0007,
+      "step": 15084
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3711967545638946e-06,
+      "loss": 0.0065,
+      "step": 15085
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3711549319336694e-06,
+      "loss": 0.0034,
+      "step": 15086
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.371113109303444e-06,
+      "loss": 1.3486,
+      "step": 15087
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3710712866732187e-06,
+      "loss": 0.0009,
+      "step": 15088
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3710294640429935e-06,
+      "loss": 0.0013,
+      "step": 15089
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3709876414127683e-06,
+      "loss": 2.0659,
+      "step": 15090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.370945818782543e-06,
+      "loss": 0.0017,
+      "step": 15091
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3709039961523179e-06,
+      "loss": 0.0007,
+      "step": 15092
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3708621735220927e-06,
+      "loss": 0.0011,
+      "step": 15093
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3708203508918674e-06,
+      "loss": 0.4202,
+      "step": 15094
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3707785282616424e-06,
+      "loss": 2.0616,
+      "step": 15095
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3707367056314172e-06,
+      "loss": 0.0024,
+      "step": 15096
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.370694883001192e-06,
+      "loss": 0.011,
+      "step": 15097
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3706530603709668e-06,
+      "loss": 0.0017,
+      "step": 15098
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3706112377407416e-06,
+      "loss": 0.5581,
+      "step": 15099
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3705694151105162e-06,
+      "loss": 0.1114,
+      "step": 15100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.370527592480291e-06,
+      "loss": 0.009,
+      "step": 15101
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3704857698500657e-06,
+      "loss": 0.0016,
+      "step": 15102
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3704439472198405e-06,
+      "loss": 2.6775,
+      "step": 15103
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3704021245896153e-06,
+      "loss": 0.3411,
+      "step": 15104
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.37036030195939e-06,
+      "loss": 0.0116,
+      "step": 15105
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3703184793291649e-06,
+      "loss": 0.0017,
+      "step": 15106
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3702766566989397e-06,
+      "loss": 0.0116,
+      "step": 15107
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3702348340687145e-06,
+      "loss": 1.9878,
+      "step": 15108
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3701930114384892e-06,
+      "loss": 0.003,
+      "step": 15109
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3701511888082642e-06,
+      "loss": 0.0011,
+      "step": 15110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.370109366178039e-06,
+      "loss": 0.0031,
+      "step": 15111
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3700675435478138e-06,
+      "loss": 0.007,
+      "step": 15112
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3700257209175886e-06,
+      "loss": 0.0177,
+      "step": 15113
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3699838982873632e-06,
+      "loss": 0.0051,
+      "step": 15114
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369942075657138e-06,
+      "loss": 0.0009,
+      "step": 15115
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3699002530269127e-06,
+      "loss": 0.0006,
+      "step": 15116
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3698584303966875e-06,
+      "loss": 0.0393,
+      "step": 15117
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3698166077664623e-06,
+      "loss": 0.0012,
+      "step": 15118
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3697747851362371e-06,
+      "loss": 0.0005,
+      "step": 15119
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3697329625060119e-06,
+      "loss": 0.0006,
+      "step": 15120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3696911398757867e-06,
+      "loss": 0.0166,
+      "step": 15121
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3696493172455615e-06,
+      "loss": 0.0012,
+      "step": 15122
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3696074946153363e-06,
+      "loss": 1.8574,
+      "step": 15123
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369565671985111e-06,
+      "loss": 0.0035,
+      "step": 15124
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369523849354886e-06,
+      "loss": 0.0007,
+      "step": 15125
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3694820267246608e-06,
+      "loss": 0.0011,
+      "step": 15126
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3694402040944354e-06,
+      "loss": 0.0266,
+      "step": 15127
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3693983814642102e-06,
+      "loss": 0.0013,
+      "step": 15128
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369356558833985e-06,
+      "loss": 0.0022,
+      "step": 15129
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3693147362037598e-06,
+      "loss": 0.0006,
+      "step": 15130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3692729135735345e-06,
+      "loss": 0.775,
+      "step": 15131
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3692310909433093e-06,
+      "loss": 0.0007,
+      "step": 15132
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3691892683130841e-06,
+      "loss": 0.0565,
+      "step": 15133
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369147445682859e-06,
+      "loss": 0.0025,
+      "step": 15134
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3691056230526337e-06,
+      "loss": 0.0008,
+      "step": 15135
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3690638004224085e-06,
+      "loss": 0.0027,
+      "step": 15136
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3690219777921833e-06,
+      "loss": 0.0011,
+      "step": 15137
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368980155161958e-06,
+      "loss": 0.0287,
+      "step": 15138
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368938332531733e-06,
+      "loss": 0.001,
+      "step": 15139
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3688965099015078e-06,
+      "loss": 2.022,
+      "step": 15140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3688546872712824e-06,
+      "loss": 0.0039,
+      "step": 15141
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3688128646410572e-06,
+      "loss": 0.003,
+      "step": 15142
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368771042010832e-06,
+      "loss": 0.0035,
+      "step": 15143
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3687292193806068e-06,
+      "loss": 1.1828,
+      "step": 15144
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3686873967503816e-06,
+      "loss": 1.1819,
+      "step": 15145
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3686455741201563e-06,
+      "loss": 0.0004,
+      "step": 15146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3686037514899311e-06,
+      "loss": 0.0114,
+      "step": 15147
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368561928859706e-06,
+      "loss": 0.0044,
+      "step": 15148
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3685201062294807e-06,
+      "loss": 0.0053,
+      "step": 15149
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3684782835992555e-06,
+      "loss": 0.0002,
+      "step": 15150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3684364609690303e-06,
+      "loss": 0.0011,
+      "step": 15151
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368394638338805e-06,
+      "loss": 0.0003,
+      "step": 15152
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3683528157085798e-06,
+      "loss": 0.0025,
+      "step": 15153
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3683109930783548e-06,
+      "loss": 0.0006,
+      "step": 15154
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3682691704481294e-06,
+      "loss": 0.0195,
+      "step": 15155
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3682273478179042e-06,
+      "loss": 0.0019,
+      "step": 15156
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368185525187679e-06,
+      "loss": 0.0012,
+      "step": 15157
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3681437025574538e-06,
+      "loss": 0.0018,
+      "step": 15158
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3681018799272286e-06,
+      "loss": 0.0011,
+      "step": 15159
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3680600572970033e-06,
+      "loss": 0.0013,
+      "step": 15160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3680182346667781e-06,
+      "loss": 0.0012,
+      "step": 15161
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367976412036553e-06,
+      "loss": 1.7462,
+      "step": 15162
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3679345894063277e-06,
+      "loss": 0.001,
+      "step": 15163
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3678927667761025e-06,
+      "loss": 0.0004,
+      "step": 15164
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3678509441458773e-06,
+      "loss": 0.0007,
+      "step": 15165
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367809121515652e-06,
+      "loss": 0.0062,
+      "step": 15166
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3677672988854269e-06,
+      "loss": 0.0175,
+      "step": 15167
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3677254762552016e-06,
+      "loss": 0.0018,
+      "step": 15168
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3676836536249764e-06,
+      "loss": 0.0006,
+      "step": 15169
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3676418309947512e-06,
+      "loss": 0.3265,
+      "step": 15170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367600008364526e-06,
+      "loss": 0.128,
+      "step": 15171
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3675581857343008e-06,
+      "loss": 0.3763,
+      "step": 15172
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3675163631040756e-06,
+      "loss": 4.541,
+      "step": 15173
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3674745404738504e-06,
+      "loss": 0.0011,
+      "step": 15174
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3674327178436251e-06,
+      "loss": 0.0065,
+      "step": 15175
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3673908952134e-06,
+      "loss": 0.0062,
+      "step": 15176
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3673490725831747e-06,
+      "loss": 0.0069,
+      "step": 15177
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3673072499529495e-06,
+      "loss": 0.0005,
+      "step": 15178
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3672654273227243e-06,
+      "loss": 0.0021,
+      "step": 15179
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367223604692499e-06,
+      "loss": 0.1028,
+      "step": 15180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3671817820622736e-06,
+      "loss": 0.001,
+      "step": 15181
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3671399594320484e-06,
+      "loss": 0.0018,
+      "step": 15182
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3670981368018234e-06,
+      "loss": 0.3627,
+      "step": 15183
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3670563141715982e-06,
+      "loss": 0.0045,
+      "step": 15184
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367014491541373e-06,
+      "loss": 0.0026,
+      "step": 15185
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3669726689111478e-06,
+      "loss": 0.0025,
+      "step": 15186
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3669308462809226e-06,
+      "loss": 2.0337,
+      "step": 15187
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3668890236506974e-06,
+      "loss": 0.0005,
+      "step": 15188
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3668472010204721e-06,
+      "loss": 0.0017,
+      "step": 15189
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366805378390247e-06,
+      "loss": 0.0003,
+      "step": 15190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3667635557600217e-06,
+      "loss": 0.0652,
+      "step": 15191
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3667217331297965e-06,
+      "loss": 0.0021,
+      "step": 15192
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3666799104995713e-06,
+      "loss": 0.0017,
+      "step": 15193
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366638087869346e-06,
+      "loss": 0.0061,
+      "step": 15194
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3665962652391207e-06,
+      "loss": 2.0025,
+      "step": 15195
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3665544426088954e-06,
+      "loss": 0.0022,
+      "step": 15196
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3665126199786702e-06,
+      "loss": 0.0051,
+      "step": 15197
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3664707973484452e-06,
+      "loss": 0.0029,
+      "step": 15198
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.36642897471822e-06,
+      "loss": 0.0054,
+      "step": 15199
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3663871520879948e-06,
+      "loss": 0.0013,
+      "step": 15200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3663453294577696e-06,
+      "loss": 0.0006,
+      "step": 15201
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3663035068275444e-06,
+      "loss": 0.0004,
+      "step": 15202
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3662616841973192e-06,
+      "loss": 0.0003,
+      "step": 15203
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366219861567094e-06,
+      "loss": 0.0027,
+      "step": 15204
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3661780389368687e-06,
+      "loss": 0.0111,
+      "step": 15205
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3661362163066435e-06,
+      "loss": 0.002,
+      "step": 15206
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3660943936764183e-06,
+      "loss": 0.0061,
+      "step": 15207
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3660525710461929e-06,
+      "loss": 0.0021,
+      "step": 15208
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3660107484159677e-06,
+      "loss": 0.0004,
+      "step": 15209
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3659689257857425e-06,
+      "loss": 0.0129,
+      "step": 15210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3659271031555172e-06,
+      "loss": 1.4036,
+      "step": 15211
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3658852805252922e-06,
+      "loss": 0.0018,
+      "step": 15212
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365843457895067e-06,
+      "loss": 0.0038,
+      "step": 15213
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3658016352648418e-06,
+      "loss": 1.9827,
+      "step": 15214
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3657598126346166e-06,
+      "loss": 0.0003,
+      "step": 15215
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3657179900043914e-06,
+      "loss": 0.8288,
+      "step": 15216
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3656761673741662e-06,
+      "loss": 0.0008,
+      "step": 15217
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365634344743941e-06,
+      "loss": 0.0031,
+      "step": 15218
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3655925221137157e-06,
+      "loss": 0.0024,
+      "step": 15219
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3655506994834905e-06,
+      "loss": 0.0003,
+      "step": 15220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3655088768532653e-06,
+      "loss": 0.0011,
+      "step": 15221
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3654670542230399e-06,
+      "loss": 0.0033,
+      "step": 15222
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3654252315928147e-06,
+      "loss": 0.2256,
+      "step": 15223
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3653834089625895e-06,
+      "loss": 2.395,
+      "step": 15224
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3653415863323642e-06,
+      "loss": 0.0013,
+      "step": 15225
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365299763702139e-06,
+      "loss": 0.0012,
+      "step": 15226
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365257941071914e-06,
+      "loss": 3.5851,
+      "step": 15227
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3652161184416888e-06,
+      "loss": 0.2594,
+      "step": 15228
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3651742958114636e-06,
+      "loss": 0.0014,
+      "step": 15229
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3651324731812384e-06,
+      "loss": 0.0047,
+      "step": 15230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3650906505510132e-06,
+      "loss": 0.0006,
+      "step": 15231
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365048827920788e-06,
+      "loss": 0.0057,
+      "step": 15232
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3650070052905627e-06,
+      "loss": 0.0055,
+      "step": 15233
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3649651826603375e-06,
+      "loss": 0.0013,
+      "step": 15234
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3649233600301121e-06,
+      "loss": 0.0027,
+      "step": 15235
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.364881537399887e-06,
+      "loss": 0.0065,
+      "step": 15236
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3648397147696617e-06,
+      "loss": 2.6643,
+      "step": 15237
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3647978921394365e-06,
+      "loss": 0.2231,
+      "step": 15238
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3647560695092113e-06,
+      "loss": 1.8762,
+      "step": 15239
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.364714246878986e-06,
+      "loss": 0.0012,
+      "step": 15240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3646724242487608e-06,
+      "loss": 0.0044,
+      "step": 15241
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3646306016185358e-06,
+      "loss": 0.0248,
+      "step": 15242
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3645887789883106e-06,
+      "loss": 0.0081,
+      "step": 15243
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3645469563580854e-06,
+      "loss": 0.0005,
+      "step": 15244
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3645051337278602e-06,
+      "loss": 0.0005,
+      "step": 15245
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.364463311097635e-06,
+      "loss": 0.0008,
+      "step": 15246
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3644214884674098e-06,
+      "loss": 0.0007,
+      "step": 15247
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3643796658371845e-06,
+      "loss": 0.001,
+      "step": 15248
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3643378432069591e-06,
+      "loss": 2.2941,
+      "step": 15249
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.364296020576734e-06,
+      "loss": 2.547,
+      "step": 15250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3642541979465087e-06,
+      "loss": 0.6906,
+      "step": 15251
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3642123753162835e-06,
+      "loss": 0.0017,
+      "step": 15252
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3641705526860583e-06,
+      "loss": 0.0182,
+      "step": 15253
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.364128730055833e-06,
+      "loss": 0.0011,
+      "step": 15254
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3640869074256078e-06,
+      "loss": 0.8505,
+      "step": 15255
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3640450847953828e-06,
+      "loss": 0.0037,
+      "step": 15256
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3640032621651576e-06,
+      "loss": 0.0003,
+      "step": 15257
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3639614395349324e-06,
+      "loss": 0.0041,
+      "step": 15258
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3639196169047072e-06,
+      "loss": 0.0004,
+      "step": 15259
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.363877794274482e-06,
+      "loss": 0.0007,
+      "step": 15260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3638359716442568e-06,
+      "loss": 0.0045,
+      "step": 15261
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3637941490140316e-06,
+      "loss": 1.8794,
+      "step": 15262
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3637523263838061e-06,
+      "loss": 0.0095,
+      "step": 15263
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.363710503753581e-06,
+      "loss": 0.2402,
+      "step": 15264
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3636686811233557e-06,
+      "loss": 0.0032,
+      "step": 15265
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3636268584931305e-06,
+      "loss": 1.7685,
+      "step": 15266
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3635850358629053e-06,
+      "loss": 3.4722,
+      "step": 15267
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.36354321323268e-06,
+      "loss": 0.92,
+      "step": 15268
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3635013906024548e-06,
+      "loss": 0.0053,
+      "step": 15269
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3634595679722296e-06,
+      "loss": 0.0039,
+      "step": 15270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3634177453420046e-06,
+      "loss": 0.0004,
+      "step": 15271
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3633759227117794e-06,
+      "loss": 0.0007,
+      "step": 15272
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3633341000815542e-06,
+      "loss": 0.0202,
+      "step": 15273
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.363292277451329e-06,
+      "loss": 0.6428,
+      "step": 15274
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3632504548211038e-06,
+      "loss": 0.2416,
+      "step": 15275
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3632086321908783e-06,
+      "loss": 0.0044,
+      "step": 15276
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3631668095606531e-06,
+      "loss": 0.0024,
+      "step": 15277
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.363124986930428e-06,
+      "loss": 0.0056,
+      "step": 15278
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3630831643002027e-06,
+      "loss": 0.0007,
+      "step": 15279
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3630413416699775e-06,
+      "loss": 0.0074,
+      "step": 15280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3629995190397523e-06,
+      "loss": 0.0004,
+      "step": 15281
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362957696409527e-06,
+      "loss": 0.0013,
+      "step": 15282
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3629158737793019e-06,
+      "loss": 0.001,
+      "step": 15283
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3628740511490766e-06,
+      "loss": 0.0006,
+      "step": 15284
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3628322285188514e-06,
+      "loss": 0.0056,
+      "step": 15285
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3627904058886264e-06,
+      "loss": 0.0003,
+      "step": 15286
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3627485832584012e-06,
+      "loss": 0.0019,
+      "step": 15287
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362706760628176e-06,
+      "loss": 3.025,
+      "step": 15288
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3626649379979508e-06,
+      "loss": 0.0024,
+      "step": 15289
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3626231153677254e-06,
+      "loss": 0.0016,
+      "step": 15290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3625812927375001e-06,
+      "loss": 0.001,
+      "step": 15291
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362539470107275e-06,
+      "loss": 0.002,
+      "step": 15292
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3624976474770497e-06,
+      "loss": 0.0006,
+      "step": 15293
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3624558248468245e-06,
+      "loss": 1.9832,
+      "step": 15294
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3624140022165993e-06,
+      "loss": 0.0022,
+      "step": 15295
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362372179586374e-06,
+      "loss": 0.0008,
+      "step": 15296
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3623303569561489e-06,
+      "loss": 0.1916,
+      "step": 15297
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3622885343259236e-06,
+      "loss": 3.2768,
+      "step": 15298
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3622467116956984e-06,
+      "loss": 0.7644,
+      "step": 15299
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3622048890654734e-06,
+      "loss": 0.0008,
+      "step": 15300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3621630664352482e-06,
+      "loss": 0.0021,
+      "step": 15301
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362121243805023e-06,
+      "loss": 2.2518,
+      "step": 15302
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3620794211747976e-06,
+      "loss": 0.2976,
+      "step": 15303
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3620375985445724e-06,
+      "loss": 0.0009,
+      "step": 15304
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3619957759143472e-06,
+      "loss": 0.0009,
+      "step": 15305
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361953953284122e-06,
+      "loss": 0.0041,
+      "step": 15306
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3619121306538967e-06,
+      "loss": 0.5157,
+      "step": 15307
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3618703080236715e-06,
+      "loss": 0.0064,
+      "step": 15308
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3618284853934463e-06,
+      "loss": 0.0005,
+      "step": 15309
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361786662763221e-06,
+      "loss": 4.3183,
+      "step": 15310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3617448401329959e-06,
+      "loss": 0.0006,
+      "step": 15311
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3617030175027707e-06,
+      "loss": 0.0041,
+      "step": 15312
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3616611948725454e-06,
+      "loss": 0.0054,
+      "step": 15313
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3616193722423202e-06,
+      "loss": 0.0119,
+      "step": 15314
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3615775496120952e-06,
+      "loss": 0.9575,
+      "step": 15315
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.36153572698187e-06,
+      "loss": 0.0006,
+      "step": 15316
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3614939043516446e-06,
+      "loss": 0.0024,
+      "step": 15317
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3614520817214194e-06,
+      "loss": 0.0039,
+      "step": 15318
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3614102590911942e-06,
+      "loss": 0.0004,
+      "step": 15319
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361368436460969e-06,
+      "loss": 0.0005,
+      "step": 15320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3613266138307437e-06,
+      "loss": 0.0382,
+      "step": 15321
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3612847912005185e-06,
+      "loss": 0.0031,
+      "step": 15322
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3612429685702933e-06,
+      "loss": 0.0039,
+      "step": 15323
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361201145940068e-06,
+      "loss": 0.003,
+      "step": 15324
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3611593233098429e-06,
+      "loss": 0.0026,
+      "step": 15325
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3611175006796177e-06,
+      "loss": 0.0006,
+      "step": 15326
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3610756780493925e-06,
+      "loss": 0.4143,
+      "step": 15327
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3610338554191672e-06,
+      "loss": 0.0238,
+      "step": 15328
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3609920327889422e-06,
+      "loss": 0.003,
+      "step": 15329
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360950210158717e-06,
+      "loss": 0.0011,
+      "step": 15330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3609083875284916e-06,
+      "loss": 0.6793,
+      "step": 15331
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3608665648982664e-06,
+      "loss": 0.0017,
+      "step": 15332
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3608247422680412e-06,
+      "loss": 0.3587,
+      "step": 15333
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360782919637816e-06,
+      "loss": 0.0057,
+      "step": 15334
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3607410970075907e-06,
+      "loss": 0.0015,
+      "step": 15335
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3606992743773655e-06,
+      "loss": 0.0011,
+      "step": 15336
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3606574517471403e-06,
+      "loss": 0.001,
+      "step": 15337
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360615629116915e-06,
+      "loss": 0.2522,
+      "step": 15338
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3605738064866899e-06,
+      "loss": 0.0018,
+      "step": 15339
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3605319838564647e-06,
+      "loss": 0.0006,
+      "step": 15340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3604901612262395e-06,
+      "loss": 0.001,
+      "step": 15341
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3604483385960142e-06,
+      "loss": 0.3318,
+      "step": 15342
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3604065159657888e-06,
+      "loss": 1.3801,
+      "step": 15343
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3603646933355638e-06,
+      "loss": 1.3089,
+      "step": 15344
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3603228707053386e-06,
+      "loss": 0.001,
+      "step": 15345
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3602810480751134e-06,
+      "loss": 0.0008,
+      "step": 15346
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3602392254448882e-06,
+      "loss": 0.0014,
+      "step": 15347
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360197402814663e-06,
+      "loss": 0.0194,
+      "step": 15348
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3601555801844378e-06,
+      "loss": 3.4686,
+      "step": 15349
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3601137575542125e-06,
+      "loss": 0.0007,
+      "step": 15350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3600719349239873e-06,
+      "loss": 1.283,
+      "step": 15351
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3600301122937621e-06,
+      "loss": 1.3599,
+      "step": 15352
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359988289663537e-06,
+      "loss": 0.0033,
+      "step": 15353
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3599464670333117e-06,
+      "loss": 0.0024,
+      "step": 15354
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3599046444030865e-06,
+      "loss": 0.0043,
+      "step": 15355
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3598628217728613e-06,
+      "loss": 0.0125,
+      "step": 15356
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3598209991426358e-06,
+      "loss": 0.0034,
+      "step": 15357
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3597791765124106e-06,
+      "loss": 0.0006,
+      "step": 15358
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3597373538821856e-06,
+      "loss": 0.1923,
+      "step": 15359
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3596955312519604e-06,
+      "loss": 0.0181,
+      "step": 15360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3596537086217352e-06,
+      "loss": 0.0048,
+      "step": 15361
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.35961188599151e-06,
+      "loss": 1.8899,
+      "step": 15362
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3595700633612848e-06,
+      "loss": 0.0009,
+      "step": 15363
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3595282407310595e-06,
+      "loss": 0.02,
+      "step": 15364
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3594864181008343e-06,
+      "loss": 0.0004,
+      "step": 15365
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3594445954706091e-06,
+      "loss": 2.1024,
+      "step": 15366
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359402772840384e-06,
+      "loss": 0.0007,
+      "step": 15367
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3593609502101587e-06,
+      "loss": 0.0102,
+      "step": 15368
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3593191275799335e-06,
+      "loss": 0.0028,
+      "step": 15369
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3592773049497083e-06,
+      "loss": 0.0057,
+      "step": 15370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3592354823194828e-06,
+      "loss": 0.0031,
+      "step": 15371
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3591936596892576e-06,
+      "loss": 1.2878,
+      "step": 15372
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3591518370590326e-06,
+      "loss": 0.002,
+      "step": 15373
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3591100144288074e-06,
+      "loss": 0.0019,
+      "step": 15374
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3590681917985822e-06,
+      "loss": 0.5909,
+      "step": 15375
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359026369168357e-06,
+      "loss": 0.0007,
+      "step": 15376
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3589845465381318e-06,
+      "loss": 0.0006,
+      "step": 15377
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3589427239079066e-06,
+      "loss": 2.4294,
+      "step": 15378
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3589009012776813e-06,
+      "loss": 0.0165,
+      "step": 15379
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3588590786474561e-06,
+      "loss": 0.0028,
+      "step": 15380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358817256017231e-06,
+      "loss": 0.0008,
+      "step": 15381
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3587754333870057e-06,
+      "loss": 2.1369,
+      "step": 15382
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3587336107567805e-06,
+      "loss": 0.0007,
+      "step": 15383
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358691788126555e-06,
+      "loss": 0.0153,
+      "step": 15384
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3586499654963298e-06,
+      "loss": 0.0014,
+      "step": 15385
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3586081428661046e-06,
+      "loss": 0.0032,
+      "step": 15386
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3585663202358794e-06,
+      "loss": 0.001,
+      "step": 15387
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3585244976056544e-06,
+      "loss": 0.0013,
+      "step": 15388
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3584826749754292e-06,
+      "loss": 0.0022,
+      "step": 15389
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358440852345204e-06,
+      "loss": 1.4298,
+      "step": 15390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3583990297149788e-06,
+      "loss": 0.2159,
+      "step": 15391
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3583572070847536e-06,
+      "loss": 0.003,
+      "step": 15392
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3583153844545283e-06,
+      "loss": 0.0012,
+      "step": 15393
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3582735618243031e-06,
+      "loss": 0.0008,
+      "step": 15394
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358231739194078e-06,
+      "loss": 0.0078,
+      "step": 15395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3581899165638527e-06,
+      "loss": 0.0042,
+      "step": 15396
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3581480939336275e-06,
+      "loss": 0.0005,
+      "step": 15397
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358106271303402e-06,
+      "loss": 0.0071,
+      "step": 15398
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3580644486731769e-06,
+      "loss": 0.0053,
+      "step": 15399
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3580226260429516e-06,
+      "loss": 0.0004,
+      "step": 15400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3579808034127264e-06,
+      "loss": 0.001,
+      "step": 15401
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3579389807825012e-06,
+      "loss": 0.0077,
+      "step": 15402
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3578971581522762e-06,
+      "loss": 0.0011,
+      "step": 15403
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.357855335522051e-06,
+      "loss": 0.0115,
+      "step": 15404
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3578135128918258e-06,
+      "loss": 0.8835,
+      "step": 15405
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3577716902616006e-06,
+      "loss": 0.0006,
+      "step": 15406
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3577298676313754e-06,
+      "loss": 0.0014,
+      "step": 15407
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3576880450011501e-06,
+      "loss": 1.088,
+      "step": 15408
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.357646222370925e-06,
+      "loss": 0.0028,
+      "step": 15409
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3576043997406997e-06,
+      "loss": 0.8913,
+      "step": 15410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3575625771104743e-06,
+      "loss": 0.0012,
+      "step": 15411
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.357520754480249e-06,
+      "loss": 0.4085,
+      "step": 15412
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3574789318500239e-06,
+      "loss": 2.0964,
+      "step": 15413
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3574371092197986e-06,
+      "loss": 0.0008,
+      "step": 15414
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3573952865895734e-06,
+      "loss": 0.008,
+      "step": 15415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3573534639593482e-06,
+      "loss": 0.0004,
+      "step": 15416
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3573116413291232e-06,
+      "loss": 0.0004,
+      "step": 15417
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.357269818698898e-06,
+      "loss": 0.9466,
+      "step": 15418
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3572279960686728e-06,
+      "loss": 2.2381,
+      "step": 15419
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3571861734384476e-06,
+      "loss": 0.0009,
+      "step": 15420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3571443508082224e-06,
+      "loss": 0.0019,
+      "step": 15421
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3571025281779972e-06,
+      "loss": 0.0021,
+      "step": 15422
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.357060705547772e-06,
+      "loss": 1.4263,
+      "step": 15423
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3570188829175467e-06,
+      "loss": 0.0015,
+      "step": 15424
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3569770602873213e-06,
+      "loss": 0.1009,
+      "step": 15425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.356935237657096e-06,
+      "loss": 0.0013,
+      "step": 15426
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3568934150268709e-06,
+      "loss": 0.0056,
+      "step": 15427
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3568515923966457e-06,
+      "loss": 0.0007,
+      "step": 15428
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3568097697664204e-06,
+      "loss": 1.9872,
+      "step": 15429
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3567679471361952e-06,
+      "loss": 0.0951,
+      "step": 15430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.35672612450597e-06,
+      "loss": 0.0018,
+      "step": 15431
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.356684301875745e-06,
+      "loss": 0.003,
+      "step": 15432
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3566424792455198e-06,
+      "loss": 0.0105,
+      "step": 15433
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3566006566152946e-06,
+      "loss": 0.0005,
+      "step": 15434
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3565588339850694e-06,
+      "loss": 0.0031,
+      "step": 15435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3565170113548442e-06,
+      "loss": 0.0047,
+      "step": 15436
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.356475188724619e-06,
+      "loss": 0.0016,
+      "step": 15437
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3564333660943935e-06,
+      "loss": 0.0012,
+      "step": 15438
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3563915434641683e-06,
+      "loss": 2.4086,
+      "step": 15439
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.356349720833943e-06,
+      "loss": 0.0009,
+      "step": 15440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3563078982037179e-06,
+      "loss": 0.0019,
+      "step": 15441
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3562660755734927e-06,
+      "loss": 0.0003,
+      "step": 15442
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3562242529432675e-06,
+      "loss": 0.0008,
+      "step": 15443
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3561824303130422e-06,
+      "loss": 1.4355,
+      "step": 15444
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.356140607682817e-06,
+      "loss": 0.6422,
+      "step": 15445
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3560987850525918e-06,
+      "loss": 0.0068,
+      "step": 15446
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3560569624223668e-06,
+      "loss": 0.0033,
+      "step": 15447
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3560151397921416e-06,
+      "loss": 0.0052,
+      "step": 15448
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3559733171619164e-06,
+      "loss": 3.5408,
+      "step": 15449
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3559314945316912e-06,
+      "loss": 2.8878,
+      "step": 15450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.355889671901466e-06,
+      "loss": 1.7173,
+      "step": 15451
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3558478492712405e-06,
+      "loss": 2.0004,
+      "step": 15452
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3558060266410153e-06,
+      "loss": 0.0018,
+      "step": 15453
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.35576420401079e-06,
+      "loss": 1.0763,
+      "step": 15454
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3557223813805649e-06,
+      "loss": 0.0105,
+      "step": 15455
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3556805587503397e-06,
+      "loss": 0.1737,
+      "step": 15456
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3556387361201145e-06,
+      "loss": 0.0024,
+      "step": 15457
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3555969134898892e-06,
+      "loss": 0.0373,
+      "step": 15458
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.355555090859664e-06,
+      "loss": 0.0032,
+      "step": 15459
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3555132682294388e-06,
+      "loss": 0.0018,
+      "step": 15460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3554714455992138e-06,
+      "loss": 0.7708,
+      "step": 15461
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3554296229689886e-06,
+      "loss": 0.0013,
+      "step": 15462
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3553878003387634e-06,
+      "loss": 0.0059,
+      "step": 15463
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3553459777085382e-06,
+      "loss": 2.1375,
+      "step": 15464
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.355304155078313e-06,
+      "loss": 0.0026,
+      "step": 15465
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3552623324480875e-06,
+      "loss": 0.5197,
+      "step": 15466
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3552205098178623e-06,
+      "loss": 0.0035,
+      "step": 15467
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3551786871876371e-06,
+      "loss": 0.0008,
+      "step": 15468
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.355136864557412e-06,
+      "loss": 0.0036,
+      "step": 15469
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3550950419271867e-06,
+      "loss": 0.0095,
+      "step": 15470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3550532192969615e-06,
+      "loss": 0.0416,
+      "step": 15471
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3550113966667363e-06,
+      "loss": 0.0029,
+      "step": 15472
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.354969574036511e-06,
+      "loss": 0.1048,
+      "step": 15473
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3549277514062858e-06,
+      "loss": 1.3256,
+      "step": 15474
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3548859287760606e-06,
+      "loss": 0.7485,
+      "step": 15475
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3548441061458356e-06,
+      "loss": 0.7213,
+      "step": 15476
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3548022835156104e-06,
+      "loss": 0.0051,
+      "step": 15477
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3547604608853852e-06,
+      "loss": 0.0041,
+      "step": 15478
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3547186382551598e-06,
+      "loss": 0.0009,
+      "step": 15479
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3546768156249345e-06,
+      "loss": 0.0019,
+      "step": 15480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3546349929947093e-06,
+      "loss": 0.0032,
+      "step": 15481
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3545931703644841e-06,
+      "loss": 2.3916,
+      "step": 15482
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.354551347734259e-06,
+      "loss": 0.0071,
+      "step": 15483
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3545095251040337e-06,
+      "loss": 0.229,
+      "step": 15484
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3544677024738085e-06,
+      "loss": 0.0007,
+      "step": 15485
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3544258798435833e-06,
+      "loss": 0.002,
+      "step": 15486
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.354384057213358e-06,
+      "loss": 0.0005,
+      "step": 15487
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3543422345831328e-06,
+      "loss": 0.0024,
+      "step": 15488
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3543004119529076e-06,
+      "loss": 0.0043,
+      "step": 15489
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3542585893226824e-06,
+      "loss": 0.0038,
+      "step": 15490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3542167666924574e-06,
+      "loss": 0.0013,
+      "step": 15491
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3541749440622322e-06,
+      "loss": 0.7934,
+      "step": 15492
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3541331214320068e-06,
+      "loss": 0.7513,
+      "step": 15493
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3540912988017816e-06,
+      "loss": 0.0013,
+      "step": 15494
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3540494761715563e-06,
+      "loss": 0.2588,
+      "step": 15495
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3540076535413311e-06,
+      "loss": 0.0164,
+      "step": 15496
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353965830911106e-06,
+      "loss": 0.0014,
+      "step": 15497
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3539240082808807e-06,
+      "loss": 0.0076,
+      "step": 15498
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3538821856506555e-06,
+      "loss": 0.0037,
+      "step": 15499
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3538403630204303e-06,
+      "loss": 0.0379,
+      "step": 15500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353798540390205e-06,
+      "loss": 0.0013,
+      "step": 15501
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3537567177599798e-06,
+      "loss": 0.0009,
+      "step": 15502
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3537148951297546e-06,
+      "loss": 0.3365,
+      "step": 15503
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3536730724995294e-06,
+      "loss": 0.0004,
+      "step": 15504
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3536312498693044e-06,
+      "loss": 0.0006,
+      "step": 15505
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353589427239079e-06,
+      "loss": 3.439,
+      "step": 15506
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3535476046088538e-06,
+      "loss": 0.004,
+      "step": 15507
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3535057819786286e-06,
+      "loss": 1.4465,
+      "step": 15508
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3534639593484034e-06,
+      "loss": 1.2731,
+      "step": 15509
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3534221367181781e-06,
+      "loss": 0.0023,
+      "step": 15510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353380314087953e-06,
+      "loss": 0.0018,
+      "step": 15511
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3533384914577277e-06,
+      "loss": 1.269,
+      "step": 15512
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3532966688275025e-06,
+      "loss": 0.0012,
+      "step": 15513
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3532548461972773e-06,
+      "loss": 0.1929,
+      "step": 15514
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353213023567052e-06,
+      "loss": 1.163,
+      "step": 15515
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3531712009368269e-06,
+      "loss": 0.0223,
+      "step": 15516
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3531293783066016e-06,
+      "loss": 0.001,
+      "step": 15517
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3530875556763764e-06,
+      "loss": 0.0022,
+      "step": 15518
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353045733046151e-06,
+      "loss": 0.0044,
+      "step": 15519
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353003910415926e-06,
+      "loss": 0.0013,
+      "step": 15520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3529620877857008e-06,
+      "loss": 0.8753,
+      "step": 15521
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3529202651554756e-06,
+      "loss": 0.0003,
+      "step": 15522
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3528784425252504e-06,
+      "loss": 0.0004,
+      "step": 15523
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3528366198950251e-06,
+      "loss": 0.0104,
+      "step": 15524
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3527947972648e-06,
+      "loss": 0.0007,
+      "step": 15525
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3527529746345747e-06,
+      "loss": 1.4257,
+      "step": 15526
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3527111520043495e-06,
+      "loss": 0.2819,
+      "step": 15527
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3526693293741243e-06,
+      "loss": 0.0017,
+      "step": 15528
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352627506743899e-06,
+      "loss": 0.0039,
+      "step": 15529
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3525856841136739e-06,
+      "loss": 0.0043,
+      "step": 15530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3525438614834487e-06,
+      "loss": 0.0006,
+      "step": 15531
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3525020388532234e-06,
+      "loss": 0.0008,
+      "step": 15532
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352460216222998e-06,
+      "loss": 2.3363,
+      "step": 15533
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352418393592773e-06,
+      "loss": 0.0028,
+      "step": 15534
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3523765709625478e-06,
+      "loss": 0.0011,
+      "step": 15535
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3523347483323226e-06,
+      "loss": 0.0007,
+      "step": 15536
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3522929257020974e-06,
+      "loss": 0.001,
+      "step": 15537
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3522511030718722e-06,
+      "loss": 0.0017,
+      "step": 15538
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352209280441647e-06,
+      "loss": 0.0011,
+      "step": 15539
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3521674578114217e-06,
+      "loss": 0.0004,
+      "step": 15540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3521256351811965e-06,
+      "loss": 0.0007,
+      "step": 15541
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3520838125509713e-06,
+      "loss": 2.6577,
+      "step": 15542
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352041989920746e-06,
+      "loss": 0.0029,
+      "step": 15543
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3520001672905209e-06,
+      "loss": 2.4827,
+      "step": 15544
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3519583446602957e-06,
+      "loss": 0.0044,
+      "step": 15545
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3519165220300702e-06,
+      "loss": 0.6496,
+      "step": 15546
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351874699399845e-06,
+      "loss": 0.0003,
+      "step": 15547
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3518328767696198e-06,
+      "loss": 0.0009,
+      "step": 15548
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3517910541393948e-06,
+      "loss": 0.0068,
+      "step": 15549
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3517492315091696e-06,
+      "loss": 0.1066,
+      "step": 15550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3517074088789444e-06,
+      "loss": 0.0072,
+      "step": 15551
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3516655862487192e-06,
+      "loss": 0.003,
+      "step": 15552
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351623763618494e-06,
+      "loss": 0.0003,
+      "step": 15553
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3515819409882687e-06,
+      "loss": 0.0013,
+      "step": 15554
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3515401183580435e-06,
+      "loss": 0.003,
+      "step": 15555
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3514982957278183e-06,
+      "loss": 0.0023,
+      "step": 15556
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351456473097593e-06,
+      "loss": 0.1874,
+      "step": 15557
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3514146504673679e-06,
+      "loss": 0.0005,
+      "step": 15558
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3513728278371427e-06,
+      "loss": 0.0007,
+      "step": 15559
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3513310052069172e-06,
+      "loss": 0.0014,
+      "step": 15560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.351289182576692e-06,
+      "loss": 0.0055,
+      "step": 15561
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3512473599464668e-06,
+      "loss": 0.0006,
+      "step": 15562
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3512055373162416e-06,
+      "loss": 3.3835,
+      "step": 15563
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3511637146860166e-06,
+      "loss": 0.002,
+      "step": 15564
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3511218920557914e-06,
+      "loss": 0.0002,
+      "step": 15565
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3510800694255662e-06,
+      "loss": 2.6766,
+      "step": 15566
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.351038246795341e-06,
+      "loss": 0.0011,
+      "step": 15567
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3509964241651157e-06,
+      "loss": 0.0008,
+      "step": 15568
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3509546015348905e-06,
+      "loss": 0.004,
+      "step": 15569
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3509127789046653e-06,
+      "loss": 0.0045,
+      "step": 15570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.35087095627444e-06,
+      "loss": 2.6463,
+      "step": 15571
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3508291336442149e-06,
+      "loss": 0.0003,
+      "step": 15572
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3507873110139897e-06,
+      "loss": 1.1555,
+      "step": 15573
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3507454883837643e-06,
+      "loss": 0.0143,
+      "step": 15574
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.350703665753539e-06,
+      "loss": 1.5971,
+      "step": 15575
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3506618431233138e-06,
+      "loss": 0.0024,
+      "step": 15576
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3506200204930886e-06,
+      "loss": 0.0005,
+      "step": 15577
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3505781978628636e-06,
+      "loss": 0.0003,
+      "step": 15578
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3505363752326384e-06,
+      "loss": 0.0006,
+      "step": 15579
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3504945526024132e-06,
+      "loss": 0.0005,
+      "step": 15580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.350452729972188e-06,
+      "loss": 0.0008,
+      "step": 15581
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3504109073419628e-06,
+      "loss": 2.2148,
+      "step": 15582
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3503690847117375e-06,
+      "loss": 0.4013,
+      "step": 15583
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3503272620815123e-06,
+      "loss": 0.0008,
+      "step": 15584
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3502854394512871e-06,
+      "loss": 0.001,
+      "step": 15585
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.350243616821062e-06,
+      "loss": 0.001,
+      "step": 15586
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3502017941908365e-06,
+      "loss": 2.5093,
+      "step": 15587
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3501599715606113e-06,
+      "loss": 0.1125,
+      "step": 15588
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.350118148930386e-06,
+      "loss": 0.0033,
+      "step": 15589
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3500763263001608e-06,
+      "loss": 0.0009,
+      "step": 15590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3500345036699356e-06,
+      "loss": 0.0003,
+      "step": 15591
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3499926810397104e-06,
+      "loss": 0.0067,
+      "step": 15592
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3499508584094854e-06,
+      "loss": 0.3871,
+      "step": 15593
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3499090357792602e-06,
+      "loss": 0.003,
+      "step": 15594
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.349867213149035e-06,
+      "loss": 0.0013,
+      "step": 15595
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3498253905188098e-06,
+      "loss": 0.4606,
+      "step": 15596
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3497835678885845e-06,
+      "loss": 0.0013,
+      "step": 15597
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3497417452583593e-06,
+      "loss": 1.7507,
+      "step": 15598
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3496999226281341e-06,
+      "loss": 2.0914,
+      "step": 15599
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.349658099997909e-06,
+      "loss": 0.006,
+      "step": 15600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3496162773676835e-06,
+      "loss": 0.0045,
+      "step": 15601
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3495744547374583e-06,
+      "loss": 0.004,
+      "step": 15602
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.349532632107233e-06,
+      "loss": 0.144,
+      "step": 15603
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3494908094770078e-06,
+      "loss": 0.0003,
+      "step": 15604
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3494489868467826e-06,
+      "loss": 0.0017,
+      "step": 15605
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3494071642165574e-06,
+      "loss": 0.0006,
+      "step": 15606
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3493653415863322e-06,
+      "loss": 0.0012,
+      "step": 15607
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3493235189561072e-06,
+      "loss": 0.0004,
+      "step": 15608
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.349281696325882e-06,
+      "loss": 0.0101,
+      "step": 15609
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3492398736956568e-06,
+      "loss": 0.6612,
+      "step": 15610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3491980510654316e-06,
+      "loss": 0.0011,
+      "step": 15611
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3491562284352063e-06,
+      "loss": 0.0037,
+      "step": 15612
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3491144058049811e-06,
+      "loss": 0.0037,
+      "step": 15613
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3490725831747557e-06,
+      "loss": 0.0018,
+      "step": 15614
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3490307605445305e-06,
+      "loss": 0.0023,
+      "step": 15615
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3489889379143053e-06,
+      "loss": 0.0067,
+      "step": 15616
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.34894711528408e-06,
+      "loss": 0.3911,
+      "step": 15617
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3489052926538548e-06,
+      "loss": 0.0047,
+      "step": 15618
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3488634700236296e-06,
+      "loss": 0.0005,
+      "step": 15619
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3488216473934044e-06,
+      "loss": 0.0009,
+      "step": 15620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3487798247631792e-06,
+      "loss": 2.5274,
+      "step": 15621
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3487380021329542e-06,
+      "loss": 2.1823,
+      "step": 15622
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348696179502729e-06,
+      "loss": 0.0007,
+      "step": 15623
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3486543568725038e-06,
+      "loss": 0.0032,
+      "step": 15624
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3486125342422786e-06,
+      "loss": 0.0034,
+      "step": 15625
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3485707116120534e-06,
+      "loss": 0.0007,
+      "step": 15626
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3485288889818281e-06,
+      "loss": 0.0018,
+      "step": 15627
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3484870663516027e-06,
+      "loss": 0.0016,
+      "step": 15628
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3484452437213775e-06,
+      "loss": 0.0045,
+      "step": 15629
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3484034210911523e-06,
+      "loss": 0.0009,
+      "step": 15630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348361598460927e-06,
+      "loss": 0.0004,
+      "step": 15631
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3483197758307019e-06,
+      "loss": 0.0141,
+      "step": 15632
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3482779532004766e-06,
+      "loss": 1.6066,
+      "step": 15633
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3482361305702514e-06,
+      "loss": 0.0004,
+      "step": 15634
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3481943079400262e-06,
+      "loss": 0.0031,
+      "step": 15635
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348152485309801e-06,
+      "loss": 0.0023,
+      "step": 15636
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348110662679576e-06,
+      "loss": 0.3864,
+      "step": 15637
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3480688400493508e-06,
+      "loss": 0.0006,
+      "step": 15638
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3480270174191256e-06,
+      "loss": 0.0116,
+      "step": 15639
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3479851947889004e-06,
+      "loss": 0.0039,
+      "step": 15640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347943372158675e-06,
+      "loss": 0.0023,
+      "step": 15641
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3479015495284497e-06,
+      "loss": 0.0028,
+      "step": 15642
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3478597268982245e-06,
+      "loss": 2.3448,
+      "step": 15643
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3478179042679993e-06,
+      "loss": 1.8947,
+      "step": 15644
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347776081637774e-06,
+      "loss": 0.0017,
+      "step": 15645
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3477342590075489e-06,
+      "loss": 0.0015,
+      "step": 15646
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3476924363773237e-06,
+      "loss": 1.8032,
+      "step": 15647
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3476506137470984e-06,
+      "loss": 1.253,
+      "step": 15648
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3476087911168732e-06,
+      "loss": 0.4595,
+      "step": 15649
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347566968486648e-06,
+      "loss": 0.0051,
+      "step": 15650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3475251458564228e-06,
+      "loss": 0.0043,
+      "step": 15651
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3474833232261978e-06,
+      "loss": 0.0103,
+      "step": 15652
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3474415005959726e-06,
+      "loss": 0.0198,
+      "step": 15653
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3473996779657474e-06,
+      "loss": 0.0172,
+      "step": 15654
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347357855335522e-06,
+      "loss": 0.2926,
+      "step": 15655
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3473160327052967e-06,
+      "loss": 0.0575,
+      "step": 15656
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3472742100750715e-06,
+      "loss": 1.7355,
+      "step": 15657
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3472323874448463e-06,
+      "loss": 0.0113,
+      "step": 15658
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347190564814621e-06,
+      "loss": 0.0007,
+      "step": 15659
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3471487421843959e-06,
+      "loss": 1.8428,
+      "step": 15660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3471069195541707e-06,
+      "loss": 0.0011,
+      "step": 15661
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3470650969239454e-06,
+      "loss": 0.0029,
+      "step": 15662
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3470232742937202e-06,
+      "loss": 3.3795,
+      "step": 15663
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346981451663495e-06,
+      "loss": 0.0009,
+      "step": 15664
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3469396290332698e-06,
+      "loss": 0.0031,
+      "step": 15665
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3468978064030448e-06,
+      "loss": 0.0005,
+      "step": 15666
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3468559837728196e-06,
+      "loss": 0.0016,
+      "step": 15667
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3468141611425944e-06,
+      "loss": 0.0007,
+      "step": 15668
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346772338512369e-06,
+      "loss": 1.1458,
+      "step": 15669
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3467305158821437e-06,
+      "loss": 0.0013,
+      "step": 15670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3466886932519185e-06,
+      "loss": 0.0023,
+      "step": 15671
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3466468706216933e-06,
+      "loss": 0.0082,
+      "step": 15672
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346605047991468e-06,
+      "loss": 0.0017,
+      "step": 15673
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3465632253612429e-06,
+      "loss": 0.5745,
+      "step": 15674
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3465214027310177e-06,
+      "loss": 0.0005,
+      "step": 15675
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3464795801007925e-06,
+      "loss": 0.0042,
+      "step": 15676
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3464377574705672e-06,
+      "loss": 0.001,
+      "step": 15677
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346395934840342e-06,
+      "loss": 0.0365,
+      "step": 15678
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3463541122101168e-06,
+      "loss": 0.0046,
+      "step": 15679
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3463122895798916e-06,
+      "loss": 0.0023,
+      "step": 15680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3462704669496666e-06,
+      "loss": 0.0016,
+      "step": 15681
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3462286443194412e-06,
+      "loss": 0.0066,
+      "step": 15682
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346186821689216e-06,
+      "loss": 0.0013,
+      "step": 15683
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3461449990589907e-06,
+      "loss": 2.3038,
+      "step": 15684
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3461031764287655e-06,
+      "loss": 0.0097,
+      "step": 15685
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3460613537985403e-06,
+      "loss": 0.0068,
+      "step": 15686
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3460195311683151e-06,
+      "loss": 0.0004,
+      "step": 15687
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3459777085380899e-06,
+      "loss": 0.0308,
+      "step": 15688
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3459358859078647e-06,
+      "loss": 0.0267,
+      "step": 15689
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3458940632776395e-06,
+      "loss": 0.0023,
+      "step": 15690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3458522406474143e-06,
+      "loss": 0.0095,
+      "step": 15691
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345810418017189e-06,
+      "loss": 0.0018,
+      "step": 15692
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3457685953869638e-06,
+      "loss": 0.0048,
+      "step": 15693
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3457267727567386e-06,
+      "loss": 0.001,
+      "step": 15694
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3456849501265132e-06,
+      "loss": 0.001,
+      "step": 15695
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3456431274962882e-06,
+      "loss": 0.011,
+      "step": 15696
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345601304866063e-06,
+      "loss": 1.2632,
+      "step": 15697
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3455594822358378e-06,
+      "loss": 0.0003,
+      "step": 15698
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3455176596056125e-06,
+      "loss": 0.0006,
+      "step": 15699
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3454758369753873e-06,
+      "loss": 0.0006,
+      "step": 15700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3454340143451621e-06,
+      "loss": 0.0007,
+      "step": 15701
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345392191714937e-06,
+      "loss": 0.0012,
+      "step": 15702
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3453503690847117e-06,
+      "loss": 0.0008,
+      "step": 15703
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3453085464544865e-06,
+      "loss": 0.4335,
+      "step": 15704
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3452667238242613e-06,
+      "loss": 0.0019,
+      "step": 15705
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345224901194036e-06,
+      "loss": 0.0004,
+      "step": 15706
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3451830785638108e-06,
+      "loss": 0.0022,
+      "step": 15707
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3451412559335856e-06,
+      "loss": 0.0033,
+      "step": 15708
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3450994333033602e-06,
+      "loss": 0.2343,
+      "step": 15709
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3450576106731352e-06,
+      "loss": 0.0006,
+      "step": 15710
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.34501578804291e-06,
+      "loss": 0.0003,
+      "step": 15711
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3449739654126848e-06,
+      "loss": 0.0054,
+      "step": 15712
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3449321427824596e-06,
+      "loss": 0.7411,
+      "step": 15713
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3448903201522343e-06,
+      "loss": 0.0004,
+      "step": 15714
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3448484975220091e-06,
+      "loss": 0.0008,
+      "step": 15715
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.344806674891784e-06,
+      "loss": 0.0008,
+      "step": 15716
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3447648522615587e-06,
+      "loss": 0.8158,
+      "step": 15717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3447230296313335e-06,
+      "loss": 0.0003,
+      "step": 15718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3446812070011083e-06,
+      "loss": 0.0005,
+      "step": 15719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.344639384370883e-06,
+      "loss": 0.0021,
+      "step": 15720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3445975617406578e-06,
+      "loss": 0.0007,
+      "step": 15721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3445557391104324e-06,
+      "loss": 0.0006,
+      "step": 15722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3445139164802072e-06,
+      "loss": 2.3526,
+      "step": 15723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.344472093849982e-06,
+      "loss": 0.3512,
+      "step": 15724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.344430271219757e-06,
+      "loss": 0.0008,
+      "step": 15725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3443884485895318e-06,
+      "loss": 0.6248,
+      "step": 15726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3443466259593066e-06,
+      "loss": 0.0014,
+      "step": 15727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3443048033290813e-06,
+      "loss": 0.0003,
+      "step": 15728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3442629806988561e-06,
+      "loss": 0.0008,
+      "step": 15729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.344221158068631e-06,
+      "loss": 0.8162,
+      "step": 15730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3441793354384057e-06,
+      "loss": 0.0007,
+      "step": 15731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3441375128081805e-06,
+      "loss": 1.962,
+      "step": 15732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3440956901779553e-06,
+      "loss": 1.2589,
+      "step": 15733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.34405386754773e-06,
+      "loss": 0.0013,
+      "step": 15734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3440120449175048e-06,
+      "loss": 0.0019,
+      "step": 15735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3439702222872794e-06,
+      "loss": 0.0011,
+      "step": 15736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3439283996570542e-06,
+      "loss": 3.9787,
+      "step": 15737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.343886577026829e-06,
+      "loss": 0.0015,
+      "step": 15738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.343844754396604e-06,
+      "loss": 0.009,
+      "step": 15739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3438029317663788e-06,
+      "loss": 2.0858,
+      "step": 15740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3437611091361536e-06,
+      "loss": 0.0067,
+      "step": 15741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3437192865059284e-06,
+      "loss": 0.0057,
+      "step": 15742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3436774638757031e-06,
+      "loss": 0.0004,
+      "step": 15743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.343635641245478e-06,
+      "loss": 4.5463,
+      "step": 15744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3435938186152527e-06,
+      "loss": 0.0013,
+      "step": 15745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3435519959850275e-06,
+      "loss": 0.4131,
+      "step": 15746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3435101733548023e-06,
+      "loss": 0.0036,
+      "step": 15747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.343468350724577e-06,
+      "loss": 1.3616,
+      "step": 15748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3434265280943516e-06,
+      "loss": 0.6673,
+      "step": 15749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3433847054641264e-06,
+      "loss": 0.0007,
+      "step": 15750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3433428828339012e-06,
+      "loss": 0.0003,
+      "step": 15751
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.343301060203676e-06,
+      "loss": 0.0034,
+      "step": 15752
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3432592375734508e-06,
+      "loss": 0.2453,
+      "step": 15753
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3432174149432258e-06,
+      "loss": 0.0044,
+      "step": 15754
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3431755923130006e-06,
+      "loss": 2.2073,
+      "step": 15755
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3431337696827754e-06,
+      "loss": 0.0008,
+      "step": 15756
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3430919470525501e-06,
+      "loss": 0.1251,
+      "step": 15757
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.343050124422325e-06,
+      "loss": 0.0006,
+      "step": 15758
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3430083017920997e-06,
+      "loss": 0.1941,
+      "step": 15759
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3429664791618745e-06,
+      "loss": 0.0045,
+      "step": 15760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3429246565316493e-06,
+      "loss": 0.0004,
+      "step": 15761
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.342882833901424e-06,
+      "loss": 0.0144,
+      "step": 15762
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3428410112711987e-06,
+      "loss": 0.0039,
+      "step": 15763
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3427991886409734e-06,
+      "loss": 1.6681,
+      "step": 15764
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3427573660107482e-06,
+      "loss": 0.0032,
+      "step": 15765
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.342715543380523e-06,
+      "loss": 0.0022,
+      "step": 15766
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3426737207502978e-06,
+      "loss": 0.0029,
+      "step": 15767
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3426318981200726e-06,
+      "loss": 3.8916,
+      "step": 15768
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3425900754898476e-06,
+      "loss": 0.1218,
+      "step": 15769
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3425482528596224e-06,
+      "loss": 0.002,
+      "step": 15770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3425064302293972e-06,
+      "loss": 0.0049,
+      "step": 15771
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.342464607599172e-06,
+      "loss": 0.0073,
+      "step": 15772
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3424227849689467e-06,
+      "loss": 0.0016,
+      "step": 15773
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3423809623387215e-06,
+      "loss": 0.0005,
+      "step": 15774
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3423391397084963e-06,
+      "loss": 0.0581,
+      "step": 15775
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.342297317078271e-06,
+      "loss": 0.0019,
+      "step": 15776
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3422554944480457e-06,
+      "loss": 2.4781,
+      "step": 15777
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3422136718178205e-06,
+      "loss": 0.0006,
+      "step": 15778
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3421718491875952e-06,
+      "loss": 0.0018,
+      "step": 15779
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.34213002655737e-06,
+      "loss": 0.0009,
+      "step": 15780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3420882039271448e-06,
+      "loss": 0.0737,
+      "step": 15781
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3420463812969196e-06,
+      "loss": 1.8074,
+      "step": 15782
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3420045586666946e-06,
+      "loss": 0.0034,
+      "step": 15783
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3419627360364694e-06,
+      "loss": 0.807,
+      "step": 15784
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3419209134062442e-06,
+      "loss": 0.0175,
+      "step": 15785
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341879090776019e-06,
+      "loss": 0.3588,
+      "step": 15786
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3418372681457937e-06,
+      "loss": 0.0013,
+      "step": 15787
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3417954455155685e-06,
+      "loss": 1.7281,
+      "step": 15788
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3417536228853433e-06,
+      "loss": 0.8605,
+      "step": 15789
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3417118002551179e-06,
+      "loss": 0.0058,
+      "step": 15790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3416699776248927e-06,
+      "loss": 0.0007,
+      "step": 15791
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3416281549946675e-06,
+      "loss": 0.0072,
+      "step": 15792
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3415863323644422e-06,
+      "loss": 0.0025,
+      "step": 15793
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341544509734217e-06,
+      "loss": 0.6347,
+      "step": 15794
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3415026871039918e-06,
+      "loss": 0.0062,
+      "step": 15795
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3414608644737666e-06,
+      "loss": 0.0015,
+      "step": 15796
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3414190418435414e-06,
+      "loss": 0.0059,
+      "step": 15797
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3413772192133164e-06,
+      "loss": 0.0125,
+      "step": 15798
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3413353965830912e-06,
+      "loss": 0.0023,
+      "step": 15799
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341293573952866e-06,
+      "loss": 0.0035,
+      "step": 15800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3412517513226407e-06,
+      "loss": 0.0186,
+      "step": 15801
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3412099286924155e-06,
+      "loss": 0.0019,
+      "step": 15802
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3411681060621903e-06,
+      "loss": 0.002,
+      "step": 15803
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341126283431965e-06,
+      "loss": 0.0051,
+      "step": 15804
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3410844608017397e-06,
+      "loss": 0.0006,
+      "step": 15805
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3410426381715145e-06,
+      "loss": 0.0006,
+      "step": 15806
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3410008155412893e-06,
+      "loss": 0.0024,
+      "step": 15807
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340958992911064e-06,
+      "loss": 0.0064,
+      "step": 15808
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3409171702808388e-06,
+      "loss": 0.0006,
+      "step": 15809
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3408753476506136e-06,
+      "loss": 1.8254,
+      "step": 15810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3408335250203884e-06,
+      "loss": 0.0007,
+      "step": 15811
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3407917023901632e-06,
+      "loss": 0.0048,
+      "step": 15812
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3407498797599382e-06,
+      "loss": 0.1105,
+      "step": 15813
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340708057129713e-06,
+      "loss": 1.0757,
+      "step": 15814
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3406662344994878e-06,
+      "loss": 0.007,
+      "step": 15815
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3406244118692625e-06,
+      "loss": 0.0013,
+      "step": 15816
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3405825892390371e-06,
+      "loss": 0.0008,
+      "step": 15817
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340540766608812e-06,
+      "loss": 0.4913,
+      "step": 15818
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3404989439785867e-06,
+      "loss": 0.0009,
+      "step": 15819
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3404571213483615e-06,
+      "loss": 0.0006,
+      "step": 15820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3404152987181363e-06,
+      "loss": 0.0016,
+      "step": 15821
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340373476087911e-06,
+      "loss": 0.0012,
+      "step": 15822
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3403316534576858e-06,
+      "loss": 0.0115,
+      "step": 15823
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3402898308274606e-06,
+      "loss": 0.0007,
+      "step": 15824
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3402480081972354e-06,
+      "loss": 0.0013,
+      "step": 15825
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3402061855670102e-06,
+      "loss": 0.0032,
+      "step": 15826
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3401643629367852e-06,
+      "loss": 2.2479,
+      "step": 15827
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.34012254030656e-06,
+      "loss": 0.0025,
+      "step": 15828
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3400807176763348e-06,
+      "loss": 1.951,
+      "step": 15829
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3400388950461096e-06,
+      "loss": 0.544,
+      "step": 15830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3399970724158841e-06,
+      "loss": 0.0011,
+      "step": 15831
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339955249785659e-06,
+      "loss": 0.0058,
+      "step": 15832
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3399134271554337e-06,
+      "loss": 2.4205,
+      "step": 15833
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3398716045252085e-06,
+      "loss": 0.0005,
+      "step": 15834
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3398297818949833e-06,
+      "loss": 0.0537,
+      "step": 15835
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339787959264758e-06,
+      "loss": 0.0017,
+      "step": 15836
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3397461366345328e-06,
+      "loss": 0.0036,
+      "step": 15837
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3397043140043076e-06,
+      "loss": 0.0142,
+      "step": 15838
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3396624913740824e-06,
+      "loss": 0.0091,
+      "step": 15839
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3396206687438572e-06,
+      "loss": 0.0005,
+      "step": 15840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339578846113632e-06,
+      "loss": 0.0019,
+      "step": 15841
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339537023483407e-06,
+      "loss": 0.0005,
+      "step": 15842
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3394952008531818e-06,
+      "loss": 0.4767,
+      "step": 15843
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3394533782229563e-06,
+      "loss": 0.0003,
+      "step": 15844
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3394115555927311e-06,
+      "loss": 0.0021,
+      "step": 15845
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339369732962506e-06,
+      "loss": 0.0024,
+      "step": 15846
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3393279103322807e-06,
+      "loss": 0.0041,
+      "step": 15847
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3392860877020555e-06,
+      "loss": 0.8338,
+      "step": 15848
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3392442650718303e-06,
+      "loss": 0.0046,
+      "step": 15849
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339202442441605e-06,
+      "loss": 1.3927,
+      "step": 15850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3391606198113799e-06,
+      "loss": 0.0008,
+      "step": 15851
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3391187971811546e-06,
+      "loss": 0.0114,
+      "step": 15852
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3390769745509294e-06,
+      "loss": 0.0023,
+      "step": 15853
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3390351519207042e-06,
+      "loss": 0.0004,
+      "step": 15854
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338993329290479e-06,
+      "loss": 0.0007,
+      "step": 15855
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3389515066602538e-06,
+      "loss": 0.0012,
+      "step": 15856
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3389096840300288e-06,
+      "loss": 0.0006,
+      "step": 15857
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3388678613998034e-06,
+      "loss": 0.0062,
+      "step": 15858
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3388260387695781e-06,
+      "loss": 0.1085,
+      "step": 15859
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338784216139353e-06,
+      "loss": 0.1111,
+      "step": 15860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3387423935091277e-06,
+      "loss": 0.0016,
+      "step": 15861
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3387005708789025e-06,
+      "loss": 0.0005,
+      "step": 15862
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3386587482486773e-06,
+      "loss": 0.0138,
+      "step": 15863
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338616925618452e-06,
+      "loss": 0.0005,
+      "step": 15864
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3385751029882269e-06,
+      "loss": 0.0007,
+      "step": 15865
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3385332803580016e-06,
+      "loss": 0.1783,
+      "step": 15866
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3384914577277764e-06,
+      "loss": 0.003,
+      "step": 15867
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3384496350975512e-06,
+      "loss": 0.0005,
+      "step": 15868
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338407812467326e-06,
+      "loss": 0.0002,
+      "step": 15869
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3383659898371008e-06,
+      "loss": 3.3413,
+      "step": 15870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3383241672068758e-06,
+      "loss": 0.0012,
+      "step": 15871
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3382823445766504e-06,
+      "loss": 0.0056,
+      "step": 15872
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3382405219464252e-06,
+      "loss": 0.0031,
+      "step": 15873
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3381986993162e-06,
+      "loss": 0.0283,
+      "step": 15874
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3381568766859747e-06,
+      "loss": 0.0004,
+      "step": 15875
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3381150540557495e-06,
+      "loss": 0.0184,
+      "step": 15876
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3380732314255243e-06,
+      "loss": 0.0017,
+      "step": 15877
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.338031408795299e-06,
+      "loss": 0.0002,
+      "step": 15878
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3379895861650739e-06,
+      "loss": 0.0044,
+      "step": 15879
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3379477635348487e-06,
+      "loss": 0.0017,
+      "step": 15880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3379059409046234e-06,
+      "loss": 1.7142,
+      "step": 15881
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3378641182743982e-06,
+      "loss": 0.0012,
+      "step": 15882
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.337822295644173e-06,
+      "loss": 0.0047,
+      "step": 15883
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3377804730139478e-06,
+      "loss": 0.0007,
+      "step": 15884
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3377386503837224e-06,
+      "loss": 0.0071,
+      "step": 15885
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3376968277534974e-06,
+      "loss": 1.7806,
+      "step": 15886
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3376550051232722e-06,
+      "loss": 0.0014,
+      "step": 15887
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.337613182493047e-06,
+      "loss": 0.0058,
+      "step": 15888
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3375713598628217e-06,
+      "loss": 0.0024,
+      "step": 15889
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3375295372325965e-06,
+      "loss": 0.0002,
+      "step": 15890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3374877146023713e-06,
+      "loss": 0.0017,
+      "step": 15891
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.337445891972146e-06,
+      "loss": 0.0009,
+      "step": 15892
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3374040693419209e-06,
+      "loss": 0.0027,
+      "step": 15893
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3373622467116957e-06,
+      "loss": 0.0004,
+      "step": 15894
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3373204240814705e-06,
+      "loss": 0.0024,
+      "step": 15895
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3372786014512452e-06,
+      "loss": 0.0003,
+      "step": 15896
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.33723677882102e-06,
+      "loss": 0.0094,
+      "step": 15897
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3371949561907946e-06,
+      "loss": 0.0019,
+      "step": 15898
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3371531335605694e-06,
+      "loss": 0.002,
+      "step": 15899
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3371113109303442e-06,
+      "loss": 0.4389,
+      "step": 15900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3370694883001192e-06,
+      "loss": 0.0035,
+      "step": 15901
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.337027665669894e-06,
+      "loss": 0.0003,
+      "step": 15902
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3369858430396687e-06,
+      "loss": 0.0017,
+      "step": 15903
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3369440204094435e-06,
+      "loss": 0.0051,
+      "step": 15904
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3369021977792183e-06,
+      "loss": 0.0005,
+      "step": 15905
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.336860375148993e-06,
+      "loss": 0.0071,
+      "step": 15906
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3368185525187679e-06,
+      "loss": 1.0675,
+      "step": 15907
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3367767298885427e-06,
+      "loss": 0.0105,
+      "step": 15908
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3367349072583175e-06,
+      "loss": 0.0006,
+      "step": 15909
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3366930846280922e-06,
+      "loss": 0.001,
+      "step": 15910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.336651261997867e-06,
+      "loss": 0.0003,
+      "step": 15911
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3366094393676416e-06,
+      "loss": 0.0003,
+      "step": 15912
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3365676167374164e-06,
+      "loss": 0.0019,
+      "step": 15913
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3365257941071912e-06,
+      "loss": 0.1618,
+      "step": 15914
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3364839714769662e-06,
+      "loss": 0.0005,
+      "step": 15915
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.336442148846741e-06,
+      "loss": 0.703,
+      "step": 15916
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3364003262165158e-06,
+      "loss": 0.0007,
+      "step": 15917
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3363585035862905e-06,
+      "loss": 0.5448,
+      "step": 15918
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3363166809560653e-06,
+      "loss": 2.135,
+      "step": 15919
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3362748583258401e-06,
+      "loss": 0.0005,
+      "step": 15920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.336233035695615e-06,
+      "loss": 0.9302,
+      "step": 15921
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3361912130653897e-06,
+      "loss": 0.0004,
+      "step": 15922
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3361493904351645e-06,
+      "loss": 0.0054,
+      "step": 15923
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3361075678049393e-06,
+      "loss": 1.9666,
+      "step": 15924
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3360657451747138e-06,
+      "loss": 2.1468,
+      "step": 15925
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3360239225444886e-06,
+      "loss": 0.0006,
+      "step": 15926
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3359820999142634e-06,
+      "loss": 0.0003,
+      "step": 15927
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3359402772840382e-06,
+      "loss": 0.0007,
+      "step": 15928
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.335898454653813e-06,
+      "loss": 0.0005,
+      "step": 15929
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.335856632023588e-06,
+      "loss": 0.0046,
+      "step": 15930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3358148093933628e-06,
+      "loss": 0.0017,
+      "step": 15931
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3357729867631375e-06,
+      "loss": 0.0005,
+      "step": 15932
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3357311641329123e-06,
+      "loss": 0.0032,
+      "step": 15933
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3356893415026871e-06,
+      "loss": 0.0006,
+      "step": 15934
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.335647518872462e-06,
+      "loss": 0.0004,
+      "step": 15935
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3356056962422367e-06,
+      "loss": 0.0003,
+      "step": 15936
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3355638736120115e-06,
+      "loss": 1.8443,
+      "step": 15937
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3355220509817863e-06,
+      "loss": 0.0008,
+      "step": 15938
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3354802283515608e-06,
+      "loss": 1.1962,
+      "step": 15939
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3354384057213356e-06,
+      "loss": 0.0005,
+      "step": 15940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3353965830911104e-06,
+      "loss": 0.0008,
+      "step": 15941
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3353547604608852e-06,
+      "loss": 2.1432,
+      "step": 15942
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.33531293783066e-06,
+      "loss": 0.0012,
+      "step": 15943
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.335271115200435e-06,
+      "loss": 0.0005,
+      "step": 15944
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3352292925702098e-06,
+      "loss": 0.0056,
+      "step": 15945
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3351874699399846e-06,
+      "loss": 2.7143,
+      "step": 15946
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3351456473097593e-06,
+      "loss": 0.0009,
+      "step": 15947
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3351038246795341e-06,
+      "loss": 0.8554,
+      "step": 15948
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.335062002049309e-06,
+      "loss": 0.0055,
+      "step": 15949
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3350201794190837e-06,
+      "loss": 0.0016,
+      "step": 15950
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3349783567888585e-06,
+      "loss": 0.0007,
+      "step": 15951
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.334936534158633e-06,
+      "loss": 0.0002,
+      "step": 15952
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3348947115284078e-06,
+      "loss": 2.2161,
+      "step": 15953
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3348528888981826e-06,
+      "loss": 0.0009,
+      "step": 15954
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3348110662679574e-06,
+      "loss": 0.8234,
+      "step": 15955
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3347692436377322e-06,
+      "loss": 0.0295,
+      "step": 15956
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.334727421007507e-06,
+      "loss": 0.0008,
+      "step": 15957
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.9688547704840984,
+      "eval_loss": 0.349341481924057,
+      "eval_runtime": 512.3903,
+      "eval_samples_per_second": 62.287,
+      "eval_steps_per_second": 0.974,
+      "step": 15957
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 47871,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "total_flos": 1.535026485851376e+16,
+  "trial_name": null,
+  "trial_params": null
+}