diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,42300 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 7045,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.358490566037736e-08,
+      "loss": 1.3232,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.716981132075472e-08,
+      "loss": 1.3271,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.075471698113208e-08,
+      "loss": 1.3477,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.433962264150944e-08,
+      "loss": 1.334,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.179245283018868e-07,
+      "loss": 1.3379,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4150943396226417e-07,
+      "loss": 1.3525,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6509433962264153e-07,
+      "loss": 1.3887,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.886792452830189e-07,
+      "loss": 1.3379,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1226415094339622e-07,
+      "loss": 1.3311,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.358490566037736e-07,
+      "loss": 1.3672,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.59433962264151e-07,
+      "loss": 1.3203,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8301886792452833e-07,
+      "loss": 1.3379,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0660377358490567e-07,
+      "loss": 1.3027,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3018867924528305e-07,
+      "loss": 1.3564,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.537735849056604e-07,
+      "loss": 1.3711,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.773584905660378e-07,
+      "loss": 1.3555,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.009433962264151e-07,
+      "loss": 1.3701,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2452830188679244e-07,
+      "loss": 1.3154,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4811320754716983e-07,
+      "loss": 1.333,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.716981132075472e-07,
+      "loss": 1.3467,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.952830188679246e-07,
+      "loss": 1.2949,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.18867924528302e-07,
+      "loss": 1.3311,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.424528301886793e-07,
+      "loss": 1.3164,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.660377358490567e-07,
+      "loss": 1.3096,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.89622641509434e-07,
+      "loss": 1.3213,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.132075471698113e-07,
+      "loss": 1.3467,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.367924528301888e-07,
+      "loss": 1.3213,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.603773584905661e-07,
+      "loss": 1.3271,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.839622641509434e-07,
+      "loss": 1.3594,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.075471698113208e-07,
+      "loss": 1.2861,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.311320754716981e-07,
+      "loss": 1.3086,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.547169811320755e-07,
+      "loss": 1.333,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.783018867924529e-07,
+      "loss": 1.2344,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.018867924528302e-07,
+      "loss": 1.2988,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.254716981132076e-07,
+      "loss": 1.2959,
+      "step": 35
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.490566037735849e-07,
+      "loss": 1.3242,
+      "step": 36
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.726415094339623e-07,
+      "loss": 1.3135,
+      "step": 37
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.962264150943397e-07,
+      "loss": 1.3145,
+      "step": 38
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.19811320754717e-07,
+      "loss": 1.2529,
+      "step": 39
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.433962264150944e-07,
+      "loss": 1.2734,
+      "step": 40
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.669811320754719e-07,
+      "loss": 1.2686,
+      "step": 41
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.90566037735849e-07,
+      "loss": 1.2588,
+      "step": 42
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0141509433962265e-06,
+      "loss": 1.2676,
+      "step": 43
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.037735849056604e-06,
+      "loss": 1.249,
+      "step": 44
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0613207547169812e-06,
+      "loss": 1.2314,
+      "step": 45
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0849056603773587e-06,
+      "loss": 1.2705,
+      "step": 46
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1084905660377359e-06,
+      "loss": 1.2188,
+      "step": 47
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1320754716981133e-06,
+      "loss": 1.252,
+      "step": 48
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1556603773584908e-06,
+      "loss": 1.2383,
+      "step": 49
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.179245283018868e-06,
+      "loss": 1.2236,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2028301886792454e-06,
+      "loss": 1.2666,
+      "step": 51
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2264150943396227e-06,
+      "loss": 1.2314,
+      "step": 52
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.25e-06,
+      "loss": 1.2158,
+      "step": 53
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2735849056603775e-06,
+      "loss": 1.2344,
+      "step": 54
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2971698113207548e-06,
+      "loss": 1.2578,
+      "step": 55
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3207547169811322e-06,
+      "loss": 1.2168,
+      "step": 56
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3443396226415094e-06,
+      "loss": 1.2246,
+      "step": 57
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3679245283018869e-06,
+      "loss": 1.2295,
+      "step": 58
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3915094339622643e-06,
+      "loss": 1.2646,
+      "step": 59
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4150943396226415e-06,
+      "loss": 1.2422,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.438679245283019e-06,
+      "loss": 1.2354,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4622641509433962e-06,
+      "loss": 1.2217,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4858490566037737e-06,
+      "loss": 1.2334,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.509433962264151e-06,
+      "loss": 1.2275,
+      "step": 64
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5330188679245283e-06,
+      "loss": 1.2305,
+      "step": 65
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5566037735849058e-06,
+      "loss": 1.2324,
+      "step": 66
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.580188679245283e-06,
+      "loss": 1.1934,
+      "step": 67
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6037735849056604e-06,
+      "loss": 1.1729,
+      "step": 68
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6273584905660379e-06,
+      "loss": 1.1987,
+      "step": 69
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.650943396226415e-06,
+      "loss": 1.1855,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6745283018867925e-06,
+      "loss": 1.1992,
+      "step": 71
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6981132075471698e-06,
+      "loss": 1.209,
+      "step": 72
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7216981132075472e-06,
+      "loss": 1.2334,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7452830188679247e-06,
+      "loss": 1.2217,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7688679245283019e-06,
+      "loss": 1.2227,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7924528301886793e-06,
+      "loss": 1.21,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8160377358490566e-06,
+      "loss": 1.2188,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.839622641509434e-06,
+      "loss": 1.2383,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8632075471698114e-06,
+      "loss": 1.2119,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8867924528301889e-06,
+      "loss": 1.2168,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9103773584905665e-06,
+      "loss": 1.208,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9339622641509438e-06,
+      "loss": 1.2451,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.957547169811321e-06,
+      "loss": 1.1904,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.981132075471698e-06,
+      "loss": 1.1729,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.004716981132076e-06,
+      "loss": 1.2148,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.028301886792453e-06,
+      "loss": 1.1816,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0518867924528303e-06,
+      "loss": 1.2305,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.075471698113208e-06,
+      "loss": 1.2305,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.099056603773585e-06,
+      "loss": 1.2148,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1226415094339624e-06,
+      "loss": 1.2383,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.14622641509434e-06,
+      "loss": 1.1836,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1698113207547173e-06,
+      "loss": 1.2344,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1933962264150945e-06,
+      "loss": 1.1982,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2169811320754718e-06,
+      "loss": 1.2256,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2405660377358494e-06,
+      "loss": 1.2266,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2641509433962266e-06,
+      "loss": 1.1953,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.287735849056604e-06,
+      "loss": 1.2207,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3113207547169815e-06,
+      "loss": 1.1904,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3349056603773588e-06,
+      "loss": 1.168,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.358490566037736e-06,
+      "loss": 1.1689,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3820754716981136e-06,
+      "loss": 1.1807,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.405660377358491e-06,
+      "loss": 1.2002,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.429245283018868e-06,
+      "loss": 1.1953,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4528301886792453e-06,
+      "loss": 1.2168,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.476415094339623e-06,
+      "loss": 1.1855,
+      "step": 105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5e-06,
+      "loss": 1.2197,
+      "step": 106
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.523584905660378e-06,
+      "loss": 1.2285,
+      "step": 107
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.547169811320755e-06,
+      "loss": 1.2217,
+      "step": 108
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5707547169811327e-06,
+      "loss": 1.1982,
+      "step": 109
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5943396226415095e-06,
+      "loss": 1.1758,
+      "step": 110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.617924528301887e-06,
+      "loss": 1.2197,
+      "step": 111
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6415094339622644e-06,
+      "loss": 1.1943,
+      "step": 112
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.665094339622642e-06,
+      "loss": 1.2002,
+      "step": 113
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.688679245283019e-06,
+      "loss": 1.1953,
+      "step": 114
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7122641509433965e-06,
+      "loss": 1.1807,
+      "step": 115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7358490566037738e-06,
+      "loss": 1.208,
+      "step": 116
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7594339622641514e-06,
+      "loss": 1.2354,
+      "step": 117
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7830188679245286e-06,
+      "loss": 1.2061,
+      "step": 118
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.8066037735849063e-06,
+      "loss": 1.2041,
+      "step": 119
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.830188679245283e-06,
+      "loss": 1.1592,
+      "step": 120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.8537735849056608e-06,
+      "loss": 1.1973,
+      "step": 121
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.877358490566038e-06,
+      "loss": 1.1768,
+      "step": 122
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9009433962264156e-06,
+      "loss": 1.207,
+      "step": 123
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9245283018867924e-06,
+      "loss": 1.1787,
+      "step": 124
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.94811320754717e-06,
+      "loss": 1.1953,
+      "step": 125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9716981132075473e-06,
+      "loss": 1.2002,
+      "step": 126
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995283018867925e-06,
+      "loss": 1.2012,
+      "step": 127
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.018867924528302e-06,
+      "loss": 1.2109,
+      "step": 128
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.04245283018868e-06,
+      "loss": 1.2363,
+      "step": 129
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0660377358490567e-06,
+      "loss": 1.1885,
+      "step": 130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0896226415094343e-06,
+      "loss": 1.1523,
+      "step": 131
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1132075471698115e-06,
+      "loss": 1.2217,
+      "step": 132
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.136792452830189e-06,
+      "loss": 1.1865,
+      "step": 133
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.160377358490566e-06,
+      "loss": 1.1992,
+      "step": 134
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1839622641509436e-06,
+      "loss": 1.1963,
+      "step": 135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.207547169811321e-06,
+      "loss": 1.1934,
+      "step": 136
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2311320754716985e-06,
+      "loss": 1.1895,
+      "step": 137
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2547169811320758e-06,
+      "loss": 1.1768,
+      "step": 138
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2783018867924534e-06,
+      "loss": 1.2002,
+      "step": 139
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.30188679245283e-06,
+      "loss": 1.1748,
+      "step": 140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.325471698113208e-06,
+      "loss": 1.2344,
+      "step": 141
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.349056603773585e-06,
+      "loss": 1.1924,
+      "step": 142
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.3726415094339627e-06,
+      "loss": 1.1523,
+      "step": 143
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.3962264150943395e-06,
+      "loss": 1.2168,
+      "step": 144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.419811320754717e-06,
+      "loss": 1.1748,
+      "step": 145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.4433962264150944e-06,
+      "loss": 1.1719,
+      "step": 146
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.466981132075472e-06,
+      "loss": 1.1768,
+      "step": 147
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.4905660377358493e-06,
+      "loss": 1.1826,
+      "step": 148
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.514150943396227e-06,
+      "loss": 1.2031,
+      "step": 149
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.5377358490566038e-06,
+      "loss": 1.165,
+      "step": 150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.5613207547169814e-06,
+      "loss": 1.1826,
+      "step": 151
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.5849056603773586e-06,
+      "loss": 1.2041,
+      "step": 152
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.6084905660377363e-06,
+      "loss": 1.1689,
+      "step": 153
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.632075471698113e-06,
+      "loss": 1.1934,
+      "step": 154
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.6556603773584908e-06,
+      "loss": 1.1709,
+      "step": 155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.679245283018868e-06,
+      "loss": 1.1885,
+      "step": 156
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7028301886792456e-06,
+      "loss": 1.2148,
+      "step": 157
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.726415094339623e-06,
+      "loss": 1.2354,
+      "step": 158
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 1.21,
+      "step": 159
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7735849056603777e-06,
+      "loss": 1.1914,
+      "step": 160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.797169811320755e-06,
+      "loss": 1.1582,
+      "step": 161
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.820754716981133e-06,
+      "loss": 1.1846,
+      "step": 162
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.8443396226415094e-06,
+      "loss": 1.2158,
+      "step": 163
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.8679245283018875e-06,
+      "loss": 1.2354,
+      "step": 164
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.891509433962265e-06,
+      "loss": 1.2002,
+      "step": 165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.915094339622642e-06,
+      "loss": 1.1445,
+      "step": 166
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.938679245283019e-06,
+      "loss": 1.1572,
+      "step": 167
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.962264150943396e-06,
+      "loss": 1.1816,
+      "step": 168
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.985849056603774e-06,
+      "loss": 1.2197,
+      "step": 169
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.009433962264152e-06,
+      "loss": 1.1787,
+      "step": 170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.033018867924529e-06,
+      "loss": 1.1826,
+      "step": 171
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.056603773584906e-06,
+      "loss": 1.166,
+      "step": 172
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.080188679245283e-06,
+      "loss": 1.1699,
+      "step": 173
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.103773584905661e-06,
+      "loss": 1.167,
+      "step": 174
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.127358490566038e-06,
+      "loss": 1.2236,
+      "step": 175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.150943396226416e-06,
+      "loss": 1.1631,
+      "step": 176
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.174528301886792e-06,
+      "loss": 1.1963,
+      "step": 177
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.19811320754717e-06,
+      "loss": 1.207,
+      "step": 178
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.221698113207548e-06,
+      "loss": 1.1924,
+      "step": 179
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.245283018867925e-06,
+      "loss": 1.1816,
+      "step": 180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.268867924528302e-06,
+      "loss": 1.165,
+      "step": 181
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.29245283018868e-06,
+      "loss": 1.1875,
+      "step": 182
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.3160377358490565e-06,
+      "loss": 1.1846,
+      "step": 183
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.339622641509435e-06,
+      "loss": 1.2217,
+      "step": 184
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.363207547169812e-06,
+      "loss": 1.1865,
+      "step": 185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.386792452830189e-06,
+      "loss": 1.208,
+      "step": 186
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.410377358490566e-06,
+      "loss": 1.1611,
+      "step": 187
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.4339622641509435e-06,
+      "loss": 1.2227,
+      "step": 188
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.457547169811321e-06,
+      "loss": 1.1455,
+      "step": 189
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.481132075471699e-06,
+      "loss": 1.126,
+      "step": 190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.504716981132076e-06,
+      "loss": 1.207,
+      "step": 191
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.528301886792453e-06,
+      "loss": 1.209,
+      "step": 192
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.5518867924528305e-06,
+      "loss": 1.1689,
+      "step": 193
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.575471698113208e-06,
+      "loss": 1.1963,
+      "step": 194
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.599056603773585e-06,
+      "loss": 1.1943,
+      "step": 195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.622641509433963e-06,
+      "loss": 1.1523,
+      "step": 196
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.6462264150943394e-06,
+      "loss": 1.1592,
+      "step": 197
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.6698113207547175e-06,
+      "loss": 1.1699,
+      "step": 198
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.693396226415095e-06,
+      "loss": 1.1758,
+      "step": 199
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.716981132075472e-06,
+      "loss": 1.1777,
+      "step": 200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.740566037735849e-06,
+      "loss": 1.2109,
+      "step": 201
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.764150943396227e-06,
+      "loss": 1.2061,
+      "step": 202
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.787735849056604e-06,
+      "loss": 1.1729,
+      "step": 203
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.811320754716982e-06,
+      "loss": 1.2227,
+      "step": 204
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.834905660377359e-06,
+      "loss": 1.209,
+      "step": 205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.858490566037736e-06,
+      "loss": 1.1865,
+      "step": 206
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.882075471698113e-06,
+      "loss": 1.21,
+      "step": 207
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.905660377358491e-06,
+      "loss": 1.1631,
+      "step": 208
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.929245283018868e-06,
+      "loss": 1.1406,
+      "step": 209
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.952830188679246e-06,
+      "loss": 1.168,
+      "step": 210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.976415094339623e-06,
+      "loss": 1.1807,
+      "step": 211
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 5e-06,
+      "loss": 1.1797,
+      "step": 212
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999999735767089e-06,
+      "loss": 1.1309,
+      "step": 213
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999998943068413e-06,
+      "loss": 1.1699,
+      "step": 214
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999997621904137e-06,
+      "loss": 1.1943,
+      "step": 215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999995772274542e-06,
+      "loss": 1.1924,
+      "step": 216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999993394180018e-06,
+      "loss": 1.1885,
+      "step": 217
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9999904876210694e-06,
+      "loss": 1.1836,
+      "step": 218
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.99998705259831e-06,
+      "loss": 1.1289,
+      "step": 219
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999983089112464e-06,
+      "loss": 1.1572,
+      "step": 220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999978597164372e-06,
+      "loss": 1.2119,
+      "step": 221
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999973576754983e-06,
+      "loss": 1.1729,
+      "step": 222
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999968027885356e-06,
+      "loss": 1.1914,
+      "step": 223
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999961950556666e-06,
+      "loss": 1.1709,
+      "step": 224
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999955344770198e-06,
+      "loss": 1.1777,
+      "step": 225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999948210527345e-06,
+      "loss": 1.1963,
+      "step": 226
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9999405478296205e-06,
+      "loss": 1.1533,
+      "step": 227
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.99993235667864e-06,
+      "loss": 1.2158,
+      "step": 228
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999923637076137e-06,
+      "loss": 1.1758,
+      "step": 229
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999914389023953e-06,
+      "loss": 1.1934,
+      "step": 230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999904612524045e-06,
+      "loss": 1.1699,
+      "step": 231
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999894307578479e-06,
+      "loss": 1.1816,
+      "step": 232
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999883474189432e-06,
+      "loss": 1.1982,
+      "step": 233
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999872112359195e-06,
+      "loss": 1.1582,
+      "step": 234
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.99986022209017e-06,
+      "loss": 1.2051,
+      "step": 235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9998478033848704e-06,
+      "loss": 1.1768,
+      "step": 236
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999834856245922e-06,
+      "loss": 1.1602,
+      "step": 237
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999821380676059e-06,
+      "loss": 1.1738,
+      "step": 238
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999807376678133e-06,
+      "loss": 1.1846,
+      "step": 239
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9997928442551025e-06,
+      "loss": 1.1523,
+      "step": 240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.99977778341004e-06,
+      "loss": 1.166,
+      "step": 241
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999762194146128e-06,
+      "loss": 1.1787,
+      "step": 242
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999746076466665e-06,
+      "loss": 1.1729,
+      "step": 243
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999729430375054e-06,
+      "loss": 1.2139,
+      "step": 244
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999712255874817e-06,
+      "loss": 1.2129,
+      "step": 245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999694552969583e-06,
+      "loss": 1.1934,
+      "step": 246
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9996763216630935e-06,
+      "loss": 1.1011,
+      "step": 247
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999657561959205e-06,
+      "loss": 1.1504,
+      "step": 248
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9996382738618795e-06,
+      "loss": 1.1665,
+      "step": 249
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9996184573751965e-06,
+      "loss": 1.1797,
+      "step": 250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9995981125033446e-06,
+      "loss": 1.1611,
+      "step": 251
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999577239250625e-06,
+      "loss": 1.1279,
+      "step": 252
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999555837621448e-06,
+      "loss": 1.1963,
+      "step": 253
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999533907620339e-06,
+      "loss": 1.1309,
+      "step": 254
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999511449251934e-06,
+      "loss": 1.1484,
+      "step": 255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99948846252098e-06,
+      "loss": 1.1621,
+      "step": 256
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999464947432337e-06,
+      "loss": 1.1621,
+      "step": 257
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999440903990973e-06,
+      "loss": 1.1299,
+      "step": 258
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999416332201973e-06,
+      "loss": 1.1328,
+      "step": 259
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999391232070531e-06,
+      "loss": 1.1895,
+      "step": 260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999365603601951e-06,
+      "loss": 1.126,
+      "step": 261
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999339446801653e-06,
+      "loss": 1.1621,
+      "step": 262
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999312761675165e-06,
+      "loss": 1.1846,
+      "step": 263
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9992855482281265e-06,
+      "loss": 1.2119,
+      "step": 264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999257806466292e-06,
+      "loss": 1.1621,
+      "step": 265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999229536395525e-06,
+      "loss": 1.2021,
+      "step": 266
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999200738021802e-06,
+      "loss": 1.1455,
+      "step": 267
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9991714113512104e-06,
+      "loss": 1.1943,
+      "step": 268
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999141556389948e-06,
+      "loss": 1.2002,
+      "step": 269
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9991111731443275e-06,
+      "loss": 1.1865,
+      "step": 270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999080261620771e-06,
+      "loss": 1.1787,
+      "step": 271
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999048821825813e-06,
+      "loss": 1.1953,
+      "step": 272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999016853766098e-06,
+      "loss": 1.1807,
+      "step": 273
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998984357448385e-06,
+      "loss": 1.2012,
+      "step": 274
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998951332879544e-06,
+      "loss": 1.1689,
+      "step": 275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998917780066554e-06,
+      "loss": 1.21,
+      "step": 276
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998883699016509e-06,
+      "loss": 1.1768,
+      "step": 277
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998849089736612e-06,
+      "loss": 1.1807,
+      "step": 278
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99881395223418e-06,
+      "loss": 1.1777,
+      "step": 279
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9987782865166415e-06,
+      "loss": 1.1865,
+      "step": 280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998742092591533e-06,
+      "loss": 1.1562,
+      "step": 281
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9987053704665075e-06,
+      "loss": 1.1787,
+      "step": 282
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998668120149327e-06,
+      "loss": 1.1768,
+      "step": 283
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998630341647866e-06,
+      "loss": 1.1621,
+      "step": 284
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99859203497011e-06,
+      "loss": 1.1934,
+      "step": 285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998553200124157e-06,
+      "loss": 1.1426,
+      "step": 286
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998513837118214e-06,
+      "loss": 1.1758,
+      "step": 287
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998473945960605e-06,
+      "loss": 1.1797,
+      "step": 288
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998433526659761e-06,
+      "loss": 1.1748,
+      "step": 289
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998392579224226e-06,
+      "loss": 1.1396,
+      "step": 290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9983511036626555e-06,
+      "loss": 1.2168,
+      "step": 291
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998309099983817e-06,
+      "loss": 1.1104,
+      "step": 292
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99826656819659e-06,
+      "loss": 1.1953,
+      "step": 293
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998223508309965e-06,
+      "loss": 1.1504,
+      "step": 294
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998179920333044e-06,
+      "loss": 1.1523,
+      "step": 295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99813580427504e-06,
+      "loss": 1.1973,
+      "step": 296
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998091160145281e-06,
+      "loss": 1.1943,
+      "step": 297
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998045987953202e-06,
+      "loss": 1.2119,
+      "step": 298
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998000287708353e-06,
+      "loss": 1.1523,
+      "step": 299
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997954059420393e-06,
+      "loss": 1.1602,
+      "step": 300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997907303099095e-06,
+      "loss": 1.1768,
+      "step": 301
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9978600187543435e-06,
+      "loss": 1.1279,
+      "step": 302
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997812206396132e-06,
+      "loss": 1.1367,
+      "step": 303
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997763866034568e-06,
+      "loss": 1.1787,
+      "step": 304
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9977149976798715e-06,
+      "loss": 1.1475,
+      "step": 305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9976656013423695e-06,
+      "loss": 1.1143,
+      "step": 306
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997615677032507e-06,
+      "loss": 1.1963,
+      "step": 307
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997565224760835e-06,
+      "loss": 1.1699,
+      "step": 308
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997514244538019e-06,
+      "loss": 1.165,
+      "step": 309
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997462736374836e-06,
+      "loss": 1.1426,
+      "step": 310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997410700282174e-06,
+      "loss": 1.1504,
+      "step": 311
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997358136271032e-06,
+      "loss": 1.168,
+      "step": 312
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997305044352523e-06,
+      "loss": 1.1553,
+      "step": 313
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997251424537868e-06,
+      "loss": 1.1831,
+      "step": 314
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997197276838403e-06,
+      "loss": 1.126,
+      "step": 315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997142601265573e-06,
+      "loss": 1.1855,
+      "step": 316
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997087397830936e-06,
+      "loss": 1.1426,
+      "step": 317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.997031666546161e-06,
+      "loss": 1.167,
+      "step": 318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9969754074230294e-06,
+      "loss": 1.1943,
+      "step": 319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9969186204734336e-06,
+      "loss": 1.1816,
+      "step": 320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9968613057093775e-06,
+      "loss": 1.1484,
+      "step": 321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996803463142976e-06,
+      "loss": 1.1514,
+      "step": 322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996745092786457e-06,
+      "loss": 1.2002,
+      "step": 323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996686194652158e-06,
+      "loss": 1.1738,
+      "step": 324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9966267687525314e-06,
+      "loss": 1.1689,
+      "step": 325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996566815100137e-06,
+      "loss": 1.1406,
+      "step": 326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99650633370765e-06,
+      "loss": 1.1475,
+      "step": 327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996445324587853e-06,
+      "loss": 1.1855,
+      "step": 328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996383787753645e-06,
+      "loss": 1.166,
+      "step": 329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996321723218032e-06,
+      "loss": 1.1465,
+      "step": 330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996259130994135e-06,
+      "loss": 1.1494,
+      "step": 331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996196011095184e-06,
+      "loss": 1.1514,
+      "step": 332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996132363534524e-06,
+      "loss": 1.1416,
+      "step": 333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996068188325606e-06,
+      "loss": 1.1641,
+      "step": 334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996003485481996e-06,
+      "loss": 1.1914,
+      "step": 335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9959382550173755e-06,
+      "loss": 1.1709,
+      "step": 336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9958724969455285e-06,
+      "loss": 1.1689,
+      "step": 337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995806211280357e-06,
+      "loss": 1.1504,
+      "step": 338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995739398035874e-06,
+      "loss": 1.1445,
+      "step": 339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995672057226202e-06,
+      "loss": 1.1045,
+      "step": 340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995604188865576e-06,
+      "loss": 1.1553,
+      "step": 341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995535792968342e-06,
+      "loss": 1.1641,
+      "step": 342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9954668695489585e-06,
+      "loss": 1.165,
+      "step": 343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995397418621994e-06,
+      "loss": 1.1514,
+      "step": 344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995327440202132e-06,
+      "loss": 1.1475,
+      "step": 345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995256934304162e-06,
+      "loss": 1.1318,
+      "step": 346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99518590094299e-06,
+      "loss": 1.1777,
+      "step": 347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995114340133631e-06,
+      "loss": 1.2002,
+      "step": 348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995042251891211e-06,
+      "loss": 1.1279,
+      "step": 349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99496963623097e-06,
+      "loss": 1.125,
+      "step": 350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9948964931682564e-06,
+      "loss": 1.1133,
+      "step": 351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994822822718532e-06,
+      "loss": 1.1494,
+      "step": 352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994748624897371e-06,
+      "loss": 1.2168,
+      "step": 353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994673899720457e-06,
+      "loss": 1.1738,
+      "step": 354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9945986472035855e-06,
+      "loss": 1.1357,
+      "step": 355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994522867362664e-06,
+      "loss": 1.1846,
+      "step": 356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994446560213712e-06,
+      "loss": 1.1377,
+      "step": 357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9943697257728584e-06,
+      "loss": 1.1719,
+      "step": 358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994292364056346e-06,
+      "loss": 1.1729,
+      "step": 359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994214475080529e-06,
+      "loss": 1.1572,
+      "step": 360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994136058861869e-06,
+      "loss": 1.1621,
+      "step": 361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994057115416946e-06,
+      "loss": 1.166,
+      "step": 362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993977644762445e-06,
+      "loss": 1.1729,
+      "step": 363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993897646915165e-06,
+      "loss": 1.1289,
+      "step": 364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993817121892017e-06,
+      "loss": 1.1484,
+      "step": 365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993736069710023e-06,
+      "loss": 1.1152,
+      "step": 366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993654490386317e-06,
+      "loss": 1.1758,
+      "step": 367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993572383938143e-06,
+      "loss": 1.125,
+      "step": 368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993489750382856e-06,
+      "loss": 1.1855,
+      "step": 369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993406589737926e-06,
+      "loss": 1.1328,
+      "step": 370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99332290202093e-06,
+      "loss": 1.1943,
+      "step": 371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99323868724956e-06,
+      "loss": 1.1621,
+      "step": 372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993153945441617e-06,
+      "loss": 1.1592,
+      "step": 373
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993068676615014e-06,
+      "loss": 1.1797,
+      "step": 374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992982880787775e-06,
+      "loss": 1.1523,
+      "step": 375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992896557978039e-06,
+      "loss": 1.1562,
+      "step": 376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99280970820405e-06,
+      "loss": 1.123,
+      "step": 377
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99272233148417e-06,
+      "loss": 1.1885,
+      "step": 378
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992634427836867e-06,
+      "loss": 1.1758,
+      "step": 379
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992545997280723e-06,
+      "loss": 1.1885,
+      "step": 380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992457039834431e-06,
+      "loss": 1.1523,
+      "step": 381
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992367555516796e-06,
+      "loss": 1.1699,
+      "step": 382
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992277544346734e-06,
+      "loss": 1.1709,
+      "step": 383
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992187006343271e-06,
+      "loss": 1.1387,
+      "step": 384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992095941525546e-06,
+      "loss": 1.1504,
+      "step": 385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9920043499128095e-06,
+      "loss": 1.1465,
+      "step": 386
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.991912231524421e-06,
+      "loss": 1.1631,
+      "step": 387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991819586379856e-06,
+      "loss": 1.1416,
+      "step": 388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991726414498695e-06,
+      "loss": 1.1816,
+      "step": 389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9916327159006366e-06,
+      "loss": 1.1719,
+      "step": 390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991538490605485e-06,
+      "loss": 1.1826,
+      "step": 391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991443738633157e-06,
+      "loss": 1.1992,
+      "step": 392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9913484600036854e-06,
+      "loss": 1.1836,
+      "step": 393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9912526547372096e-06,
+      "loss": 1.1211,
+      "step": 394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.99115632285398e-06,
+      "loss": 1.1768,
+      "step": 395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991059464374361e-06,
+      "loss": 1.127,
+      "step": 396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990962079318828e-06,
+      "loss": 1.1484,
+      "step": 397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990864167707965e-06,
+      "loss": 1.209,
+      "step": 398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.99076572956247e-06,
+      "loss": 1.1475,
+      "step": 399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990666764903152e-06,
+      "loss": 1.1729,
+      "step": 400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.99056727375093e-06,
+      "loss": 1.208,
+      "step": 401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990467256126835e-06,
+      "loss": 1.1943,
+      "step": 402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9903667120520104e-06,
+      "loss": 1.1768,
+      "step": 403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990265641547709e-06,
+      "loss": 1.1436,
+      "step": 404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990164044635296e-06,
+      "loss": 1.1611,
+      "step": 405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990061921336248e-06,
+      "loss": 1.1963,
+      "step": 406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989959271672151e-06,
+      "loss": 1.165,
+      "step": 407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989856095664706e-06,
+      "loss": 1.1797,
+      "step": 408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.98975239333572e-06,
+      "loss": 1.1553,
+      "step": 409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989648164707118e-06,
+      "loss": 1.1885,
+      "step": 410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989543409800929e-06,
+      "loss": 1.1353,
+      "step": 411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989438128639299e-06,
+      "loss": 1.1748,
+      "step": 412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989332321244482e-06,
+      "loss": 1.2236,
+      "step": 413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989225987638844e-06,
+      "loss": 1.1406,
+      "step": 414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9891191278448635e-06,
+      "loss": 1.1729,
+      "step": 415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989011741885128e-06,
+      "loss": 1.1895,
+      "step": 416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988903829782339e-06,
+      "loss": 1.165,
+      "step": 417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988795391559307e-06,
+      "loss": 1.1611,
+      "step": 418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988686427238953e-06,
+      "loss": 1.1611,
+      "step": 419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9885769368443124e-06,
+      "loss": 1.1113,
+      "step": 420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988466920398528e-06,
+      "loss": 1.1641,
+      "step": 421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9883563779248575e-06,
+      "loss": 1.125,
+      "step": 422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988245309446669e-06,
+      "loss": 1.1533,
+      "step": 423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988133714987438e-06,
+      "loss": 1.1162,
+      "step": 424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988021594570756e-06,
+      "loss": 1.2109,
+      "step": 425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9879089482203226e-06,
+      "loss": 1.1387,
+      "step": 426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9877957759599504e-06,
+      "loss": 1.1738,
+      "step": 427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987682077813562e-06,
+      "loss": 1.1299,
+      "step": 428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987567853805193e-06,
+      "loss": 1.1768,
+      "step": 429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987453103958988e-06,
+      "loss": 1.1514,
+      "step": 430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987337828299203e-06,
+      "loss": 1.1953,
+      "step": 431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987222026850205e-06,
+      "loss": 1.1543,
+      "step": 432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987105699636474e-06,
+      "loss": 1.1787,
+      "step": 433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986988846682601e-06,
+      "loss": 1.1611,
+      "step": 434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986871468013285e-06,
+      "loss": 1.1826,
+      "step": 435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986753563653339e-06,
+      "loss": 1.1855,
+      "step": 436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986635133627687e-06,
+      "loss": 1.1387,
+      "step": 437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986516177961364e-06,
+      "loss": 1.1914,
+      "step": 438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986396696679514e-06,
+      "loss": 1.1836,
+      "step": 439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986276689807394e-06,
+      "loss": 1.1895,
+      "step": 440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986156157370372e-06,
+      "loss": 1.1504,
+      "step": 441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986035099393928e-06,
+      "loss": 1.127,
+      "step": 442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.98591351590365e-06,
+      "loss": 1.1592,
+      "step": 443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985791406925241e-06,
+      "loss": 1.1523,
+      "step": 444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985668772484512e-06,
+      "loss": 1.1172,
+      "step": 445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985545612607387e-06,
+      "loss": 1.1396,
+      "step": 446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9854219273199e-06,
+      "loss": 1.1719,
+      "step": 447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9852977166481955e-06,
+      "loss": 1.1406,
+      "step": 448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985172980618531e-06,
+      "loss": 1.209,
+      "step": 449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985047719257274e-06,
+      "loss": 1.1445,
+      "step": 450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.984921932590903e-06,
+      "loss": 1.0962,
+      "step": 451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9847956206460065e-06,
+      "loss": 1.1895,
+      "step": 452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9846687834492864e-06,
+      "loss": 1.1611,
+      "step": 453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9845414210275545e-06,
+      "loss": 1.1099,
+      "step": 454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.984413533407733e-06,
+      "loss": 1.168,
+      "step": 455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.984285120616854e-06,
+      "loss": 1.1455,
+      "step": 456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.984156182682065e-06,
+      "loss": 1.1787,
+      "step": 457
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.98402671963062e-06,
+      "loss": 1.1699,
+      "step": 458
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9838967314898865e-06,
+      "loss": 1.1729,
+      "step": 459
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983766218287341e-06,
+      "loss": 1.1855,
+      "step": 460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983635180050573e-06,
+      "loss": 1.1836,
+      "step": 461
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983503616807283e-06,
+      "loss": 1.1333,
+      "step": 462
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983371528585281e-06,
+      "loss": 1.1787,
+      "step": 463
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983238915412488e-06,
+      "loss": 1.1606,
+      "step": 464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983105777316938e-06,
+      "loss": 1.1592,
+      "step": 465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9829721143267724e-06,
+      "loss": 1.2002,
+      "step": 466
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982837926470248e-06,
+      "loss": 1.1465,
+      "step": 467
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982703213775729e-06,
+      "loss": 1.1162,
+      "step": 468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982567976271692e-06,
+      "loss": 1.1582,
+      "step": 469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982432213986725e-06,
+      "loss": 1.1641,
+      "step": 470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982295926949526e-06,
+      "loss": 1.1855,
+      "step": 471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982159115188904e-06,
+      "loss": 1.1602,
+      "step": 472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982021778733779e-06,
+      "loss": 1.1787,
+      "step": 473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981883917613182e-06,
+      "loss": 1.1338,
+      "step": 474
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981745531856255e-06,
+      "loss": 1.1465,
+      "step": 475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981606621492251e-06,
+      "loss": 1.1514,
+      "step": 476
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981467186550534e-06,
+      "loss": 1.1494,
+      "step": 477
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981327227060579e-06,
+      "loss": 1.1641,
+      "step": 478
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.98118674305197e-06,
+      "loss": 1.1914,
+      "step": 479
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981045734554405e-06,
+      "loss": 1.1177,
+      "step": 480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.98090420159769e-06,
+      "loss": 1.1572,
+      "step": 481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980762144211744e-06,
+      "loss": 1.1807,
+      "step": 482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980619562426596e-06,
+      "loss": 1.1572,
+      "step": 483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980476456272386e-06,
+      "loss": 1.1338,
+      "step": 484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980332825779364e-06,
+      "loss": 1.1484,
+      "step": 485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980188670977891e-06,
+      "loss": 1.1553,
+      "step": 486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980043991898441e-06,
+      "loss": 1.165,
+      "step": 487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979898788571595e-06,
+      "loss": 1.1758,
+      "step": 488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9797530610280495e-06,
+      "loss": 1.1582,
+      "step": 489
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979606809298608e-06,
+      "loss": 1.1064,
+      "step": 490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979460033414186e-06,
+      "loss": 1.1934,
+      "step": 491
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979312733405811e-06,
+      "loss": 1.1426,
+      "step": 492
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979164909304619e-06,
+      "loss": 1.1816,
+      "step": 493
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979016561141857e-06,
+      "loss": 1.1465,
+      "step": 494
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9788676889488865e-06,
+      "loss": 1.1514,
+      "step": 495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.978718292757176e-06,
+      "loss": 1.167,
+      "step": 496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.978568372598304e-06,
+      "loss": 1.1875,
+      "step": 497
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9784179285039646e-06,
+      "loss": 1.1504,
+      "step": 498
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.978266960505957e-06,
+      "loss": 1.209,
+      "step": 499
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.978115468636195e-06,
+      "loss": 1.1562,
+      "step": 500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.977963452926703e-06,
+      "loss": 1.1729,
+      "step": 501
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9778109134096125e-06,
+      "loss": 1.1201,
+      "step": 502
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.97765785011717e-06,
+      "loss": 1.1426,
+      "step": 503
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.97750426308173e-06,
+      "loss": 1.1992,
+      "step": 504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.97735015233576e-06,
+      "loss": 1.1206,
+      "step": 505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.977195517911836e-06,
+      "loss": 1.1494,
+      "step": 506
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9770403598426465e-06,
+      "loss": 1.168,
+      "step": 507
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976884678160988e-06,
+      "loss": 1.1768,
+      "step": 508
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976728472899771e-06,
+      "loss": 1.1113,
+      "step": 509
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976571744092015e-06,
+      "loss": 1.1582,
+      "step": 510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976414491770849e-06,
+      "loss": 1.1572,
+      "step": 511
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9762567159695155e-06,
+      "loss": 1.1699,
+      "step": 512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976098416721366e-06,
+      "loss": 1.1553,
+      "step": 513
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.975939594059862e-06,
+      "loss": 1.1553,
+      "step": 514
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.975780248018578e-06,
+      "loss": 1.1387,
+      "step": 515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.975620378631195e-06,
+      "loss": 1.1611,
+      "step": 516
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.97545998593151e-06,
+      "loss": 1.1875,
+      "step": 517
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.975299069953426e-06,
+      "loss": 1.1709,
+      "step": 518
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9751376307309585e-06,
+      "loss": 1.1621,
+      "step": 519
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9749756682982344e-06,
+      "loss": 1.1797,
+      "step": 520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.97481318268949e-06,
+      "loss": 1.1475,
+      "step": 521
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.974650173939072e-06,
+      "loss": 1.1719,
+      "step": 522
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.974486642081439e-06,
+      "loss": 1.1377,
+      "step": 523
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.974322587151159e-06,
+      "loss": 1.1523,
+      "step": 524
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.974158009182912e-06,
+      "loss": 1.1738,
+      "step": 525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.973992908211486e-06,
+      "loss": 1.1689,
+      "step": 526
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.973827284271781e-06,
+      "loss": 1.1416,
+      "step": 527
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.973661137398809e-06,
+      "loss": 1.1416,
+      "step": 528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.97349446762769e-06,
+      "loss": 1.1875,
+      "step": 529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.973327274993657e-06,
+      "loss": 1.1475,
+      "step": 530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.973159559532051e-06,
+      "loss": 1.1875,
+      "step": 531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.972991321278325e-06,
+      "loss": 1.1602,
+      "step": 532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9728225602680425e-06,
+      "loss": 1.1699,
+      "step": 533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.972653276536878e-06,
+      "loss": 1.1787,
+      "step": 534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.972483470120614e-06,
+      "loss": 1.166,
+      "step": 535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.972313141055146e-06,
+      "loss": 1.1177,
+      "step": 536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9721422893764805e-06,
+      "loss": 1.1895,
+      "step": 537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9719709151207315e-06,
+      "loss": 1.1514,
+      "step": 538
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9717990183241265e-06,
+      "loss": 1.168,
+      "step": 539
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971626599023002e-06,
+      "loss": 1.1475,
+      "step": 540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971453657253803e-06,
+      "loss": 1.1641,
+      "step": 541
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971280193053089e-06,
+      "loss": 1.1543,
+      "step": 542
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971106206457529e-06,
+      "loss": 1.1992,
+      "step": 543
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970931697503899e-06,
+      "loss": 1.1455,
+      "step": 544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970756666229089e-06,
+      "loss": 1.1445,
+      "step": 545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9705811126700975e-06,
+      "loss": 1.1514,
+      "step": 546
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970405036864035e-06,
+      "loss": 1.165,
+      "step": 547
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970228438848122e-06,
+      "loss": 1.1943,
+      "step": 548
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970051318659687e-06,
+      "loss": 1.1152,
+      "step": 549
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969873676336171e-06,
+      "loss": 1.1162,
+      "step": 550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969695511915127e-06,
+      "loss": 1.1826,
+      "step": 551
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969516825434215e-06,
+      "loss": 1.1543,
+      "step": 552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969337616931208e-06,
+      "loss": 1.1943,
+      "step": 553
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969157886443988e-06,
+      "loss": 1.0889,
+      "step": 554
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9689776340105466e-06,
+      "loss": 1.1572,
+      "step": 555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.968796859668987e-06,
+      "loss": 1.1294,
+      "step": 556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.968615563457523e-06,
+      "loss": 1.1416,
+      "step": 557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.968433745414478e-06,
+      "loss": 1.1387,
+      "step": 558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.968251405578286e-06,
+      "loss": 1.1709,
+      "step": 559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9680685439874895e-06,
+      "loss": 1.1465,
+      "step": 560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.967885160680746e-06,
+      "loss": 1.1631,
+      "step": 561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9677012556968175e-06,
+      "loss": 1.1772,
+      "step": 562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.96751682907458e-06,
+      "loss": 1.1436,
+      "step": 563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.967331880853019e-06,
+      "loss": 1.1582,
+      "step": 564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9671464110712306e-06,
+      "loss": 1.1514,
+      "step": 565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.966960419768419e-06,
+      "loss": 1.1152,
+      "step": 566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9667739069839e-06,
+      "loss": 1.168,
+      "step": 567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9665868727571024e-06,
+      "loss": 1.1025,
+      "step": 568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.966399317127561e-06,
+      "loss": 1.168,
+      "step": 569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.966211240134922e-06,
+      "loss": 1.1758,
+      "step": 570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9660226418189435e-06,
+      "loss": 1.165,
+      "step": 571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.965833522219491e-06,
+      "loss": 1.1816,
+      "step": 572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.965643881376544e-06,
+      "loss": 1.1416,
+      "step": 573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.965453719330189e-06,
+      "loss": 1.1787,
+      "step": 574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9652630361206224e-06,
+      "loss": 1.1133,
+      "step": 575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.965071831788153e-06,
+      "loss": 1.1426,
+      "step": 576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9648801063732e-06,
+      "loss": 1.0801,
+      "step": 577
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.96468785991629e-06,
+      "loss": 1.168,
+      "step": 578
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.964495092458061e-06,
+      "loss": 1.1924,
+      "step": 579
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.964301804039263e-06,
+      "loss": 1.1445,
+      "step": 580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.964107994700753e-06,
+      "loss": 1.1592,
+      "step": 581
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9639136644835e-06,
+      "loss": 1.1318,
+      "step": 582
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.963718813428584e-06,
+      "loss": 1.1465,
+      "step": 583
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.963523441577193e-06,
+      "loss": 1.1699,
+      "step": 584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.963327548970624e-06,
+      "loss": 1.1436,
+      "step": 585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.96313113565029e-06,
+      "loss": 1.1631,
+      "step": 586
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.962934201657706e-06,
+      "loss": 1.1357,
+      "step": 587
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.962736747034504e-06,
+      "loss": 1.145,
+      "step": 588
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.962538771822423e-06,
+      "loss": 1.1152,
+      "step": 589
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.96234027606331e-06,
+      "loss": 1.1494,
+      "step": 590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.962141259799127e-06,
+      "loss": 1.1353,
+      "step": 591
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.961941723071942e-06,
+      "loss": 1.1611,
+      "step": 592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9617416659239335e-06,
+      "loss": 1.1465,
+      "step": 593
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.961541088397392e-06,
+      "loss": 1.1562,
+      "step": 594
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.961339990534718e-06,
+      "loss": 1.1641,
+      "step": 595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.961138372378418e-06,
+      "loss": 1.1123,
+      "step": 596
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.960936233971115e-06,
+      "loss": 1.1387,
+      "step": 597
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.960733575355534e-06,
+      "loss": 1.1904,
+      "step": 598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.960530396574517e-06,
+      "loss": 1.1553,
+      "step": 599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.960326697671012e-06,
+      "loss": 1.1846,
+      "step": 600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.960122478688079e-06,
+      "loss": 1.1719,
+      "step": 601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959917739668888e-06,
+      "loss": 1.1416,
+      "step": 602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959712480656715e-06,
+      "loss": 1.127,
+      "step": 603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959506701694953e-06,
+      "loss": 1.1006,
+      "step": 604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959300402827098e-06,
+      "loss": 1.1689,
+      "step": 605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959093584096758e-06,
+      "loss": 1.1768,
+      "step": 606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958886245547654e-06,
+      "loss": 1.1436,
+      "step": 607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958678387223614e-06,
+      "loss": 1.1323,
+      "step": 608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9584700091685765e-06,
+      "loss": 1.1318,
+      "step": 609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958261111426589e-06,
+      "loss": 1.1416,
+      "step": 610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958051694041809e-06,
+      "loss": 1.1592,
+      "step": 611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.957841757058506e-06,
+      "loss": 1.1357,
+      "step": 612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.957631300521058e-06,
+      "loss": 1.1211,
+      "step": 613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.95742032447395e-06,
+      "loss": 1.1318,
+      "step": 614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.957208828961784e-06,
+      "loss": 1.1572,
+      "step": 615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.956996814029262e-06,
+      "loss": 1.1445,
+      "step": 616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.956784279721205e-06,
+      "loss": 1.166,
+      "step": 617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.956571226082538e-06,
+      "loss": 1.126,
+      "step": 618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.956357653158299e-06,
+      "loss": 1.1504,
+      "step": 619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.956143560993633e-06,
+      "loss": 1.1797,
+      "step": 620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.955928949633796e-06,
+      "loss": 1.1396,
+      "step": 621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.955713819124155e-06,
+      "loss": 1.1611,
+      "step": 622
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.955498169510186e-06,
+      "loss": 1.1582,
+      "step": 623
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.955282000837472e-06,
+      "loss": 1.1631,
+      "step": 624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.955065313151711e-06,
+      "loss": 1.1675,
+      "step": 625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.954848106498706e-06,
+      "loss": 1.1367,
+      "step": 626
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.954630380924373e-06,
+      "loss": 1.1709,
+      "step": 627
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9544121364747335e-06,
+      "loss": 1.1216,
+      "step": 628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.954193373195925e-06,
+      "loss": 1.1338,
+      "step": 629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9539740911341874e-06,
+      "loss": 1.1436,
+      "step": 630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.953754290335877e-06,
+      "loss": 1.1392,
+      "step": 631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.953533970847455e-06,
+      "loss": 1.1768,
+      "step": 632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.953313132715494e-06,
+      "loss": 1.0889,
+      "step": 633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.953091775986677e-06,
+      "loss": 1.1582,
+      "step": 634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.952869900707795e-06,
+      "loss": 1.168,
+      "step": 635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.952647506925749e-06,
+      "loss": 1.1729,
+      "step": 636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.952424594687553e-06,
+      "loss": 1.1328,
+      "step": 637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.952201164040323e-06,
+      "loss": 1.1455,
+      "step": 638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.951977215031293e-06,
+      "loss": 1.1245,
+      "step": 639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9517527477078e-06,
+      "loss": 1.1182,
+      "step": 640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.951527762117295e-06,
+      "loss": 1.165,
+      "step": 641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9513022583073364e-06,
+      "loss": 1.166,
+      "step": 642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.951076236325593e-06,
+      "loss": 1.1118,
+      "step": 643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.950849696219842e-06,
+      "loss": 1.1787,
+      "step": 644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9506226380379715e-06,
+      "loss": 1.1504,
+      "step": 645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.95039506182798e-06,
+      "loss": 1.1191,
+      "step": 646
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.95016696763797e-06,
+      "loss": 1.1523,
+      "step": 647
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.94993835551616e-06,
+      "loss": 1.1572,
+      "step": 648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.949709225510876e-06,
+      "loss": 1.1953,
+      "step": 649
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.949479577670552e-06,
+      "loss": 1.1504,
+      "step": 650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.949249412043733e-06,
+      "loss": 1.1436,
+      "step": 651
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.949018728679071e-06,
+      "loss": 1.1309,
+      "step": 652
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.948787527625332e-06,
+      "loss": 1.1895,
+      "step": 653
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.948555808931388e-06,
+      "loss": 1.1465,
+      "step": 654
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.94832357264622e-06,
+      "loss": 1.1377,
+      "step": 655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.94809081881892e-06,
+      "loss": 1.0732,
+      "step": 656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.94785754749869e-06,
+      "loss": 1.1494,
+      "step": 657
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.94762375873484e-06,
+      "loss": 1.1406,
+      "step": 658
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9473894525767885e-06,
+      "loss": 1.1416,
+      "step": 659
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.947154629074066e-06,
+      "loss": 1.1201,
+      "step": 660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.94691928827631e-06,
+      "loss": 1.1777,
+      "step": 661
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.946683430233269e-06,
+      "loss": 1.1201,
+      "step": 662
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9464470549948e-06,
+      "loss": 1.1074,
+      "step": 663
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9462101626108696e-06,
+      "loss": 1.166,
+      "step": 664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945972753131554e-06,
+      "loss": 1.1514,
+      "step": 665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945734826607037e-06,
+      "loss": 1.1162,
+      "step": 666
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945496383087613e-06,
+      "loss": 1.1582,
+      "step": 667
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945257422623688e-06,
+      "loss": 1.1377,
+      "step": 668
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945017945265773e-06,
+      "loss": 1.1382,
+      "step": 669
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.944777951064491e-06,
+      "loss": 1.1748,
+      "step": 670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.944537440070572e-06,
+      "loss": 1.1582,
+      "step": 671
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9442964123348604e-06,
+      "loss": 1.168,
+      "step": 672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.944054867908302e-06,
+      "loss": 1.1543,
+      "step": 673
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.943812806841958e-06,
+      "loss": 1.1699,
+      "step": 674
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.943570229186996e-06,
+      "loss": 1.1631,
+      "step": 675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.943327134994695e-06,
+      "loss": 1.1436,
+      "step": 676
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.94308352431644e-06,
+      "loss": 1.1519,
+      "step": 677
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.942839397203728e-06,
+      "loss": 1.1152,
+      "step": 678
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.942594753708165e-06,
+      "loss": 1.1816,
+      "step": 679
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.942349593881464e-06,
+      "loss": 1.1719,
+      "step": 680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.942103917775448e-06,
+      "loss": 1.1689,
+      "step": 681
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.941857725442051e-06,
+      "loss": 1.1348,
+      "step": 682
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.941611016933313e-06,
+      "loss": 1.1211,
+      "step": 683
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.941363792301387e-06,
+      "loss": 1.1699,
+      "step": 684
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.941116051598531e-06,
+      "loss": 1.1719,
+      "step": 685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.940867794877116e-06,
+      "loss": 1.1504,
+      "step": 686
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.940619022189616e-06,
+      "loss": 1.1523,
+      "step": 687
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.940369733588624e-06,
+      "loss": 1.168,
+      "step": 688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9401199291268314e-06,
+      "loss": 1.1289,
+      "step": 689
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9398696088570455e-06,
+      "loss": 1.1641,
+      "step": 690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.93961877283218e-06,
+      "loss": 1.123,
+      "step": 691
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.939367421105259e-06,
+      "loss": 1.1318,
+      "step": 692
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.939115553729415e-06,
+      "loss": 1.125,
+      "step": 693
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.938863170757887e-06,
+      "loss": 1.1641,
+      "step": 694
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.938610272244027e-06,
+      "loss": 1.1836,
+      "step": 695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.938356858241295e-06,
+      "loss": 1.165,
+      "step": 696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.938102928803259e-06,
+      "loss": 1.1836,
+      "step": 697
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.937848483983594e-06,
+      "loss": 1.1348,
+      "step": 698
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9375935238360896e-06,
+      "loss": 1.126,
+      "step": 699
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.937338048414638e-06,
+      "loss": 1.1211,
+      "step": 700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.937082057773245e-06,
+      "loss": 1.1416,
+      "step": 701
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9368255519660215e-06,
+      "loss": 1.1455,
+      "step": 702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.936568531047192e-06,
+      "loss": 1.1328,
+      "step": 703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.936310995071085e-06,
+      "loss": 1.1035,
+      "step": 704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.936052944092142e-06,
+      "loss": 1.1357,
+      "step": 705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.93579437816491e-06,
+      "loss": 1.2178,
+      "step": 706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.935535297344047e-06,
+      "loss": 1.1592,
+      "step": 707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.935275701684318e-06,
+      "loss": 1.1738,
+      "step": 708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.935015591240599e-06,
+      "loss": 1.1445,
+      "step": 709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9347549660678745e-06,
+      "loss": 1.1338,
+      "step": 710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.934493826221236e-06,
+      "loss": 1.1338,
+      "step": 711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.934232171755885e-06,
+      "loss": 1.1738,
+      "step": 712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.933970002727131e-06,
+      "loss": 1.1543,
+      "step": 713
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.933707319190394e-06,
+      "loss": 1.1611,
+      "step": 714
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.933444121201202e-06,
+      "loss": 1.1167,
+      "step": 715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.933180408815191e-06,
+      "loss": 1.1152,
+      "step": 716
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.932916182088105e-06,
+      "loss": 1.0952,
+      "step": 717
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.932651441075801e-06,
+      "loss": 1.1592,
+      "step": 718
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.932386185834238e-06,
+      "loss": 1.1338,
+      "step": 719
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.932120416419489e-06,
+      "loss": 1.1572,
+      "step": 720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.931854132887735e-06,
+      "loss": 1.1528,
+      "step": 721
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.931587335295263e-06,
+      "loss": 1.1748,
+      "step": 722
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.931320023698471e-06,
+      "loss": 1.1738,
+      "step": 723
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.931052198153865e-06,
+      "loss": 1.1221,
+      "step": 724
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.93078385871806e-06,
+      "loss": 1.1416,
+      "step": 725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.930515005447779e-06,
+      "loss": 1.1709,
+      "step": 726
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.930245638399854e-06,
+      "loss": 1.1738,
+      "step": 727
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.929975757631225e-06,
+      "loss": 1.1377,
+      "step": 728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.929705363198943e-06,
+      "loss": 1.0908,
+      "step": 729
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.929434455160163e-06,
+      "loss": 1.1401,
+      "step": 730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.929163033572153e-06,
+      "loss": 1.1865,
+      "step": 731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.928891098492286e-06,
+      "loss": 1.1719,
+      "step": 732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9286186499780485e-06,
+      "loss": 1.1494,
+      "step": 733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.92834568808703e-06,
+      "loss": 1.2041,
+      "step": 734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9280722128769305e-06,
+      "loss": 1.2012,
+      "step": 735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.927798224405561e-06,
+      "loss": 1.1084,
+      "step": 736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.927523722730837e-06,
+      "loss": 1.1182,
+      "step": 737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9272487079107855e-06,
+      "loss": 1.165,
+      "step": 738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9269731800035416e-06,
+      "loss": 1.1357,
+      "step": 739
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9266971390673456e-06,
+      "loss": 1.1377,
+      "step": 740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9264205851605504e-06,
+      "loss": 1.1299,
+      "step": 741
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.926143518341617e-06,
+      "loss": 1.1777,
+      "step": 742
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.925865938669111e-06,
+      "loss": 1.1108,
+      "step": 743
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.92558784620171e-06,
+      "loss": 1.1631,
+      "step": 744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.925309240998199e-06,
+      "loss": 1.1357,
+      "step": 745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.925030123117472e-06,
+      "loss": 1.1494,
+      "step": 746
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.92475049261853e-06,
+      "loss": 1.1123,
+      "step": 747
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9244703495604825e-06,
+      "loss": 1.1406,
+      "step": 748
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.92418969400255e-06,
+      "loss": 1.1631,
+      "step": 749
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.923908526004057e-06,
+      "loss": 1.123,
+      "step": 750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.923626845624438e-06,
+      "loss": 1.166,
+      "step": 751
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.92334465292324e-06,
+      "loss": 1.1533,
+      "step": 752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.923061947960111e-06,
+      "loss": 1.1689,
+      "step": 753
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.922778730794813e-06,
+      "loss": 1.1279,
+      "step": 754
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.922495001487213e-06,
+      "loss": 1.1367,
+      "step": 755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.922210760097289e-06,
+      "loss": 1.124,
+      "step": 756
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.921926006685123e-06,
+      "loss": 1.1533,
+      "step": 757
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.921640741310913e-06,
+      "loss": 1.1621,
+      "step": 758
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.921354964034954e-06,
+      "loss": 1.1436,
+      "step": 759
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.92106867491766e-06,
+      "loss": 1.1738,
+      "step": 760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.920781874019547e-06,
+      "loss": 1.1543,
+      "step": 761
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.920494561401241e-06,
+      "loss": 1.1787,
+      "step": 762
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9202067371234755e-06,
+      "loss": 1.1592,
+      "step": 763
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9199184012470925e-06,
+      "loss": 1.1709,
+      "step": 764
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.919629553833043e-06,
+      "loss": 1.1592,
+      "step": 765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9193401949423855e-06,
+      "loss": 1.1387,
+      "step": 766
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.919050324636285e-06,
+      "loss": 1.1455,
+      "step": 767
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.918759942976018e-06,
+      "loss": 1.1621,
+      "step": 768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.918469050022967e-06,
+      "loss": 1.1274,
+      "step": 769
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.918177645838622e-06,
+      "loss": 1.1768,
+      "step": 770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.917885730484582e-06,
+      "loss": 1.1533,
+      "step": 771
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9175933040225535e-06,
+      "loss": 1.167,
+      "step": 772
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.917300366514352e-06,
+      "loss": 1.168,
+      "step": 773
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9170069180219005e-06,
+      "loss": 1.1562,
+      "step": 774
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.916712958607231e-06,
+      "loss": 1.1914,
+      "step": 775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.916418488332479e-06,
+      "loss": 1.1895,
+      "step": 776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.916123507259896e-06,
+      "loss": 1.1562,
+      "step": 777
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.915828015451835e-06,
+      "loss": 1.1177,
+      "step": 778
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.915532012970758e-06,
+      "loss": 1.1895,
+      "step": 779
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.915235499879236e-06,
+      "loss": 1.1289,
+      "step": 780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.91493847623995e-06,
+      "loss": 1.1553,
+      "step": 781
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.914640942115685e-06,
+      "loss": 1.1709,
+      "step": 782
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9143428975693355e-06,
+      "loss": 1.1348,
+      "step": 783
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9140443426639044e-06,
+      "loss": 1.2012,
+      "step": 784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.913745277462503e-06,
+      "loss": 1.1382,
+      "step": 785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.913445702028348e-06,
+      "loss": 1.1387,
+      "step": 786
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.913145616424766e-06,
+      "loss": 1.1719,
+      "step": 787
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9128450207151915e-06,
+      "loss": 1.1416,
+      "step": 788
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.912543914963167e-06,
+      "loss": 1.1367,
+      "step": 789
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.912242299232341e-06,
+      "loss": 1.1172,
+      "step": 790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.911940173586471e-06,
+      "loss": 1.1006,
+      "step": 791
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.911637538089421e-06,
+      "loss": 1.1211,
+      "step": 792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.911334392805167e-06,
+      "loss": 1.127,
+      "step": 793
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.911030737797788e-06,
+      "loss": 1.1377,
+      "step": 794
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.910726573131474e-06,
+      "loss": 1.1621,
+      "step": 795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.910421898870518e-06,
+      "loss": 1.1709,
+      "step": 796
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.910116715079327e-06,
+      "loss": 1.1768,
+      "step": 797
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.909811021822411e-06,
+      "loss": 1.1416,
+      "step": 798
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9095048191643904e-06,
+      "loss": 1.126,
+      "step": 799
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.909198107169991e-06,
+      "loss": 1.1484,
+      "step": 800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.90889088590405e-06,
+      "loss": 1.167,
+      "step": 801
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.908583155431508e-06,
+      "loss": 1.1836,
+      "step": 802
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9082749158174146e-06,
+      "loss": 1.125,
+      "step": 803
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.907966167126927e-06,
+      "loss": 1.1641,
+      "step": 804
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.907656909425314e-06,
+      "loss": 1.0947,
+      "step": 805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9073471427779436e-06,
+      "loss": 1.1309,
+      "step": 806
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9070368672503e-06,
+      "loss": 1.1934,
+      "step": 807
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.906726082907969e-06,
+      "loss": 1.1699,
+      "step": 808
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.906414789816647e-06,
+      "loss": 1.1406,
+      "step": 809
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.906102988042137e-06,
+      "loss": 1.1582,
+      "step": 810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.90579067765035e-06,
+      "loss": 1.1533,
+      "step": 811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.905477858707304e-06,
+      "loss": 1.1328,
+      "step": 812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.905164531279124e-06,
+      "loss": 1.1709,
+      "step": 813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.904850695432043e-06,
+      "loss": 1.1133,
+      "step": 814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9045363512324034e-06,
+      "loss": 1.0947,
+      "step": 815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.904221498746652e-06,
+      "loss": 1.0977,
+      "step": 816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9039061380413435e-06,
+      "loss": 1.1162,
+      "step": 817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.903590269183143e-06,
+      "loss": 1.1514,
+      "step": 818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.903273892238818e-06,
+      "loss": 1.1523,
+      "step": 819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.902957007275249e-06,
+      "loss": 1.1436,
+      "step": 820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.90263961435942e-06,
+      "loss": 1.1396,
+      "step": 821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.902321713558423e-06,
+      "loss": 1.168,
+      "step": 822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.902003304939459e-06,
+      "loss": 1.1182,
+      "step": 823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.901684388569834e-06,
+      "loss": 1.1484,
+      "step": 824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9013649645169634e-06,
+      "loss": 1.1611,
+      "step": 825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.901045032848369e-06,
+      "loss": 1.1279,
+      "step": 826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.900724593631679e-06,
+      "loss": 1.1191,
+      "step": 827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.900403646934632e-06,
+      "loss": 1.1738,
+      "step": 828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.90008219282507e-06,
+      "loss": 1.1797,
+      "step": 829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.899760231370945e-06,
+      "loss": 1.1953,
+      "step": 830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.899437762640313e-06,
+      "loss": 1.1465,
+      "step": 831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.899114786701343e-06,
+      "loss": 1.1729,
+      "step": 832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.898791303622305e-06,
+      "loss": 1.1934,
+      "step": 833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.89846731347158e-06,
+      "loss": 1.1719,
+      "step": 834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.898142816317654e-06,
+      "loss": 1.1484,
+      "step": 835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.897817812229124e-06,
+      "loss": 1.1279,
+      "step": 836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.897492301274689e-06,
+      "loss": 1.1484,
+      "step": 837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.897166283523157e-06,
+      "loss": 1.125,
+      "step": 838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.896839759043445e-06,
+      "loss": 1.1406,
+      "step": 839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8965127279045776e-06,
+      "loss": 1.1387,
+      "step": 840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.896185190175681e-06,
+      "loss": 1.2129,
+      "step": 841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.895857145925995e-06,
+      "loss": 1.1377,
+      "step": 842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8955285952248615e-06,
+      "loss": 1.1846,
+      "step": 843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.895199538141733e-06,
+      "loss": 1.1587,
+      "step": 844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.894869974746168e-06,
+      "loss": 1.1592,
+      "step": 845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.89453990510783e-06,
+      "loss": 1.1318,
+      "step": 846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.894209329296494e-06,
+      "loss": 1.1392,
+      "step": 847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.893878247382037e-06,
+      "loss": 1.1533,
+      "step": 848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.893546659434445e-06,
+      "loss": 1.1592,
+      "step": 849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.893214565523811e-06,
+      "loss": 1.1074,
+      "step": 850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.892881965720338e-06,
+      "loss": 1.1738,
+      "step": 851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.892548860094331e-06,
+      "loss": 1.1455,
+      "step": 852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8922152487162025e-06,
+      "loss": 1.1357,
+      "step": 853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.891881131656476e-06,
+      "loss": 1.2129,
+      "step": 854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.891546508985778e-06,
+      "loss": 1.1963,
+      "step": 855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.891211380774843e-06,
+      "loss": 1.1348,
+      "step": 856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.890875747094513e-06,
+      "loss": 1.1738,
+      "step": 857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.890539608015737e-06,
+      "loss": 1.1533,
+      "step": 858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.890202963609568e-06,
+      "loss": 1.1416,
+      "step": 859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.88986581394717e-06,
+      "loss": 1.1348,
+      "step": 860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8895281590998115e-06,
+      "loss": 1.1494,
+      "step": 861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.889189999138868e-06,
+      "loss": 1.1523,
+      "step": 862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8888513341358215e-06,
+      "loss": 1.1934,
+      "step": 863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.888512164162262e-06,
+      "loss": 1.1484,
+      "step": 864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.888172489289884e-06,
+      "loss": 1.1826,
+      "step": 865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.887832309590492e-06,
+      "loss": 1.144,
+      "step": 866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8874916251359935e-06,
+      "loss": 1.1611,
+      "step": 867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.887150435998406e-06,
+      "loss": 1.1025,
+      "step": 868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.886808742249851e-06,
+      "loss": 1.1128,
+      "step": 869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.886466543962559e-06,
+      "loss": 1.1411,
+      "step": 870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.886123841208865e-06,
+      "loss": 1.146,
+      "step": 871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.885780634061212e-06,
+      "loss": 1.1416,
+      "step": 872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.88543692259215e-06,
+      "loss": 1.1504,
+      "step": 873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.885092706874334e-06,
+      "loss": 1.1621,
+      "step": 874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.884747986980527e-06,
+      "loss": 1.1211,
+      "step": 875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.884402762983598e-06,
+      "loss": 1.1455,
+      "step": 876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.884057034956523e-06,
+      "loss": 1.1523,
+      "step": 877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.883710802972383e-06,
+      "loss": 1.1621,
+      "step": 878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.883364067104368e-06,
+      "loss": 1.1924,
+      "step": 879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.883016827425773e-06,
+      "loss": 1.1157,
+      "step": 880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.882669084009999e-06,
+      "loss": 1.1187,
+      "step": 881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.882320836930554e-06,
+      "loss": 1.1553,
+      "step": 882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8819720862610545e-06,
+      "loss": 1.1172,
+      "step": 883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.881622832075219e-06,
+      "loss": 1.1572,
+      "step": 884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.881273074446877e-06,
+      "loss": 1.1191,
+      "step": 885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.880922813449962e-06,
+      "loss": 1.1099,
+      "step": 886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8805720491585155e-06,
+      "loss": 1.1426,
+      "step": 887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.880220781646682e-06,
+      "loss": 1.1172,
+      "step": 888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.879869010988715e-06,
+      "loss": 1.127,
+      "step": 889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.879516737258976e-06,
+      "loss": 1.1592,
+      "step": 890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8791639605319294e-06,
+      "loss": 1.0967,
+      "step": 891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.878810680882148e-06,
+      "loss": 1.1279,
+      "step": 892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.878456898384309e-06,
+      "loss": 1.1299,
+      "step": 893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.878102613113198e-06,
+      "loss": 1.1201,
+      "step": 894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.877747825143707e-06,
+      "loss": 1.1582,
+      "step": 895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.877392534550832e-06,
+      "loss": 1.1689,
+      "step": 896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.877036741409676e-06,
+      "loss": 1.1377,
+      "step": 897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.876680445795452e-06,
+      "loss": 1.166,
+      "step": 898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8763236477834726e-06,
+      "loss": 1.1309,
+      "step": 899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.875966347449162e-06,
+      "loss": 1.1543,
+      "step": 900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.875608544868047e-06,
+      "loss": 1.1416,
+      "step": 901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.875250240115763e-06,
+      "loss": 1.1436,
+      "step": 902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.874891433268051e-06,
+      "loss": 1.1323,
+      "step": 903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.874532124400758e-06,
+      "loss": 1.1768,
+      "step": 904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.874172313589837e-06,
+      "loss": 1.1431,
+      "step": 905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.873812000911346e-06,
+      "loss": 1.1533,
+      "step": 906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8734511864414505e-06,
+      "loss": 1.1406,
+      "step": 907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.873089870256423e-06,
+      "loss": 1.1436,
+      "step": 908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8727280524326395e-06,
+      "loss": 1.1436,
+      "step": 909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.872365733046584e-06,
+      "loss": 1.0693,
+      "step": 910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.872002912174846e-06,
+      "loss": 1.166,
+      "step": 911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.871639589894119e-06,
+      "loss": 1.1416,
+      "step": 912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.871275766281207e-06,
+      "loss": 1.1875,
+      "step": 913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.870911441413016e-06,
+      "loss": 1.1396,
+      "step": 914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.870546615366561e-06,
+      "loss": 1.1377,
+      "step": 915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.870181288218958e-06,
+      "loss": 1.165,
+      "step": 916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.869815460047435e-06,
+      "loss": 1.1162,
+      "step": 917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.869449130929321e-06,
+      "loss": 1.1621,
+      "step": 918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.869082300942054e-06,
+      "loss": 1.1436,
+      "step": 919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.868714970163178e-06,
+      "loss": 1.1357,
+      "step": 920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.868347138670339e-06,
+      "loss": 1.1211,
+      "step": 921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.867978806541293e-06,
+      "loss": 1.168,
+      "step": 922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.867609973853902e-06,
+      "loss": 1.1309,
+      "step": 923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8672406406861295e-06,
+      "loss": 1.1025,
+      "step": 924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.866870807116049e-06,
+      "loss": 1.1299,
+      "step": 925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.866500473221838e-06,
+      "loss": 1.1641,
+      "step": 926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.866129639081779e-06,
+      "loss": 1.1758,
+      "step": 927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.865758304774263e-06,
+      "loss": 1.1533,
+      "step": 928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.865386470377785e-06,
+      "loss": 1.1299,
+      "step": 929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.865014135970944e-06,
+      "loss": 1.1777,
+      "step": 930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.864641301632446e-06,
+      "loss": 1.1377,
+      "step": 931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.864267967441106e-06,
+      "loss": 1.1338,
+      "step": 932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.863894133475839e-06,
+      "loss": 1.2051,
+      "step": 933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.86351979981567e-06,
+      "loss": 1.1357,
+      "step": 934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.863144966539728e-06,
+      "loss": 1.166,
+      "step": 935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.862769633727247e-06,
+      "loss": 1.1826,
+      "step": 936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8623938014575664e-06,
+      "loss": 1.1709,
+      "step": 937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.862017469810134e-06,
+      "loss": 1.1357,
+      "step": 938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.861640638864499e-06,
+      "loss": 1.1719,
+      "step": 939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.86126330870032e-06,
+      "loss": 1.1445,
+      "step": 940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.860885479397359e-06,
+      "loss": 1.124,
+      "step": 941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8605071510354836e-06,
+      "loss": 1.1768,
+      "step": 942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8601283236946684e-06,
+      "loss": 1.1523,
+      "step": 943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8597489974549905e-06,
+      "loss": 1.166,
+      "step": 944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.859369172396635e-06,
+      "loss": 1.1582,
+      "step": 945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.858988848599892e-06,
+      "loss": 1.1445,
+      "step": 946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.858608026145157e-06,
+      "loss": 1.1245,
+      "step": 947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.858226705112931e-06,
+      "loss": 1.1396,
+      "step": 948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.857844885583818e-06,
+      "loss": 1.0967,
+      "step": 949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8574625676385315e-06,
+      "loss": 1.1211,
+      "step": 950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.857079751357887e-06,
+      "loss": 1.1621,
+      "step": 951
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8566964368228084e-06,
+      "loss": 1.126,
+      "step": 952
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.856312624114321e-06,
+      "loss": 1.1348,
+      "step": 953
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.855928313313558e-06,
+      "loss": 1.1641,
+      "step": 954
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.855543504501759e-06,
+      "loss": 1.1582,
+      "step": 955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.855158197760266e-06,
+      "loss": 1.1279,
+      "step": 956
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.854772393170528e-06,
+      "loss": 1.1289,
+      "step": 957
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.854386090814098e-06,
+      "loss": 1.0913,
+      "step": 958
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8539992907726366e-06,
+      "loss": 1.1641,
+      "step": 959
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.853611993127906e-06,
+      "loss": 1.1602,
+      "step": 960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.853224197961778e-06,
+      "loss": 1.1797,
+      "step": 961
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.852835905356225e-06,
+      "loss": 1.1553,
+      "step": 962
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8524471153933285e-06,
+      "loss": 1.167,
+      "step": 963
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.852057828155272e-06,
+      "loss": 1.1045,
+      "step": 964
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.851668043724347e-06,
+      "loss": 1.165,
+      "step": 965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8512777621829465e-06,
+      "loss": 1.1045,
+      "step": 966
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.850886983613573e-06,
+      "loss": 1.1084,
+      "step": 967
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.850495708098831e-06,
+      "loss": 1.126,
+      "step": 968
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.850103935721429e-06,
+      "loss": 1.1426,
+      "step": 969
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.849711666564184e-06,
+      "loss": 1.1729,
+      "step": 970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.849318900710016e-06,
+      "loss": 1.1621,
+      "step": 971
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.848925638241951e-06,
+      "loss": 1.1729,
+      "step": 972
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.848531879243118e-06,
+      "loss": 1.1318,
+      "step": 973
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.848137623796754e-06,
+      "loss": 1.1484,
+      "step": 974
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.847742871986198e-06,
+      "loss": 1.1914,
+      "step": 975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.847347623894895e-06,
+      "loss": 1.1641,
+      "step": 976
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.846951879606396e-06,
+      "loss": 1.1387,
+      "step": 977
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.846555639204354e-06,
+      "loss": 1.1514,
+      "step": 978
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.846158902772532e-06,
+      "loss": 1.1416,
+      "step": 979
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.845761670394792e-06,
+      "loss": 1.167,
+      "step": 980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.845363942155104e-06,
+      "loss": 1.1611,
+      "step": 981
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.844965718137542e-06,
+      "loss": 1.1582,
+      "step": 982
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.844566998426286e-06,
+      "loss": 1.123,
+      "step": 983
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.844167783105621e-06,
+      "loss": 1.125,
+      "step": 984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.843768072259933e-06,
+      "loss": 1.1826,
+      "step": 985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.843367865973717e-06,
+      "loss": 1.1572,
+      "step": 986
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.842967164331571e-06,
+      "loss": 1.1309,
+      "step": 987
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.842565967418197e-06,
+      "loss": 1.1289,
+      "step": 988
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.842164275318404e-06,
+      "loss": 1.1621,
+      "step": 989
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.841762088117102e-06,
+      "loss": 1.1196,
+      "step": 990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.841359405899311e-06,
+      "loss": 1.103,
+      "step": 991
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.84095622875015e-06,
+      "loss": 1.1592,
+      "step": 992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.840552556754846e-06,
+      "loss": 1.1641,
+      "step": 993
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.84014838999873e-06,
+      "loss": 1.1318,
+      "step": 994
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.839743728567237e-06,
+      "loss": 1.123,
+      "step": 995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.839338572545906e-06,
+      "loss": 1.167,
+      "step": 996
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.838932922020384e-06,
+      "loss": 1.1484,
+      "step": 997
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.838526777076417e-06,
+      "loss": 1.1294,
+      "step": 998
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.83812013779986e-06,
+      "loss": 1.1953,
+      "step": 999
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.837713004276671e-06,
+      "loss": 1.1445,
+      "step": 1000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.837305376592912e-06,
+      "loss": 1.1455,
+      "step": 1001
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.836897254834752e-06,
+      "loss": 1.1289,
+      "step": 1002
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.836488639088458e-06,
+      "loss": 1.1055,
+      "step": 1003
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.83607952944041e-06,
+      "loss": 1.1426,
+      "step": 1004
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.835669925977086e-06,
+      "loss": 1.1377,
+      "step": 1005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8352598287850704e-06,
+      "loss": 1.1553,
+      "step": 1006
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.834849237951053e-06,
+      "loss": 1.1553,
+      "step": 1007
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.834438153561828e-06,
+      "loss": 1.1328,
+      "step": 1008
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.834026575704291e-06,
+      "loss": 1.165,
+      "step": 1009
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8336145044654445e-06,
+      "loss": 1.1504,
+      "step": 1010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.833201939932396e-06,
+      "loss": 1.1807,
+      "step": 1011
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.832788882192355e-06,
+      "loss": 1.1475,
+      "step": 1012
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.832375331332636e-06,
+      "loss": 1.1484,
+      "step": 1013
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.831961287440658e-06,
+      "loss": 1.1309,
+      "step": 1014
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.831546750603946e-06,
+      "loss": 1.1367,
+      "step": 1015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.831131720910125e-06,
+      "loss": 1.1172,
+      "step": 1016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8307161984469275e-06,
+      "loss": 1.1816,
+      "step": 1017
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.83030018330219e-06,
+      "loss": 1.1641,
+      "step": 1018
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8298836755638515e-06,
+      "loss": 1.127,
+      "step": 1019
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.829466675319956e-06,
+      "loss": 1.1465,
+      "step": 1020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.829049182658652e-06,
+      "loss": 1.1377,
+      "step": 1021
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8286311976681915e-06,
+      "loss": 1.1289,
+      "step": 1022
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.828212720436931e-06,
+      "loss": 1.1182,
+      "step": 1023
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.827793751053331e-06,
+      "loss": 1.1846,
+      "step": 1024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.827374289605956e-06,
+      "loss": 1.1484,
+      "step": 1025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.826954336183475e-06,
+      "loss": 1.1553,
+      "step": 1026
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.826533890874658e-06,
+      "loss": 1.1338,
+      "step": 1027
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.826112953768385e-06,
+      "loss": 1.1787,
+      "step": 1028
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.825691524953633e-06,
+      "loss": 1.167,
+      "step": 1029
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8252696045194875e-06,
+      "loss": 1.0918,
+      "step": 1030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.824847192555137e-06,
+      "loss": 1.165,
+      "step": 1031
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.824424289149873e-06,
+      "loss": 1.1807,
+      "step": 1032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8240008943930935e-06,
+      "loss": 1.1592,
+      "step": 1033
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.823577008374296e-06,
+      "loss": 1.1934,
+      "step": 1034
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.823152631183085e-06,
+      "loss": 1.1699,
+      "step": 1035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.822727762909168e-06,
+      "loss": 1.1445,
+      "step": 1036
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.822302403642357e-06,
+      "loss": 1.1367,
+      "step": 1037
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.821876553472566e-06,
+      "loss": 1.1201,
+      "step": 1038
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.821450212489814e-06,
+      "loss": 1.1367,
+      "step": 1039
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.821023380784225e-06,
+      "loss": 1.1348,
+      "step": 1040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.820596058446025e-06,
+      "loss": 1.1318,
+      "step": 1041
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.820168245565542e-06,
+      "loss": 1.1992,
+      "step": 1042
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.819739942233213e-06,
+      "loss": 1.1113,
+      "step": 1043
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.819311148539572e-06,
+      "loss": 1.1504,
+      "step": 1044
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.818881864575263e-06,
+      "loss": 1.1348,
+      "step": 1045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.81845209043103e-06,
+      "loss": 1.1348,
+      "step": 1046
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.818021826197721e-06,
+      "loss": 1.1436,
+      "step": 1047
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.817591071966288e-06,
+      "loss": 1.1699,
+      "step": 1048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.817159827827786e-06,
+      "loss": 1.1543,
+      "step": 1049
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.816728093873377e-06,
+      "loss": 1.085,
+      "step": 1050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.816295870194319e-06,
+      "loss": 1.1846,
+      "step": 1051
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.815863156881981e-06,
+      "loss": 1.1406,
+      "step": 1052
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.815429954027834e-06,
+      "loss": 1.1533,
+      "step": 1053
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.814996261723448e-06,
+      "loss": 1.1221,
+      "step": 1054
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.814562080060501e-06,
+      "loss": 1.1504,
+      "step": 1055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.814127409130773e-06,
+      "loss": 1.1416,
+      "step": 1056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.813692249026148e-06,
+      "loss": 1.1157,
+      "step": 1057
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.813256599838614e-06,
+      "loss": 1.1309,
+      "step": 1058
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.812820461660258e-06,
+      "loss": 1.1455,
+      "step": 1059
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.812383834583276e-06,
+      "loss": 1.1396,
+      "step": 1060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.811946718699964e-06,
+      "loss": 1.1729,
+      "step": 1061
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.811509114102723e-06,
+      "loss": 1.0669,
+      "step": 1062
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8110710208840575e-06,
+      "loss": 1.1621,
+      "step": 1063
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.810632439136572e-06,
+      "loss": 1.1406,
+      "step": 1064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.810193368952979e-06,
+      "loss": 1.105,
+      "step": 1065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.80975381042609e-06,
+      "loss": 1.1426,
+      "step": 1066
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.809313763648823e-06,
+      "loss": 1.127,
+      "step": 1067
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.808873228714197e-06,
+      "loss": 1.1128,
+      "step": 1068
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8084322057153365e-06,
+      "loss": 1.1318,
+      "step": 1069
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.807990694745466e-06,
+      "loss": 1.1284,
+      "step": 1070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.807548695897915e-06,
+      "loss": 1.1602,
+      "step": 1071
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.807106209266118e-06,
+      "loss": 1.124,
+      "step": 1072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8066632349436094e-06,
+      "loss": 1.1602,
+      "step": 1073
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.806219773024028e-06,
+      "loss": 1.1299,
+      "step": 1074
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.805775823601115e-06,
+      "loss": 1.1064,
+      "step": 1075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.805331386768714e-06,
+      "loss": 1.1514,
+      "step": 1076
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.804886462620777e-06,
+      "loss": 1.1621,
+      "step": 1077
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.80444105125135e-06,
+      "loss": 1.0977,
+      "step": 1078
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8039951527545914e-06,
+      "loss": 1.1543,
+      "step": 1079
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.803548767224755e-06,
+      "loss": 1.1484,
+      "step": 1080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8031018947562015e-06,
+      "loss": 1.127,
+      "step": 1081
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.802654535443394e-06,
+      "loss": 1.1514,
+      "step": 1082
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8022066893808976e-06,
+      "loss": 1.0918,
+      "step": 1083
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.801758356663381e-06,
+      "loss": 1.1074,
+      "step": 1084
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.801309537385615e-06,
+      "loss": 1.1201,
+      "step": 1085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.800860231642476e-06,
+      "loss": 1.166,
+      "step": 1086
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.800410439528938e-06,
+      "loss": 1.1729,
+      "step": 1087
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.799960161140083e-06,
+      "loss": 1.1348,
+      "step": 1088
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.7995093965710935e-06,
+      "loss": 1.1455,
+      "step": 1089
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.799058145917254e-06,
+      "loss": 1.1699,
+      "step": 1090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.798606409273954e-06,
+      "loss": 1.0811,
+      "step": 1091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.798154186736684e-06,
+      "loss": 1.1523,
+      "step": 1092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7977014784010365e-06,
+      "loss": 1.1641,
+      "step": 1093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.79724828436271e-06,
+      "loss": 1.1494,
+      "step": 1094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.796794604717502e-06,
+      "loss": 1.1226,
+      "step": 1095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.796340439561314e-06,
+      "loss": 1.0659,
+      "step": 1096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.795885788990151e-06,
+      "loss": 1.1338,
+      "step": 1097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.79543065310012e-06,
+      "loss": 1.1582,
+      "step": 1098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.79497503198743e-06,
+      "loss": 1.1787,
+      "step": 1099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.794518925748394e-06,
+      "loss": 1.124,
+      "step": 1100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.794062334479424e-06,
+      "loss": 1.1436,
+      "step": 1101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.793605258277041e-06,
+      "loss": 1.1768,
+      "step": 1102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7931476972378625e-06,
+      "loss": 1.1523,
+      "step": 1103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7926896514586104e-06,
+      "loss": 1.1475,
+      "step": 1104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.792231121036111e-06,
+      "loss": 1.127,
+      "step": 1105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.791772106067289e-06,
+      "loss": 1.167,
+      "step": 1106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.791312606649176e-06,
+      "loss": 1.1562,
+      "step": 1107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.790852622878901e-06,
+      "loss": 1.1533,
+      "step": 1108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7903921548537025e-06,
+      "loss": 1.1318,
+      "step": 1109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.789931202670914e-06,
+      "loss": 1.0967,
+      "step": 1110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7894697664279755e-06,
+      "loss": 1.1562,
+      "step": 1111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7890078462224276e-06,
+      "loss": 1.1328,
+      "step": 1112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.788545442151915e-06,
+      "loss": 1.167,
+      "step": 1113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.788082554314183e-06,
+      "loss": 1.1729,
+      "step": 1114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.787619182807081e-06,
+      "loss": 1.1475,
+      "step": 1115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.787155327728557e-06,
+      "loss": 1.1592,
+      "step": 1116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.786690989176665e-06,
+      "loss": 1.1309,
+      "step": 1117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.78622616724956e-06,
+      "loss": 1.1216,
+      "step": 1118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.785760862045499e-06,
+      "loss": 1.1411,
+      "step": 1119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.78529507366284e-06,
+      "loss": 1.1689,
+      "step": 1120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.784828802200046e-06,
+      "loss": 1.1318,
+      "step": 1121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.78436204775568e-06,
+      "loss": 1.1191,
+      "step": 1122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.783894810428405e-06,
+      "loss": 1.168,
+      "step": 1123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.783427090316992e-06,
+      "loss": 1.1553,
+      "step": 1124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7829588875203106e-06,
+      "loss": 1.1191,
+      "step": 1125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.78249020213733e-06,
+      "loss": 1.1201,
+      "step": 1126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7820210342671245e-06,
+      "loss": 1.1611,
+      "step": 1127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.781551384008871e-06,
+      "loss": 1.166,
+      "step": 1128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.781081251461847e-06,
+      "loss": 1.1162,
+      "step": 1129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.780610636725431e-06,
+      "loss": 1.1104,
+      "step": 1130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.780139539899105e-06,
+      "loss": 1.1289,
+      "step": 1131
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.779667961082452e-06,
+      "loss": 1.1377,
+      "step": 1132
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.779195900375158e-06,
+      "loss": 1.1396,
+      "step": 1133
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.778723357877012e-06,
+      "loss": 1.126,
+      "step": 1134
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.778250333687899e-06,
+      "loss": 1.1201,
+      "step": 1135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.777776827907813e-06,
+      "loss": 1.1328,
+      "step": 1136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.777302840636845e-06,
+      "loss": 1.1758,
+      "step": 1137
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.776828371975189e-06,
+      "loss": 1.1201,
+      "step": 1138
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.776353422023145e-06,
+      "loss": 1.1592,
+      "step": 1139
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.775877990881105e-06,
+      "loss": 1.1699,
+      "step": 1140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.775402078649574e-06,
+      "loss": 1.1943,
+      "step": 1141
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.774925685429149e-06,
+      "loss": 1.1719,
+      "step": 1142
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7744488113205364e-06,
+      "loss": 1.1562,
+      "step": 1143
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.77397145642454e-06,
+      "loss": 1.1592,
+      "step": 1144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7734936208420645e-06,
+      "loss": 1.1523,
+      "step": 1145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.77301530467412e-06,
+      "loss": 1.124,
+      "step": 1146
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.772536508021815e-06,
+      "loss": 1.1455,
+      "step": 1147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.772057230986359e-06,
+      "loss": 1.1592,
+      "step": 1148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.771577473669067e-06,
+      "loss": 1.123,
+      "step": 1149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.771097236171353e-06,
+      "loss": 1.1353,
+      "step": 1150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.770616518594731e-06,
+      "loss": 1.1504,
+      "step": 1151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.770135321040819e-06,
+      "loss": 1.1553,
+      "step": 1152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.769653643611337e-06,
+      "loss": 1.0986,
+      "step": 1153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.769171486408102e-06,
+      "loss": 1.1655,
+      "step": 1154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.768688849533039e-06,
+      "loss": 1.1338,
+      "step": 1155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.768205733088167e-06,
+      "loss": 1.1719,
+      "step": 1156
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.767722137175613e-06,
+      "loss": 1.1582,
+      "step": 1157
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.767238061897601e-06,
+      "loss": 1.1333,
+      "step": 1158
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.76675350735646e-06,
+      "loss": 1.1592,
+      "step": 1159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.766268473654616e-06,
+      "loss": 1.167,
+      "step": 1160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.765782960894599e-06,
+      "loss": 1.1484,
+      "step": 1161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.76529696917904e-06,
+      "loss": 1.1367,
+      "step": 1162
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.764810498610671e-06,
+      "loss": 1.1104,
+      "step": 1163
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.764323549292326e-06,
+      "loss": 1.1396,
+      "step": 1164
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.763836121326939e-06,
+      "loss": 1.1299,
+      "step": 1165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.763348214817545e-06,
+      "loss": 1.1631,
+      "step": 1166
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.762859829867281e-06,
+      "loss": 1.1592,
+      "step": 1167
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.762370966579385e-06,
+      "loss": 1.167,
+      "step": 1168
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.761881625057196e-06,
+      "loss": 1.1172,
+      "step": 1169
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.761391805404154e-06,
+      "loss": 1.1152,
+      "step": 1170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7609015077238e-06,
+      "loss": 1.1475,
+      "step": 1171
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.760410732119777e-06,
+      "loss": 1.1631,
+      "step": 1172
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.759919478695828e-06,
+      "loss": 1.1006,
+      "step": 1173
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7594277475557956e-06,
+      "loss": 1.1309,
+      "step": 1174
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.758935538803629e-06,
+      "loss": 1.1177,
+      "step": 1175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.758442852543369e-06,
+      "loss": 1.1133,
+      "step": 1176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.757949688879167e-06,
+      "loss": 1.1523,
+      "step": 1177
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.757456047915271e-06,
+      "loss": 1.1572,
+      "step": 1178
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.756961929756027e-06,
+      "loss": 1.1172,
+      "step": 1179
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.756467334505887e-06,
+      "loss": 1.1104,
+      "step": 1180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.755972262269401e-06,
+      "loss": 1.1396,
+      "step": 1181
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.75547671315122e-06,
+      "loss": 1.1279,
+      "step": 1182
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.754980687256098e-06,
+      "loss": 1.1123,
+      "step": 1183
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.754484184688887e-06,
+      "loss": 1.124,
+      "step": 1184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7539872055545395e-06,
+      "loss": 1.1406,
+      "step": 1185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.753489749958113e-06,
+      "loss": 1.1797,
+      "step": 1186
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.752991818004761e-06,
+      "loss": 1.1133,
+      "step": 1187
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.75249340979974e-06,
+      "loss": 1.1318,
+      "step": 1188
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.751994525448406e-06,
+      "loss": 1.167,
+      "step": 1189
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.751495165056216e-06,
+      "loss": 1.1113,
+      "step": 1190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.75099532872873e-06,
+      "loss": 1.1562,
+      "step": 1191
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.750495016571604e-06,
+      "loss": 1.1191,
+      "step": 1192
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7499942286906004e-06,
+      "loss": 1.1338,
+      "step": 1193
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.749492965191576e-06,
+      "loss": 1.1455,
+      "step": 1194
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7489912261804925e-06,
+      "loss": 1.1309,
+      "step": 1195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.74848901176341e-06,
+      "loss": 1.1523,
+      "step": 1196
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.74798632204649e-06,
+      "loss": 1.1396,
+      "step": 1197
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.747483157135993e-06,
+      "loss": 1.1592,
+      "step": 1198
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.746979517138284e-06,
+      "loss": 1.1914,
+      "step": 1199
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.746475402159824e-06,
+      "loss": 1.1338,
+      "step": 1200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.745970812307174e-06,
+      "loss": 1.124,
+      "step": 1201
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.745465747687002e-06,
+      "loss": 1.1631,
+      "step": 1202
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.744960208406068e-06,
+      "loss": 1.1143,
+      "step": 1203
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.744454194571238e-06,
+      "loss": 1.1396,
+      "step": 1204
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.743947706289475e-06,
+      "loss": 1.1689,
+      "step": 1205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.743440743667843e-06,
+      "loss": 1.1548,
+      "step": 1206
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.742933306813511e-06,
+      "loss": 1.106,
+      "step": 1207
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.742425395833739e-06,
+      "loss": 1.1543,
+      "step": 1208
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.741917010835897e-06,
+      "loss": 1.1758,
+      "step": 1209
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.741408151927448e-06,
+      "loss": 1.1494,
+      "step": 1210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7408988192159585e-06,
+      "loss": 1.1309,
+      "step": 1211
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.740389012809094e-06,
+      "loss": 1.165,
+      "step": 1212
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.739878732814621e-06,
+      "loss": 1.1299,
+      "step": 1213
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.739367979340407e-06,
+      "loss": 1.1562,
+      "step": 1214
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.738856752494416e-06,
+      "loss": 1.1367,
+      "step": 1215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.738345052384715e-06,
+      "loss": 1.126,
+      "step": 1216
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.737832879119472e-06,
+      "loss": 1.1602,
+      "step": 1217
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.737320232806951e-06,
+      "loss": 1.1494,
+      "step": 1218
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.736807113555521e-06,
+      "loss": 1.124,
+      "step": 1219
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.736293521473645e-06,
+      "loss": 1.1582,
+      "step": 1220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7357794566698934e-06,
+      "loss": 1.1484,
+      "step": 1221
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.73526491925293e-06,
+      "loss": 1.1069,
+      "step": 1222
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7347499093315205e-06,
+      "loss": 1.1304,
+      "step": 1223
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.734234427014533e-06,
+      "loss": 1.1084,
+      "step": 1224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.733718472410932e-06,
+      "loss": 1.1211,
+      "step": 1225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.733202045629784e-06,
+      "loss": 1.1504,
+      "step": 1226
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7326851467802534e-06,
+      "loss": 1.127,
+      "step": 1227
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.732167775971607e-06,
+      "loss": 1.167,
+      "step": 1228
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7316499333132086e-06,
+      "loss": 1.1406,
+      "step": 1229
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.731131618914523e-06,
+      "loss": 1.1436,
+      "step": 1230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7306128328851164e-06,
+      "loss": 1.1362,
+      "step": 1231
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7300935753346525e-06,
+      "loss": 1.1226,
+      "step": 1232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.729573846372893e-06,
+      "loss": 1.1025,
+      "step": 1233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.729053646109706e-06,
+      "loss": 1.1348,
+      "step": 1234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7285329746550505e-06,
+      "loss": 1.0957,
+      "step": 1235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.72801183211899e-06,
+      "loss": 1.1445,
+      "step": 1236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.72749021861169e-06,
+      "loss": 1.1216,
+      "step": 1237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.726968134243409e-06,
+      "loss": 1.1387,
+      "step": 1238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.72644557912451e-06,
+      "loss": 1.1592,
+      "step": 1239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.725922553365454e-06,
+      "loss": 1.125,
+      "step": 1240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7253990570768004e-06,
+      "loss": 1.165,
+      "step": 1241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.724875090369211e-06,
+      "loss": 1.1211,
+      "step": 1242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.724350653353444e-06,
+      "loss": 1.0571,
+      "step": 1243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.723825746140358e-06,
+      "loss": 1.167,
+      "step": 1244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.723300368840912e-06,
+      "loss": 1.1299,
+      "step": 1245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.722774521566164e-06,
+      "loss": 1.168,
+      "step": 1246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.722248204427269e-06,
+      "loss": 1.1152,
+      "step": 1247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.721721417535485e-06,
+      "loss": 1.1084,
+      "step": 1248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.721194161002167e-06,
+      "loss": 1.1309,
+      "step": 1249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7206664349387695e-06,
+      "loss": 1.1191,
+      "step": 1250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.720138239456847e-06,
+      "loss": 1.1182,
+      "step": 1251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.719609574668052e-06,
+      "loss": 1.0986,
+      "step": 1252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.719080440684139e-06,
+      "loss": 1.1533,
+      "step": 1253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.718550837616958e-06,
+      "loss": 1.1768,
+      "step": 1254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.718020765578459e-06,
+      "loss": 1.1426,
+      "step": 1255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7174902246806946e-06,
+      "loss": 1.1035,
+      "step": 1256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7169592150358116e-06,
+      "loss": 1.1504,
+      "step": 1257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.716427736756059e-06,
+      "loss": 1.1694,
+      "step": 1258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.715895789953785e-06,
+      "loss": 1.1216,
+      "step": 1259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.715363374741434e-06,
+      "loss": 1.1191,
+      "step": 1260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.714830491231553e-06,
+      "loss": 1.1416,
+      "step": 1261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.714297139536786e-06,
+      "loss": 1.1172,
+      "step": 1262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.713763319769874e-06,
+      "loss": 1.1201,
+      "step": 1263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.713229032043662e-06,
+      "loss": 1.166,
+      "step": 1264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.712694276471091e-06,
+      "loss": 1.125,
+      "step": 1265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.712159053165199e-06,
+      "loss": 1.1387,
+      "step": 1266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7116233622391265e-06,
+      "loss": 1.1133,
+      "step": 1267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.711087203806111e-06,
+      "loss": 1.1475,
+      "step": 1268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7105505779794894e-06,
+      "loss": 1.0767,
+      "step": 1269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7100134848726955e-06,
+      "loss": 1.1562,
+      "step": 1270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.709475924599265e-06,
+      "loss": 1.1094,
+      "step": 1271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.70893789727283e-06,
+      "loss": 1.1357,
+      "step": 1272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.708399403007122e-06,
+      "loss": 1.1494,
+      "step": 1273
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7078604419159715e-06,
+      "loss": 1.1309,
+      "step": 1274
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7073210141133085e-06,
+      "loss": 1.1523,
+      "step": 1275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.70678111971316e-06,
+      "loss": 1.1436,
+      "step": 1276
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.706240758829652e-06,
+      "loss": 1.1641,
+      "step": 1277
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.705699931577009e-06,
+      "loss": 1.0986,
+      "step": 1278
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.705158638069555e-06,
+      "loss": 1.1016,
+      "step": 1279
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7046168784217126e-06,
+      "loss": 1.1191,
+      "step": 1280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7040746527480025e-06,
+      "loss": 1.1338,
+      "step": 1281
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.703531961163042e-06,
+      "loss": 1.1191,
+      "step": 1282
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.70298880378155e-06,
+      "loss": 1.1289,
+      "step": 1283
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.702445180718343e-06,
+      "loss": 1.1758,
+      "step": 1284
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.701901092088335e-06,
+      "loss": 1.1162,
+      "step": 1285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.701356538006538e-06,
+      "loss": 1.1133,
+      "step": 1286
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.700811518588066e-06,
+      "loss": 1.165,
+      "step": 1287
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.700266033948124e-06,
+      "loss": 1.1279,
+      "step": 1288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.699720084202025e-06,
+      "loss": 1.1338,
+      "step": 1289
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.699173669465171e-06,
+      "loss": 1.1597,
+      "step": 1290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.6986267898530705e-06,
+      "loss": 1.1328,
+      "step": 1291
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.698079445481323e-06,
+      "loss": 1.1611,
+      "step": 1292
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.697531636465632e-06,
+      "loss": 1.1543,
+      "step": 1293
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.696983362921797e-06,
+      "loss": 1.0977,
+      "step": 1294
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.696434624965712e-06,
+      "loss": 1.1309,
+      "step": 1295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.695885422713377e-06,
+      "loss": 1.1689,
+      "step": 1296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.695335756280883e-06,
+      "loss": 1.1465,
+      "step": 1297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.694785625784425e-06,
+      "loss": 1.1265,
+      "step": 1298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.694235031340289e-06,
+      "loss": 1.1338,
+      "step": 1299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.693683973064866e-06,
+      "loss": 1.168,
+      "step": 1300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.6931324510746415e-06,
+      "loss": 1.1738,
+      "step": 1301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.6925804654862e-06,
+      "loss": 1.1465,
+      "step": 1302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.692028016416222e-06,
+      "loss": 1.1094,
+      "step": 1303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.69147510398149e-06,
+      "loss": 1.1484,
+      "step": 1304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.690921728298881e-06,
+      "loss": 1.1846,
+      "step": 1305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.690367889485371e-06,
+      "loss": 1.1348,
+      "step": 1306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.689813587658034e-06,
+      "loss": 1.123,
+      "step": 1307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.689258822934042e-06,
+      "loss": 1.1797,
+      "step": 1308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6887035954306656e-06,
+      "loss": 1.125,
+      "step": 1309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.68814790526527e-06,
+      "loss": 1.144,
+      "step": 1310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.687591752555323e-06,
+      "loss": 1.1089,
+      "step": 1311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.687035137418386e-06,
+      "loss": 1.126,
+      "step": 1312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.68647805997212e-06,
+      "loss": 1.0898,
+      "step": 1313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.685920520334284e-06,
+      "loss": 1.1475,
+      "step": 1314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.685362518622735e-06,
+      "loss": 1.1611,
+      "step": 1315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.684804054955426e-06,
+      "loss": 1.1299,
+      "step": 1316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.684245129450409e-06,
+      "loss": 1.1328,
+      "step": 1317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.683685742225832e-06,
+      "loss": 1.1201,
+      "step": 1318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.683125893399943e-06,
+      "loss": 1.1309,
+      "step": 1319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.682565583091086e-06,
+      "loss": 1.124,
+      "step": 1320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.682004811417703e-06,
+      "loss": 1.1953,
+      "step": 1321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.681443578498333e-06,
+      "loss": 1.1284,
+      "step": 1322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6808818844516145e-06,
+      "loss": 1.1045,
+      "step": 1323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.680319729396281e-06,
+      "loss": 1.2012,
+      "step": 1324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.679757113451163e-06,
+      "loss": 1.1377,
+      "step": 1325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.679194036735193e-06,
+      "loss": 1.1582,
+      "step": 1326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.678630499367395e-06,
+      "loss": 1.1543,
+      "step": 1327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6780665014668935e-06,
+      "loss": 1.1484,
+      "step": 1328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.67750204315291e-06,
+      "loss": 1.1582,
+      "step": 1329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.676937124544765e-06,
+      "loss": 1.1211,
+      "step": 1330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.676371745761873e-06,
+      "loss": 1.1118,
+      "step": 1331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.675805906923747e-06,
+      "loss": 1.1553,
+      "step": 1332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.675239608149999e-06,
+      "loss": 1.1484,
+      "step": 1333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.674672849560335e-06,
+      "loss": 1.1201,
+      "step": 1334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.674105631274561e-06,
+      "loss": 1.1289,
+      "step": 1335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.67353795341258e-06,
+      "loss": 1.1797,
+      "step": 1336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6729698160943906e-06,
+      "loss": 1.1201,
+      "step": 1337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6724012194400885e-06,
+      "loss": 1.1631,
+      "step": 1338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.671832163569868e-06,
+      "loss": 1.1299,
+      "step": 1339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6712626486040206e-06,
+      "loss": 1.1675,
+      "step": 1340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6706926746629315e-06,
+      "loss": 1.1328,
+      "step": 1341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.670122241867088e-06,
+      "loss": 1.1426,
+      "step": 1342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.669551350337071e-06,
+      "loss": 1.1748,
+      "step": 1343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.668980000193558e-06,
+      "loss": 1.1572,
+      "step": 1344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.668408191557326e-06,
+      "loss": 1.1152,
+      "step": 1345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.667835924549247e-06,
+      "loss": 1.1562,
+      "step": 1346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.667263199290292e-06,
+      "loss": 1.1699,
+      "step": 1347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.666690015901524e-06,
+      "loss": 1.1768,
+      "step": 1348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.666116374504108e-06,
+      "loss": 1.125,
+      "step": 1349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.665542275219305e-06,
+      "loss": 1.1172,
+      "step": 1350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.66496771816847e-06,
+      "loss": 1.1553,
+      "step": 1351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.664392703473057e-06,
+      "loss": 1.1348,
+      "step": 1352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.663817231254616e-06,
+      "loss": 1.1221,
+      "step": 1353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.663241301634795e-06,
+      "loss": 1.168,
+      "step": 1354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.662664914735336e-06,
+      "loss": 1.1426,
+      "step": 1355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.662088070678081e-06,
+      "loss": 1.1182,
+      "step": 1356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6615107695849675e-06,
+      "loss": 1.123,
+      "step": 1357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6609330115780264e-06,
+      "loss": 1.1328,
+      "step": 1358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.660354796779388e-06,
+      "loss": 1.1426,
+      "step": 1359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.659776125311282e-06,
+      "loss": 1.124,
+      "step": 1360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.65919699729603e-06,
+      "loss": 1.1753,
+      "step": 1361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.65861741285605e-06,
+      "loss": 1.1582,
+      "step": 1362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.658037372113861e-06,
+      "loss": 1.1255,
+      "step": 1363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.657456875192074e-06,
+      "loss": 1.127,
+      "step": 1364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6568759222134e-06,
+      "loss": 1.1147,
+      "step": 1365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.656294513300642e-06,
+      "loss": 1.1602,
+      "step": 1366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.655712648576703e-06,
+      "loss": 1.1377,
+      "step": 1367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.655130328164581e-06,
+      "loss": 1.1436,
+      "step": 1368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.654547552187372e-06,
+      "loss": 1.1299,
+      "step": 1369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.653964320768265e-06,
+      "loss": 1.1787,
+      "step": 1370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6533806340305474e-06,
+      "loss": 1.168,
+      "step": 1371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.652796492097605e-06,
+      "loss": 1.1377,
+      "step": 1372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.652211895092913e-06,
+      "loss": 1.1133,
+      "step": 1373
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6516268431400515e-06,
+      "loss": 1.1284,
+      "step": 1374
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.651041336362689e-06,
+      "loss": 1.1582,
+      "step": 1375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.650455374884597e-06,
+      "loss": 1.1709,
+      "step": 1376
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.649868958829638e-06,
+      "loss": 1.1211,
+      "step": 1377
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.649282088321772e-06,
+      "loss": 1.1406,
+      "step": 1378
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.648694763485056e-06,
+      "loss": 1.1548,
+      "step": 1379
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.648106984443641e-06,
+      "loss": 1.1055,
+      "step": 1380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6475187513217776e-06,
+      "loss": 1.1533,
+      "step": 1381
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.64693006424381e-06,
+      "loss": 1.1416,
+      "step": 1382
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6463409233341776e-06,
+      "loss": 1.209,
+      "step": 1383
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.645751328717417e-06,
+      "loss": 1.1416,
+      "step": 1384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.64516128051816e-06,
+      "loss": 1.1592,
+      "step": 1385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.644570778861135e-06,
+      "loss": 1.1699,
+      "step": 1386
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6439798238711676e-06,
+      "loss": 1.084,
+      "step": 1387
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6433884156731755e-06,
+      "loss": 1.1357,
+      "step": 1388
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.642796554392175e-06,
+      "loss": 1.083,
+      "step": 1389
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.642204240153278e-06,
+      "loss": 1.1221,
+      "step": 1390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6416114730816906e-06,
+      "loss": 1.1826,
+      "step": 1391
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.641018253302716e-06,
+      "loss": 1.1201,
+      "step": 1392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6404245809417536e-06,
+      "loss": 1.1348,
+      "step": 1393
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6398304561242965e-06,
+      "loss": 1.0996,
+      "step": 1394
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6392358789759364e-06,
+      "loss": 1.1562,
+      "step": 1395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.638640849622356e-06,
+      "loss": 1.1709,
+      "step": 1396
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.638045368189338e-06,
+      "loss": 1.1201,
+      "step": 1397
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.637449434802759e-06,
+      "loss": 1.1709,
+      "step": 1398
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.636853049588591e-06,
+      "loss": 1.0859,
+      "step": 1399
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6362562126729016e-06,
+      "loss": 1.1055,
+      "step": 1400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.635658924181854e-06,
+      "loss": 1.0571,
+      "step": 1401
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.635061184241707e-06,
+      "loss": 1.1094,
+      "step": 1402
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.634462992978815e-06,
+      "loss": 1.166,
+      "step": 1403
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.633864350519627e-06,
+      "loss": 1.1562,
+      "step": 1404
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.633265256990687e-06,
+      "loss": 1.0938,
+      "step": 1405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6326657125186356e-06,
+      "loss": 1.1055,
+      "step": 1406
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.632065717230209e-06,
+      "loss": 1.1113,
+      "step": 1407
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.631465271252239e-06,
+      "loss": 1.0879,
+      "step": 1408
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.63086437471165e-06,
+      "loss": 1.0918,
+      "step": 1409
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.630263027735463e-06,
+      "loss": 1.1069,
+      "step": 1410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.629661230450795e-06,
+      "loss": 1.125,
+      "step": 1411
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.629058982984859e-06,
+      "loss": 1.1494,
+      "step": 1412
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.62845628546496e-06,
+      "loss": 1.1504,
+      "step": 1413
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6278531380185e-06,
+      "loss": 1.0967,
+      "step": 1414
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.627249540772979e-06,
+      "loss": 1.1348,
+      "step": 1415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.626645493855986e-06,
+      "loss": 1.1475,
+      "step": 1416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.626040997395209e-06,
+      "loss": 1.1289,
+      "step": 1417
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.625436051518432e-06,
+      "loss": 1.1484,
+      "step": 1418
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.624830656353529e-06,
+      "loss": 1.1582,
+      "step": 1419
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.624224812028476e-06,
+      "loss": 1.1504,
+      "step": 1420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.623618518671338e-06,
+      "loss": 1.1406,
+      "step": 1421
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.623011776410278e-06,
+      "loss": 1.1543,
+      "step": 1422
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.622404585373551e-06,
+      "loss": 1.1533,
+      "step": 1423
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.621796945689512e-06,
+      "loss": 1.1406,
+      "step": 1424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.621188857486606e-06,
+      "loss": 1.147,
+      "step": 1425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.620580320893374e-06,
+      "loss": 1.167,
+      "step": 1426
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6199713360384545e-06,
+      "loss": 1.1523,
+      "step": 1427
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.619361903050576e-06,
+      "loss": 1.1406,
+      "step": 1428
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.618752022058566e-06,
+      "loss": 1.2109,
+      "step": 1429
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.618141693191343e-06,
+      "loss": 1.1436,
+      "step": 1430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.617530916577926e-06,
+      "loss": 1.1143,
+      "step": 1431
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.61691969234742e-06,
+      "loss": 1.1436,
+      "step": 1432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.616308020629035e-06,
+      "loss": 1.1504,
+      "step": 1433
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.615695901552064e-06,
+      "loss": 1.1514,
+      "step": 1434
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.615083335245905e-06,
+      "loss": 1.1211,
+      "step": 1435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.614470321840045e-06,
+      "loss": 1.1318,
+      "step": 1436
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.613856861464065e-06,
+      "loss": 1.1494,
+      "step": 1437
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6132429542476435e-06,
+      "loss": 1.165,
+      "step": 1438
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.612628600320554e-06,
+      "loss": 1.0762,
+      "step": 1439
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.612013799812658e-06,
+      "loss": 1.1064,
+      "step": 1440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6113985528539204e-06,
+      "loss": 1.1279,
+      "step": 1441
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.610782859574393e-06,
+      "loss": 1.1367,
+      "step": 1442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6101667201042265e-06,
+      "loss": 1.1162,
+      "step": 1443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.609550134573664e-06,
+      "loss": 1.165,
+      "step": 1444
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.608933103113044e-06,
+      "loss": 1.168,
+      "step": 1445
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.608315625852796e-06,
+      "loss": 1.1045,
+      "step": 1446
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.607697702923448e-06,
+      "loss": 1.1187,
+      "step": 1447
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.607079334455622e-06,
+      "loss": 1.1387,
+      "step": 1448
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.606460520580029e-06,
+      "loss": 1.168,
+      "step": 1449
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.605841261427482e-06,
+      "loss": 1.1357,
+      "step": 1450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.605221557128881e-06,
+      "loss": 1.1367,
+      "step": 1451
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.604601407815223e-06,
+      "loss": 1.1523,
+      "step": 1452
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.6039808136176e-06,
+      "loss": 1.1318,
+      "step": 1453
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.6033597746671975e-06,
+      "loss": 1.1621,
+      "step": 1454
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.602738291095293e-06,
+      "loss": 1.1045,
+      "step": 1455
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.602116363033262e-06,
+      "loss": 1.1143,
+      "step": 1456
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.60149399061257e-06,
+      "loss": 1.0913,
+      "step": 1457
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.600871173964778e-06,
+      "loss": 1.1592,
+      "step": 1458
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.6002479132215415e-06,
+      "loss": 1.1211,
+      "step": 1459
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5996242085146095e-06,
+      "loss": 1.1216,
+      "step": 1460
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.599000059975824e-06,
+      "loss": 1.1367,
+      "step": 1461
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.598375467737121e-06,
+      "loss": 1.1416,
+      "step": 1462
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.597750431930532e-06,
+      "loss": 1.1953,
+      "step": 1463
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5971249526881815e-06,
+      "loss": 1.1436,
+      "step": 1464
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.596499030142285e-06,
+      "loss": 1.1377,
+      "step": 1465
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.595872664425155e-06,
+      "loss": 1.1089,
+      "step": 1466
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.595245855669197e-06,
+      "loss": 1.127,
+      "step": 1467
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.59461860400691e-06,
+      "loss": 1.1572,
+      "step": 1468
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.593990909570886e-06,
+      "loss": 1.1143,
+      "step": 1469
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.593362772493811e-06,
+      "loss": 1.1758,
+      "step": 1470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.592734192908465e-06,
+      "loss": 1.1294,
+      "step": 1471
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.59210517094772e-06,
+      "loss": 1.1182,
+      "step": 1472
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.591475706744545e-06,
+      "loss": 1.1631,
+      "step": 1473
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.590845800431998e-06,
+      "loss": 1.1162,
+      "step": 1474
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.590215452143233e-06,
+      "loss": 1.1372,
+      "step": 1475
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.589584662011497e-06,
+      "loss": 1.1387,
+      "step": 1476
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.58895343017013e-06,
+      "loss": 1.1143,
+      "step": 1477
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.588321756752567e-06,
+      "loss": 1.1372,
+      "step": 1478
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.587689641892336e-06,
+      "loss": 1.1123,
+      "step": 1479
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.587057085723054e-06,
+      "loss": 1.166,
+      "step": 1480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.586424088378437e-06,
+      "loss": 1.1367,
+      "step": 1481
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.585790649992292e-06,
+      "loss": 1.1138,
+      "step": 1482
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.585156770698519e-06,
+      "loss": 1.1426,
+      "step": 1483
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5845224506311105e-06,
+      "loss": 1.0991,
+      "step": 1484
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5838876899241544e-06,
+      "loss": 1.1074,
+      "step": 1485
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.583252488711831e-06,
+      "loss": 1.1592,
+      "step": 1486
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.582616847128411e-06,
+      "loss": 1.1572,
+      "step": 1487
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5819807653082615e-06,
+      "loss": 1.1338,
+      "step": 1488
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5813442433858415e-06,
+      "loss": 1.1079,
+      "step": 1489
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.580707281495702e-06,
+      "loss": 1.1572,
+      "step": 1490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.580069879772491e-06,
+      "loss": 1.0879,
+      "step": 1491
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.579432038350943e-06,
+      "loss": 1.168,
+      "step": 1492
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5787937573658905e-06,
+      "loss": 1.1602,
+      "step": 1493
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.578155036952258e-06,
+      "loss": 1.1221,
+      "step": 1494
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.577515877245061e-06,
+      "loss": 1.1299,
+      "step": 1495
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.57687627837941e-06,
+      "loss": 1.1353,
+      "step": 1496
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.576236240490507e-06,
+      "loss": 1.1221,
+      "step": 1497
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.575595763713647e-06,
+      "loss": 1.1133,
+      "step": 1498
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.574954848184219e-06,
+      "loss": 1.1338,
+      "step": 1499
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5743134940377036e-06,
+      "loss": 1.1494,
+      "step": 1500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.573671701409673e-06,
+      "loss": 1.1172,
+      "step": 1501
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.573029470435795e-06,
+      "loss": 1.1475,
+      "step": 1502
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.572386801251828e-06,
+      "loss": 1.1934,
+      "step": 1503
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.571743693993623e-06,
+      "loss": 1.1064,
+      "step": 1504
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.571100148797124e-06,
+      "loss": 1.1836,
+      "step": 1505
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.570456165798368e-06,
+      "loss": 1.1729,
+      "step": 1506
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.569811745133484e-06,
+      "loss": 1.2031,
+      "step": 1507
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.569166886938695e-06,
+      "loss": 1.1143,
+      "step": 1508
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.568521591350313e-06,
+      "loss": 1.1338,
+      "step": 1509
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.567875858504746e-06,
+      "loss": 1.124,
+      "step": 1510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.567229688538494e-06,
+      "loss": 1.1504,
+      "step": 1511
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.566583081588146e-06,
+      "loss": 1.1348,
+      "step": 1512
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.565936037790388e-06,
+      "loss": 1.1221,
+      "step": 1513
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.565288557281995e-06,
+      "loss": 1.1416,
+      "step": 1514
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.564640640199836e-06,
+      "loss": 1.1465,
+      "step": 1515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.563992286680872e-06,
+      "loss": 1.1602,
+      "step": 1516
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.563343496862156e-06,
+      "loss": 1.127,
+      "step": 1517
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.562694270880833e-06,
+      "loss": 1.125,
+      "step": 1518
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.562044608874141e-06,
+      "loss": 1.1143,
+      "step": 1519
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.561394510979409e-06,
+      "loss": 1.2158,
+      "step": 1520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.560743977334059e-06,
+      "loss": 1.1221,
+      "step": 1521
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.560093008075606e-06,
+      "loss": 1.1416,
+      "step": 1522
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5594416033416534e-06,
+      "loss": 1.1553,
+      "step": 1523
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.558789763269902e-06,
+      "loss": 1.1299,
+      "step": 1524
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.558137487998141e-06,
+      "loss": 1.1099,
+      "step": 1525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.557484777664252e-06,
+      "loss": 1.1123,
+      "step": 1526
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.556831632406209e-06,
+      "loss": 1.168,
+      "step": 1527
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.556178052362078e-06,
+      "loss": 1.0996,
+      "step": 1528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.555524037670018e-06,
+      "loss": 1.1362,
+      "step": 1529
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.554869588468277e-06,
+      "loss": 1.1562,
+      "step": 1530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.554214704895198e-06,
+      "loss": 1.2031,
+      "step": 1531
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.553559387089215e-06,
+      "loss": 1.1533,
+      "step": 1532
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.552903635188851e-06,
+      "loss": 1.1123,
+      "step": 1533
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5522474493327246e-06,
+      "loss": 1.1641,
+      "step": 1534
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.551590829659544e-06,
+      "loss": 1.1191,
+      "step": 1535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5509337763081095e-06,
+      "loss": 1.1289,
+      "step": 1536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.550276289417314e-06,
+      "loss": 1.1611,
+      "step": 1537
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.549618369126141e-06,
+      "loss": 1.168,
+      "step": 1538
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.548960015573664e-06,
+      "loss": 1.1128,
+      "step": 1539
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.548301228899053e-06,
+      "loss": 1.1621,
+      "step": 1540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.547642009241565e-06,
+      "loss": 1.1533,
+      "step": 1541
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.546982356740549e-06,
+      "loss": 1.1357,
+      "step": 1542
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.546322271535448e-06,
+      "loss": 1.1152,
+      "step": 1543
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.545661753765794e-06,
+      "loss": 1.1519,
+      "step": 1544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.545000803571212e-06,
+      "loss": 1.1787,
+      "step": 1545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.544339421091419e-06,
+      "loss": 1.1152,
+      "step": 1546
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.54367760646622e-06,
+      "loss": 1.1484,
+      "step": 1547
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.543015359835514e-06,
+      "loss": 1.1572,
+      "step": 1548
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.542352681339292e-06,
+      "loss": 1.1562,
+      "step": 1549
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5416895711176356e-06,
+      "loss": 1.105,
+      "step": 1550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.541026029310714e-06,
+      "loss": 1.1406,
+      "step": 1551
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.540362056058796e-06,
+      "loss": 1.1279,
+      "step": 1552
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.539697651502231e-06,
+      "loss": 1.0957,
+      "step": 1553
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5390328157814686e-06,
+      "loss": 1.1445,
+      "step": 1554
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5383675490370445e-06,
+      "loss": 1.1572,
+      "step": 1555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5377018514095885e-06,
+      "loss": 1.1475,
+      "step": 1556
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.537035723039819e-06,
+      "loss": 1.1318,
+      "step": 1557
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.536369164068546e-06,
+      "loss": 1.1328,
+      "step": 1558
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.535702174636671e-06,
+      "loss": 1.1123,
+      "step": 1559
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.535034754885187e-06,
+      "loss": 1.0996,
+      "step": 1560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5343669049551766e-06,
+      "loss": 1.1211,
+      "step": 1561
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.533698624987815e-06,
+      "loss": 1.1533,
+      "step": 1562
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.533029915124368e-06,
+      "loss": 1.1055,
+      "step": 1563
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.532360775506191e-06,
+      "loss": 1.1338,
+      "step": 1564
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.53169120627473e-06,
+      "loss": 1.1494,
+      "step": 1565
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.531021207571524e-06,
+      "loss": 1.1445,
+      "step": 1566
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5303507795382015e-06,
+      "loss": 1.1152,
+      "step": 1567
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.529679922316481e-06,
+      "loss": 1.1445,
+      "step": 1568
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.529008636048173e-06,
+      "loss": 1.1494,
+      "step": 1569
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5283369208751794e-06,
+      "loss": 1.1299,
+      "step": 1570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.52766477693949e-06,
+      "loss": 1.1523,
+      "step": 1571
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.526992204383187e-06,
+      "loss": 1.168,
+      "step": 1572
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5263192033484444e-06,
+      "loss": 1.1484,
+      "step": 1573
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.525645773977524e-06,
+      "loss": 1.0923,
+      "step": 1574
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5249719164127796e-06,
+      "loss": 1.1348,
+      "step": 1575
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.524297630796656e-06,
+      "loss": 1.1455,
+      "step": 1576
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5236229172716886e-06,
+      "loss": 1.0938,
+      "step": 1577
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5229477759805005e-06,
+      "loss": 1.1221,
+      "step": 1578
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.52227220706581e-06,
+      "loss": 1.124,
+      "step": 1579
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.521596210670421e-06,
+      "loss": 1.1543,
+      "step": 1580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5209197869372316e-06,
+      "loss": 1.1396,
+      "step": 1581
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.520242936009226e-06,
+      "loss": 1.1338,
+      "step": 1582
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.519565658029485e-06,
+      "loss": 1.106,
+      "step": 1583
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.518887953141172e-06,
+      "loss": 1.127,
+      "step": 1584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.518209821487547e-06,
+      "loss": 1.1387,
+      "step": 1585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.517531263211958e-06,
+      "loss": 1.1133,
+      "step": 1586
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.516852278457841e-06,
+      "loss": 1.1768,
+      "step": 1587
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.516172867368725e-06,
+      "loss": 1.1318,
+      "step": 1588
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.515493030088229e-06,
+      "loss": 1.1465,
+      "step": 1589
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.514812766760061e-06,
+      "loss": 1.124,
+      "step": 1590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.514132077528018e-06,
+      "loss": 1.1123,
+      "step": 1591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5134509625359904e-06,
+      "loss": 1.1514,
+      "step": 1592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.512769421927955e-06,
+      "loss": 1.1367,
+      "step": 1593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.512087455847981e-06,
+      "loss": 1.1504,
+      "step": 1594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.511405064440226e-06,
+      "loss": 1.1074,
+      "step": 1595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.51072224784894e-06,
+      "loss": 1.1143,
+      "step": 1596
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.510039006218459e-06,
+      "loss": 1.1289,
+      "step": 1597
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.509355339693212e-06,
+      "loss": 1.1367,
+      "step": 1598
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5086712484177164e-06,
+      "loss": 1.1509,
+      "step": 1599
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.50798673253658e-06,
+      "loss": 1.1572,
+      "step": 1600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5073017921945e-06,
+      "loss": 1.1455,
+      "step": 1601
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.506616427536262e-06,
+      "loss": 1.126,
+      "step": 1602
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5059306387067455e-06,
+      "loss": 1.1338,
+      "step": 1603
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.505244425850915e-06,
+      "loss": 1.165,
+      "step": 1604
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.504557789113827e-06,
+      "loss": 1.1455,
+      "step": 1605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.503870728640628e-06,
+      "loss": 1.1094,
+      "step": 1606
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.503183244576551e-06,
+      "loss": 1.0869,
+      "step": 1607
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.502495337066923e-06,
+      "loss": 1.1152,
+      "step": 1608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5018070062571575e-06,
+      "loss": 1.1289,
+      "step": 1609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.501118252292756e-06,
+      "loss": 1.1021,
+      "step": 1610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.500429075319316e-06,
+      "loss": 1.1602,
+      "step": 1611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.499739475482516e-06,
+      "loss": 1.125,
+      "step": 1612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.499049452928131e-06,
+      "loss": 1.1279,
+      "step": 1613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.498359007802021e-06,
+      "loss": 1.1328,
+      "step": 1614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.497668140250137e-06,
+      "loss": 1.1787,
+      "step": 1615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.496976850418517e-06,
+      "loss": 1.1196,
+      "step": 1616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.496285138453294e-06,
+      "loss": 1.1182,
+      "step": 1617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.495593004500683e-06,
+      "loss": 1.1602,
+      "step": 1618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.494900448706994e-06,
+      "loss": 1.1104,
+      "step": 1619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4942074712186215e-06,
+      "loss": 1.1479,
+      "step": 1620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.493514072182053e-06,
+      "loss": 1.1475,
+      "step": 1621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4928202517438655e-06,
+      "loss": 1.1162,
+      "step": 1622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.492126010050719e-06,
+      "loss": 1.1357,
+      "step": 1623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.49143134724937e-06,
+      "loss": 1.165,
+      "step": 1624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.490736263486658e-06,
+      "loss": 1.1279,
+      "step": 1625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.490040758909517e-06,
+      "loss": 1.1729,
+      "step": 1626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.489344833664965e-06,
+      "loss": 1.1143,
+      "step": 1627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.488648487900112e-06,
+      "loss": 1.1348,
+      "step": 1628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.487951721762156e-06,
+      "loss": 1.1152,
+      "step": 1629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.487254535398384e-06,
+      "loss": 1.1035,
+      "step": 1630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4865569289561714e-06,
+      "loss": 1.1396,
+      "step": 1631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.485858902582982e-06,
+      "loss": 1.127,
+      "step": 1632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4851604564263705e-06,
+      "loss": 1.165,
+      "step": 1633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.484461590633978e-06,
+      "loss": 1.1367,
+      "step": 1634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.483762305353535e-06,
+      "loss": 1.1157,
+      "step": 1635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.483062600732861e-06,
+      "loss": 1.1045,
+      "step": 1636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.482362476919864e-06,
+      "loss": 1.1377,
+      "step": 1637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4816619340625415e-06,
+      "loss": 1.1406,
+      "step": 1638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.480960972308977e-06,
+      "loss": 1.1221,
+      "step": 1639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.480259591807344e-06,
+      "loss": 1.064,
+      "step": 1640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4795577927059076e-06,
+      "loss": 1.1094,
+      "step": 1641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.478855575153016e-06,
+      "loss": 1.1719,
+      "step": 1642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.47815293929711e-06,
+      "loss": 1.1338,
+      "step": 1643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.477449885286716e-06,
+      "loss": 1.1328,
+      "step": 1644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4767464132704495e-06,
+      "loss": 1.1318,
+      "step": 1645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.476042523397017e-06,
+      "loss": 1.1123,
+      "step": 1646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4753382158152095e-06,
+      "loss": 1.1084,
+      "step": 1647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.474633490673908e-06,
+      "loss": 1.1377,
+      "step": 1648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.473928348122083e-06,
+      "loss": 1.0747,
+      "step": 1649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.473222788308791e-06,
+      "loss": 1.1382,
+      "step": 1650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4725168113831774e-06,
+      "loss": 1.1504,
+      "step": 1651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.471810417494477e-06,
+      "loss": 1.1206,
+      "step": 1652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.471103606792011e-06,
+      "loss": 1.0947,
+      "step": 1653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4703963794251895e-06,
+      "loss": 1.0713,
+      "step": 1654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.469688735543512e-06,
+      "loss": 1.1328,
+      "step": 1655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.468980675296563e-06,
+      "loss": 1.1299,
+      "step": 1656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4682721988340185e-06,
+      "loss": 1.1455,
+      "step": 1657
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.46756330630564e-06,
+      "loss": 1.0786,
+      "step": 1658
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.466853997861277e-06,
+      "loss": 1.1201,
+      "step": 1659
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.466144273650867e-06,
+      "loss": 1.147,
+      "step": 1660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.465434133824439e-06,
+      "loss": 1.125,
+      "step": 1661
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.464723578532104e-06,
+      "loss": 1.1396,
+      "step": 1662
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.464012607924065e-06,
+      "loss": 1.1035,
+      "step": 1663
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4633012221506114e-06,
+      "loss": 1.1514,
+      "step": 1664
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.462589421362119e-06,
+      "loss": 1.1631,
+      "step": 1665
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.461877205709055e-06,
+      "loss": 1.0996,
+      "step": 1666
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.461164575341971e-06,
+      "loss": 1.1074,
+      "step": 1667
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.460451530411507e-06,
+      "loss": 1.1055,
+      "step": 1668
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.459738071068391e-06,
+      "loss": 1.1055,
+      "step": 1669
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.459024197463438e-06,
+      "loss": 1.1641,
+      "step": 1670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.458309909747554e-06,
+      "loss": 1.1553,
+      "step": 1671
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.457595208071726e-06,
+      "loss": 1.0732,
+      "step": 1672
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4568800925870335e-06,
+      "loss": 1.0825,
+      "step": 1673
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.456164563444643e-06,
+      "loss": 1.1348,
+      "step": 1674
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.455448620795807e-06,
+      "loss": 1.1602,
+      "step": 1675
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.454732264791866e-06,
+      "loss": 1.1304,
+      "step": 1676
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.454015495584247e-06,
+      "loss": 1.0806,
+      "step": 1677
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4532983133244665e-06,
+      "loss": 1.1523,
+      "step": 1678
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.452580718164127e-06,
+      "loss": 1.1279,
+      "step": 1679
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.451862710254916e-06,
+      "loss": 1.1357,
+      "step": 1680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.451144289748614e-06,
+      "loss": 1.1113,
+      "step": 1681
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.450425456797084e-06,
+      "loss": 1.1592,
+      "step": 1682
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.449706211552276e-06,
+      "loss": 1.0713,
+      "step": 1683
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4489865541662305e-06,
+      "loss": 1.1396,
+      "step": 1684
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.448266484791071e-06,
+      "loss": 1.1367,
+      "step": 1685
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.447546003579014e-06,
+      "loss": 1.1406,
+      "step": 1686
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.446825110682356e-06,
+      "loss": 1.1514,
+      "step": 1687
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.446103806253485e-06,
+      "loss": 1.0796,
+      "step": 1688
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.445382090444875e-06,
+      "loss": 1.165,
+      "step": 1689
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.444659963409087e-06,
+      "loss": 1.124,
+      "step": 1690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4439374252987685e-06,
+      "loss": 1.1475,
+      "step": 1691
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.443214476266655e-06,
+      "loss": 1.1396,
+      "step": 1692
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.442491116465566e-06,
+      "loss": 1.123,
+      "step": 1693
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4417673460484125e-06,
+      "loss": 1.1182,
+      "step": 1694
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.441043165168187e-06,
+      "loss": 1.2002,
+      "step": 1695
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.440318573977973e-06,
+      "loss": 1.1494,
+      "step": 1696
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.439593572630939e-06,
+      "loss": 1.0938,
+      "step": 1697
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.438868161280342e-06,
+      "loss": 1.0996,
+      "step": 1698
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.438142340079521e-06,
+      "loss": 1.1377,
+      "step": 1699
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.437416109181905e-06,
+      "loss": 1.124,
+      "step": 1700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.436689468741012e-06,
+      "loss": 1.1328,
+      "step": 1701
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.435962418910442e-06,
+      "loss": 1.1157,
+      "step": 1702
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4352349598438834e-06,
+      "loss": 1.082,
+      "step": 1703
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.434507091695112e-06,
+      "loss": 1.1377,
+      "step": 1704
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.433778814617987e-06,
+      "loss": 1.127,
+      "step": 1705
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.433050128766458e-06,
+      "loss": 1.1484,
+      "step": 1706
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.43232103429456e-06,
+      "loss": 1.1543,
+      "step": 1707
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.431591531356412e-06,
+      "loss": 1.1426,
+      "step": 1708
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.430861620106221e-06,
+      "loss": 1.1348,
+      "step": 1709
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.43013130069828e-06,
+      "loss": 1.1582,
+      "step": 1710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.429400573286972e-06,
+      "loss": 1.1143,
+      "step": 1711
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.428669438026757e-06,
+      "loss": 1.1592,
+      "step": 1712
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.427937895072192e-06,
+      "loss": 1.1172,
+      "step": 1713
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.427205944577913e-06,
+      "loss": 1.1494,
+      "step": 1714
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.426473586698643e-06,
+      "loss": 1.1104,
+      "step": 1715
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.425740821589196e-06,
+      "loss": 1.1152,
+      "step": 1716
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.425007649404466e-06,
+      "loss": 1.1689,
+      "step": 1717
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.424274070299435e-06,
+      "loss": 1.1123,
+      "step": 1718
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.423540084429174e-06,
+      "loss": 1.0957,
+      "step": 1719
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.422805691948835e-06,
+      "loss": 1.1514,
+      "step": 1720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4220708930136615e-06,
+      "loss": 1.1113,
+      "step": 1721
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.421335687778977e-06,
+      "loss": 1.168,
+      "step": 1722
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4206000764001955e-06,
+      "loss": 1.1152,
+      "step": 1723
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.419864059032815e-06,
+      "loss": 1.1309,
+      "step": 1724
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.419127635832418e-06,
+      "loss": 1.1133,
+      "step": 1725
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.418390806954678e-06,
+      "loss": 1.1387,
+      "step": 1726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.417653572555346e-06,
+      "loss": 1.1631,
+      "step": 1727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.416915932790266e-06,
+      "loss": 1.1118,
+      "step": 1728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.416177887815365e-06,
+      "loss": 1.0977,
+      "step": 1729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.415439437786654e-06,
+      "loss": 1.1367,
+      "step": 1730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.414700582860233e-06,
+      "loss": 1.1455,
+      "step": 1731
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.413961323192285e-06,
+      "loss": 1.165,
+      "step": 1732
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.413221658939078e-06,
+      "loss": 1.1475,
+      "step": 1733
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.41248159025697e-06,
+      "loss": 1.1621,
+      "step": 1734
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.4117411173024e-06,
+      "loss": 1.0874,
+      "step": 1735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.411000240231893e-06,
+      "loss": 1.1191,
+      "step": 1736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.41025895920206e-06,
+      "loss": 1.1289,
+      "step": 1737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.4095172743696e-06,
+      "loss": 1.1045,
+      "step": 1738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.408775185891293e-06,
+      "loss": 1.1104,
+      "step": 1739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.408032693924007e-06,
+      "loss": 1.1533,
+      "step": 1740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.407289798624695e-06,
+      "loss": 1.1045,
+      "step": 1741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.406546500150394e-06,
+      "loss": 1.1689,
+      "step": 1742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.405802798658228e-06,
+      "loss": 1.1338,
+      "step": 1743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.405058694305404e-06,
+      "loss": 1.1426,
+      "step": 1744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.404314187249218e-06,
+      "loss": 1.124,
+      "step": 1745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.403569277647046e-06,
+      "loss": 1.1099,
+      "step": 1746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.402823965656354e-06,
+      "loss": 1.1631,
+      "step": 1747
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.4020782514346886e-06,
+      "loss": 1.1035,
+      "step": 1748
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.401332135139685e-06,
+      "loss": 1.1523,
+      "step": 1749
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.4005856169290615e-06,
+      "loss": 1.1133,
+      "step": 1750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.399838696960621e-06,
+      "loss": 1.0869,
+      "step": 1751
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.399091375392254e-06,
+      "loss": 1.165,
+      "step": 1752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.398343652381933e-06,
+      "loss": 1.1318,
+      "step": 1753
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.397595528087717e-06,
+      "loss": 1.1157,
+      "step": 1754
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.396847002667748e-06,
+      "loss": 1.1152,
+      "step": 1755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.396098076280256e-06,
+      "loss": 1.1396,
+      "step": 1756
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.39534874908355e-06,
+      "loss": 1.1367,
+      "step": 1757
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3945990212360325e-06,
+      "loss": 1.1113,
+      "step": 1758
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.393848892896183e-06,
+      "loss": 1.1172,
+      "step": 1759
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.393098364222568e-06,
+      "loss": 1.1729,
+      "step": 1760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.39234743537384e-06,
+      "loss": 1.1221,
+      "step": 1761
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.391596106508735e-06,
+      "loss": 1.1387,
+      "step": 1762
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.390844377786074e-06,
+      "loss": 1.1079,
+      "step": 1763
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.390092249364761e-06,
+      "loss": 1.1084,
+      "step": 1764
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.389339721403787e-06,
+      "loss": 1.1309,
+      "step": 1765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.388586794062225e-06,
+      "loss": 1.1133,
+      "step": 1766
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.387833467499234e-06,
+      "loss": 1.126,
+      "step": 1767
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3870797418740575e-06,
+      "loss": 1.1289,
+      "step": 1768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.386325617346021e-06,
+      "loss": 1.124,
+      "step": 1769
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3855710940745384e-06,
+      "loss": 1.1011,
+      "step": 1770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.384816172219104e-06,
+      "loss": 1.1211,
+      "step": 1771
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.384060851939299e-06,
+      "loss": 1.0957,
+      "step": 1772
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3833051333947876e-06,
+      "loss": 1.1543,
+      "step": 1773
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.382549016745317e-06,
+      "loss": 1.1221,
+      "step": 1774
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.381792502150721e-06,
+      "loss": 1.0957,
+      "step": 1775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.381035589770918e-06,
+      "loss": 1.1299,
+      "step": 1776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.380278279765906e-06,
+      "loss": 1.1235,
+      "step": 1777
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.379520572295771e-06,
+      "loss": 1.1445,
+      "step": 1778
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.378762467520683e-06,
+      "loss": 1.1104,
+      "step": 1779
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.378003965600895e-06,
+      "loss": 1.1377,
+      "step": 1780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.377245066696742e-06,
+      "loss": 1.0977,
+      "step": 1781
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.376485770968647e-06,
+      "loss": 1.1104,
+      "step": 1782
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3757260785771134e-06,
+      "loss": 1.0918,
+      "step": 1783
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.37496598968273e-06,
+      "loss": 1.1084,
+      "step": 1784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.37420550444617e-06,
+      "loss": 1.1133,
+      "step": 1785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.373444623028189e-06,
+      "loss": 1.1602,
+      "step": 1786
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.372683345589627e-06,
+      "loss": 1.0942,
+      "step": 1787
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.371921672291407e-06,
+      "loss": 1.1182,
+      "step": 1788
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.371159603294538e-06,
+      "loss": 1.1699,
+      "step": 1789
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.370397138760109e-06,
+      "loss": 1.1543,
+      "step": 1790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.369634278849295e-06,
+      "loss": 1.0796,
+      "step": 1791
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.368871023723355e-06,
+      "loss": 1.1211,
+      "step": 1792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.368107373543631e-06,
+      "loss": 1.1709,
+      "step": 1793
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.367343328471547e-06,
+      "loss": 1.1582,
+      "step": 1794
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3665788886686126e-06,
+      "loss": 1.1426,
+      "step": 1795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.365814054296419e-06,
+      "loss": 1.0952,
+      "step": 1796
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.365048825516642e-06,
+      "loss": 1.1094,
+      "step": 1797
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.364283202491042e-06,
+      "loss": 1.1387,
+      "step": 1798
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.363517185381459e-06,
+      "loss": 1.1064,
+      "step": 1799
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.36275077434982e-06,
+      "loss": 1.0996,
+      "step": 1800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3619839695581326e-06,
+      "loss": 1.1211,
+      "step": 1801
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.361216771168491e-06,
+      "loss": 1.1465,
+      "step": 1802
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.360449179343068e-06,
+      "loss": 1.1172,
+      "step": 1803
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.359681194244123e-06,
+      "loss": 1.1523,
+      "step": 1804
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.358912816033998e-06,
+      "loss": 1.0898,
+      "step": 1805
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.358144044875118e-06,
+      "loss": 1.1152,
+      "step": 1806
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.357374880929988e-06,
+      "loss": 1.166,
+      "step": 1807
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.356605324361203e-06,
+      "loss": 1.1069,
+      "step": 1808
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.355835375331433e-06,
+      "loss": 1.1338,
+      "step": 1809
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3550650340034364e-06,
+      "loss": 1.1455,
+      "step": 1810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.354294300540053e-06,
+      "loss": 1.1943,
+      "step": 1811
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.353523175104205e-06,
+      "loss": 1.1289,
+      "step": 1812
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.352751657858897e-06,
+      "loss": 1.1279,
+      "step": 1813
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.351979748967219e-06,
+      "loss": 1.1211,
+      "step": 1814
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.351207448592339e-06,
+      "loss": 1.1465,
+      "step": 1815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.350434756897515e-06,
+      "loss": 1.1504,
+      "step": 1816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.349661674046079e-06,
+      "loss": 1.1143,
+      "step": 1817
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.348888200201454e-06,
+      "loss": 1.1523,
+      "step": 1818
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.348114335527138e-06,
+      "loss": 1.1343,
+      "step": 1819
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3473400801867195e-06,
+      "loss": 1.1484,
+      "step": 1820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.346565434343862e-06,
+      "loss": 1.1182,
+      "step": 1821
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.345790398162317e-06,
+      "loss": 1.1748,
+      "step": 1822
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.345014971805916e-06,
+      "loss": 1.1367,
+      "step": 1823
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.344239155438573e-06,
+      "loss": 1.0894,
+      "step": 1824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.343462949224286e-06,
+      "loss": 1.1523,
+      "step": 1825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.342686353327135e-06,
+      "loss": 1.1182,
+      "step": 1826
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.341909367911279e-06,
+      "loss": 1.1455,
+      "step": 1827
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.341131993140964e-06,
+      "loss": 1.1318,
+      "step": 1828
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.340354229180517e-06,
+      "loss": 1.1328,
+      "step": 1829
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.339576076194346e-06,
+      "loss": 1.1514,
+      "step": 1830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.33879753434694e-06,
+      "loss": 1.0884,
+      "step": 1831
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.338018603802875e-06,
+      "loss": 1.1465,
+      "step": 1832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3372392847268055e-06,
+      "loss": 1.1328,
+      "step": 1833
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.336459577283468e-06,
+      "loss": 1.0659,
+      "step": 1834
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.335679481637682e-06,
+      "loss": 1.1426,
+      "step": 1835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.334898997954351e-06,
+      "loss": 1.1289,
+      "step": 1836
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.334118126398456e-06,
+      "loss": 1.1167,
+      "step": 1837
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.333336867135064e-06,
+      "loss": 1.1357,
+      "step": 1838
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.332555220329322e-06,
+      "loss": 1.1201,
+      "step": 1839
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.331773186146459e-06,
+      "loss": 1.1221,
+      "step": 1840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.330990764751788e-06,
+      "loss": 1.1045,
+      "step": 1841
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.330207956310701e-06,
+      "loss": 1.1123,
+      "step": 1842
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.329424760988673e-06,
+      "loss": 1.1045,
+      "step": 1843
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.328641178951262e-06,
+      "loss": 1.1436,
+      "step": 1844
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.327857210364104e-06,
+      "loss": 1.1475,
+      "step": 1845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.327072855392922e-06,
+      "loss": 1.1504,
+      "step": 1846
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.326288114203516e-06,
+      "loss": 1.0977,
+      "step": 1847
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3255029869617694e-06,
+      "loss": 1.1621,
+      "step": 1848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3247174738336494e-06,
+      "loss": 1.1426,
+      "step": 1849
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.323931574985202e-06,
+      "loss": 1.1357,
+      "step": 1850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.323145290582554e-06,
+      "loss": 1.0654,
+      "step": 1851
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.322358620791916e-06,
+      "loss": 1.1689,
+      "step": 1852
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.32157156577958e-06,
+      "loss": 1.1719,
+      "step": 1853
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.320784125711918e-06,
+      "loss": 1.1348,
+      "step": 1854
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.319996300755384e-06,
+      "loss": 1.1328,
+      "step": 1855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.319208091076513e-06,
+      "loss": 1.1309,
+      "step": 1856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.318419496841923e-06,
+      "loss": 1.1279,
+      "step": 1857
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.317630518218312e-06,
+      "loss": 1.1484,
+      "step": 1858
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3168411553724575e-06,
+      "loss": 1.1279,
+      "step": 1859
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.316051408471222e-06,
+      "loss": 1.083,
+      "step": 1860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3152612776815465e-06,
+      "loss": 1.1123,
+      "step": 1861
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.314470763170453e-06,
+      "loss": 1.1123,
+      "step": 1862
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.313679865105046e-06,
+      "loss": 1.126,
+      "step": 1863
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3128885836525114e-06,
+      "loss": 1.0688,
+      "step": 1864
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.312096918980114e-06,
+      "loss": 1.0889,
+      "step": 1865
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.311304871255202e-06,
+      "loss": 1.1318,
+      "step": 1866
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3105124406452035e-06,
+      "loss": 1.1045,
+      "step": 1867
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.309719627317625e-06,
+      "loss": 1.127,
+      "step": 1868
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.30892643144006e-06,
+      "loss": 1.1357,
+      "step": 1869
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.308132853180177e-06,
+      "loss": 1.1191,
+      "step": 1870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3073388927057295e-06,
+      "loss": 1.1396,
+      "step": 1871
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.306544550184547e-06,
+      "loss": 1.1157,
+      "step": 1872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.305749825784545e-06,
+      "loss": 1.1177,
+      "step": 1873
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3049547196737165e-06,
+      "loss": 1.1035,
+      "step": 1874
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.304159232020135e-06,
+      "loss": 1.1143,
+      "step": 1875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3033633629919595e-06,
+      "loss": 1.1699,
+      "step": 1876
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3025671127574205e-06,
+      "loss": 1.1211,
+      "step": 1877
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.301770481484838e-06,
+      "loss": 1.1226,
+      "step": 1878
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.300973469342609e-06,
+      "loss": 1.1475,
+      "step": 1879
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.30017607649921e-06,
+      "loss": 1.1367,
+      "step": 1880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.299378303123198e-06,
+      "loss": 1.1396,
+      "step": 1881
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.298580149383213e-06,
+      "loss": 1.1357,
+      "step": 1882
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2977816154479725e-06,
+      "loss": 1.1255,
+      "step": 1883
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.296982701486278e-06,
+      "loss": 1.1514,
+      "step": 1884
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2961834076670055e-06,
+      "loss": 1.1436,
+      "step": 1885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.295383734159118e-06,
+      "loss": 1.1719,
+      "step": 1886
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.294583681131653e-06,
+      "loss": 1.1201,
+      "step": 1887
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.293783248753732e-06,
+      "loss": 1.1367,
+      "step": 1888
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.292982437194556e-06,
+      "loss": 1.1055,
+      "step": 1889
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.292181246623405e-06,
+      "loss": 1.0981,
+      "step": 1890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.291379677209641e-06,
+      "loss": 1.1016,
+      "step": 1891
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.290577729122701e-06,
+      "loss": 1.1147,
+      "step": 1892
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.289775402532111e-06,
+      "loss": 1.1631,
+      "step": 1893
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.288972697607467e-06,
+      "loss": 1.103,
+      "step": 1894
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.288169614518453e-06,
+      "loss": 1.1045,
+      "step": 1895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.287366153434829e-06,
+      "loss": 1.1533,
+      "step": 1896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.286562314526435e-06,
+      "loss": 1.1162,
+      "step": 1897
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.285758097963194e-06,
+      "loss": 1.1021,
+      "step": 1898
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.284953503915104e-06,
+      "loss": 1.1104,
+      "step": 1899
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.284148532552245e-06,
+      "loss": 1.167,
+      "step": 1900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.283343184044777e-06,
+      "loss": 1.1689,
+      "step": 1901
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.282537458562942e-06,
+      "loss": 1.1055,
+      "step": 1902
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.281731356277057e-06,
+      "loss": 1.165,
+      "step": 1903
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.280924877357521e-06,
+      "loss": 1.1436,
+      "step": 1904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.280118021974814e-06,
+      "loss": 1.123,
+      "step": 1905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.279310790299493e-06,
+      "loss": 1.1963,
+      "step": 1906
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.278503182502197e-06,
+      "loss": 1.1309,
+      "step": 1907
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.277695198753643e-06,
+      "loss": 1.1631,
+      "step": 1908
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.276886839224626e-06,
+      "loss": 1.1514,
+      "step": 1909
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2760781040860245e-06,
+      "loss": 1.1152,
+      "step": 1910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.275268993508792e-06,
+      "loss": 1.1719,
+      "step": 1911
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.274459507663965e-06,
+      "loss": 1.1504,
+      "step": 1912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.273649646722656e-06,
+      "loss": 1.1748,
+      "step": 1913
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.272839410856061e-06,
+      "loss": 1.1357,
+      "step": 1914
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.272028800235451e-06,
+      "loss": 1.1465,
+      "step": 1915
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.271217815032178e-06,
+      "loss": 1.127,
+      "step": 1916
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.270406455417674e-06,
+      "loss": 1.1523,
+      "step": 1917
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.269594721563448e-06,
+      "loss": 1.1279,
+      "step": 1918
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.268782613641091e-06,
+      "loss": 1.1147,
+      "step": 1919
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2679701318222714e-06,
+      "loss": 1.1572,
+      "step": 1920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.267157276278736e-06,
+      "loss": 1.126,
+      "step": 1921
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2663440471823115e-06,
+      "loss": 1.0889,
+      "step": 1922
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.265530444704903e-06,
+      "loss": 1.126,
+      "step": 1923
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.264716469018496e-06,
+      "loss": 1.1455,
+      "step": 1924
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.263902120295154e-06,
+      "loss": 1.1064,
+      "step": 1925
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.263087398707017e-06,
+      "loss": 1.0986,
+      "step": 1926
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.262272304426308e-06,
+      "loss": 1.1343,
+      "step": 1927
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.261456837625326e-06,
+      "loss": 1.0889,
+      "step": 1928
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2606409984764494e-06,
+      "loss": 1.1348,
+      "step": 1929
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.259824787152137e-06,
+      "loss": 1.1328,
+      "step": 1930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.259008203824923e-06,
+      "loss": 1.1729,
+      "step": 1931
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.258191248667423e-06,
+      "loss": 1.1318,
+      "step": 1932
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2573739218523284e-06,
+      "loss": 1.1357,
+      "step": 1933
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.256556223552413e-06,
+      "loss": 1.105,
+      "step": 1934
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.255738153940526e-06,
+      "loss": 1.1475,
+      "step": 1935
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.254919713189596e-06,
+      "loss": 1.1309,
+      "step": 1936
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.254100901472631e-06,
+      "loss": 1.1401,
+      "step": 1937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.253281718962716e-06,
+      "loss": 1.1064,
+      "step": 1938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.252462165833015e-06,
+      "loss": 1.1123,
+      "step": 1939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.25164224225677e-06,
+      "loss": 1.1104,
+      "step": 1940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.250821948407303e-06,
+      "loss": 1.1406,
+      "step": 1941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.250001284458012e-06,
+      "loss": 1.0869,
+      "step": 1942
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.249180250582373e-06,
+      "loss": 1.1025,
+      "step": 1943
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.248358846953944e-06,
+      "loss": 1.1113,
+      "step": 1944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.247537073746355e-06,
+      "loss": 1.1377,
+      "step": 1945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.246714931133321e-06,
+      "loss": 1.1064,
+      "step": 1946
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.24589241928863e-06,
+      "loss": 1.125,
+      "step": 1947
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.245069538386149e-06,
+      "loss": 1.1011,
+      "step": 1948
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2442462885998265e-06,
+      "loss": 1.1182,
+      "step": 1949
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.243422670103684e-06,
+      "loss": 1.1465,
+      "step": 1950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.242598683071823e-06,
+      "loss": 1.1064,
+      "step": 1951
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.241774327678424e-06,
+      "loss": 1.0889,
+      "step": 1952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2409496040977445e-06,
+      "loss": 1.1045,
+      "step": 1953
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.24012451250412e-06,
+      "loss": 1.1045,
+      "step": 1954
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.239299053071963e-06,
+      "loss": 1.124,
+      "step": 1955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.238473225975764e-06,
+      "loss": 1.1436,
+      "step": 1956
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.237647031390093e-06,
+      "loss": 1.1504,
+      "step": 1957
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.236820469489595e-06,
+      "loss": 1.1201,
+      "step": 1958
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.235993540448993e-06,
+      "loss": 1.1318,
+      "step": 1959
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.235166244443091e-06,
+      "loss": 1.0918,
+      "step": 1960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.234338581646767e-06,
+      "loss": 1.1426,
+      "step": 1961
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.233510552234976e-06,
+      "loss": 1.1396,
+      "step": 1962
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.232682156382754e-06,
+      "loss": 1.1172,
+      "step": 1963
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.231853394265212e-06,
+      "loss": 1.1011,
+      "step": 1964
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.231024266057539e-06,
+      "loss": 1.1553,
+      "step": 1965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.230194771935002e-06,
+      "loss": 1.1182,
+      "step": 1966
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.229364912072943e-06,
+      "loss": 1.1309,
+      "step": 1967
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.228534686646785e-06,
+      "loss": 1.0938,
+      "step": 1968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.227704095832025e-06,
+      "loss": 1.1338,
+      "step": 1969
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.226873139804239e-06,
+      "loss": 1.1333,
+      "step": 1970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.226041818739079e-06,
+      "loss": 1.0742,
+      "step": 1971
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.225210132812277e-06,
+      "loss": 1.1318,
+      "step": 1972
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2243780821996385e-06,
+      "loss": 1.1582,
+      "step": 1973
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.223545667077047e-06,
+      "loss": 1.1133,
+      "step": 1974
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2227128876204644e-06,
+      "loss": 1.1289,
+      "step": 1975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.221879744005929e-06,
+      "loss": 1.1289,
+      "step": 1976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.221046236409556e-06,
+      "loss": 1.1543,
+      "step": 1977
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2202123650075374e-06,
+      "loss": 1.1045,
+      "step": 1978
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.219378129976142e-06,
+      "loss": 1.125,
+      "step": 1979
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.218543531491717e-06,
+      "loss": 1.1177,
+      "step": 1980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.217708569730683e-06,
+      "loss": 1.1592,
+      "step": 1981
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.216873244869541e-06,
+      "loss": 1.0854,
+      "step": 1982
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2160375570848665e-06,
+      "loss": 1.1357,
+      "step": 1983
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2152015065533125e-06,
+      "loss": 1.1182,
+      "step": 1984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.214365093451609e-06,
+      "loss": 1.1123,
+      "step": 1985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.213528317956563e-06,
+      "loss": 1.1104,
+      "step": 1986
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.212691180245056e-06,
+      "loss": 1.1104,
+      "step": 1987
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.211853680494048e-06,
+      "loss": 1.1318,
+      "step": 1988
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.211015818880575e-06,
+      "loss": 1.0996,
+      "step": 1989
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.210177595581749e-06,
+      "loss": 1.1377,
+      "step": 1990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.209339010774761e-06,
+      "loss": 1.1201,
+      "step": 1991
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.208500064636874e-06,
+      "loss": 1.1543,
+      "step": 1992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2076607573454304e-06,
+      "loss": 1.1084,
+      "step": 1993
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.206821089077848e-06,
+      "loss": 1.1396,
+      "step": 1994
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2059810600116225e-06,
+      "loss": 1.1338,
+      "step": 1995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.205140670324324e-06,
+      "loss": 1.1621,
+      "step": 1996
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.204299920193599e-06,
+      "loss": 1.125,
+      "step": 1997
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.20345880979717e-06,
+      "loss": 1.1182,
+      "step": 1998
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.202617339312838e-06,
+      "loss": 1.1475,
+      "step": 1999
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.201775508918477e-06,
+      "loss": 1.1738,
+      "step": 2000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2009333187920375e-06,
+      "loss": 1.1045,
+      "step": 2001
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.20009076911155e-06,
+      "loss": 1.0596,
+      "step": 2002
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.199247860055115e-06,
+      "loss": 1.1172,
+      "step": 2003
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.198404591800913e-06,
+      "loss": 1.147,
+      "step": 2004
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.197560964527201e-06,
+      "loss": 1.0967,
+      "step": 2005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.196716978412307e-06,
+      "loss": 1.1182,
+      "step": 2006
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.195872633634641e-06,
+      "loss": 1.1304,
+      "step": 2007
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.195027930372685e-06,
+      "loss": 1.0908,
+      "step": 2008
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.194182868804997e-06,
+      "loss": 1.1475,
+      "step": 2009
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.193337449110213e-06,
+      "loss": 1.1533,
+      "step": 2010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.192491671467041e-06,
+      "loss": 1.1436,
+      "step": 2011
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.191645536054268e-06,
+      "loss": 1.1133,
+      "step": 2012
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.190799043050757e-06,
+      "loss": 1.1465,
+      "step": 2013
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.189952192635443e-06,
+      "loss": 1.1313,
+      "step": 2014
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.189104984987339e-06,
+      "loss": 1.1484,
+      "step": 2015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1882574202855334e-06,
+      "loss": 1.1025,
+      "step": 2016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.18740949870919e-06,
+      "loss": 1.1318,
+      "step": 2017
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.186561220437546e-06,
+      "loss": 1.1455,
+      "step": 2018
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.185712585649919e-06,
+      "loss": 1.1328,
+      "step": 2019
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.184863594525697e-06,
+      "loss": 1.0996,
+      "step": 2020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.184014247244344e-06,
+      "loss": 1.1279,
+      "step": 2021
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.183164543985402e-06,
+      "loss": 1.1299,
+      "step": 2022
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.182314484928487e-06,
+      "loss": 1.1147,
+      "step": 2023
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.181464070253289e-06,
+      "loss": 1.1328,
+      "step": 2024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.180613300139575e-06,
+      "loss": 1.1279,
+      "step": 2025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1797621747671836e-06,
+      "loss": 1.1406,
+      "step": 2026
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.178910694316034e-06,
+      "loss": 1.1338,
+      "step": 2027
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.178058858966116e-06,
+      "loss": 1.1084,
+      "step": 2028
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.177206668897496e-06,
+      "loss": 1.0859,
+      "step": 2029
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.176354124290316e-06,
+      "loss": 1.0977,
+      "step": 2030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.175501225324792e-06,
+      "loss": 1.1172,
+      "step": 2031
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.174647972181215e-06,
+      "loss": 1.1177,
+      "step": 2032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.173794365039951e-06,
+      "loss": 1.1406,
+      "step": 2033
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.172940404081441e-06,
+      "loss": 1.1631,
+      "step": 2034
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.172086089486201e-06,
+      "loss": 1.1084,
+      "step": 2035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.171231421434821e-06,
+      "loss": 1.1445,
+      "step": 2036
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.170376400107966e-06,
+      "loss": 1.1377,
+      "step": 2037
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.169521025686377e-06,
+      "loss": 1.168,
+      "step": 2038
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.168665298350867e-06,
+      "loss": 1.0713,
+      "step": 2039
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.167809218282325e-06,
+      "loss": 1.1123,
+      "step": 2040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.166952785661716e-06,
+      "loss": 1.0933,
+      "step": 2041
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.166096000670077e-06,
+      "loss": 1.1162,
+      "step": 2042
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.165238863488521e-06,
+      "loss": 1.124,
+      "step": 2043
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.164381374298235e-06,
+      "loss": 1.1143,
+      "step": 2044
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.163523533280481e-06,
+      "loss": 1.2021,
+      "step": 2045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.162665340616594e-06,
+      "loss": 1.1377,
+      "step": 2046
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.161806796487985e-06,
+      "loss": 1.1309,
+      "step": 2047
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.160947901076139e-06,
+      "loss": 1.1143,
+      "step": 2048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.160088654562613e-06,
+      "loss": 1.1494,
+      "step": 2049
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.15922905712904e-06,
+      "loss": 1.1387,
+      "step": 2050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.158369108957129e-06,
+      "loss": 1.1299,
+      "step": 2051
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.15750881022866e-06,
+      "loss": 1.1641,
+      "step": 2052
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.156648161125489e-06,
+      "loss": 1.124,
+      "step": 2053
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.155787161829544e-06,
+      "loss": 1.1494,
+      "step": 2054
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.154925812522831e-06,
+      "loss": 1.1416,
+      "step": 2055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.154064113387426e-06,
+      "loss": 1.126,
+      "step": 2056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.15320206460548e-06,
+      "loss": 1.1396,
+      "step": 2057
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.152339666359219e-06,
+      "loss": 1.1509,
+      "step": 2058
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.151476918830941e-06,
+      "loss": 1.1006,
+      "step": 2059
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.150613822203021e-06,
+      "loss": 1.1055,
+      "step": 2060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.149750376657905e-06,
+      "loss": 1.1445,
+      "step": 2061
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.148886582378113e-06,
+      "loss": 1.1592,
+      "step": 2062
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.14802243954624e-06,
+      "loss": 1.1445,
+      "step": 2063
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.147157948344954e-06,
+      "loss": 1.1592,
+      "step": 2064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.146293108956996e-06,
+      "loss": 1.1348,
+      "step": 2065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.145427921565182e-06,
+      "loss": 1.1494,
+      "step": 2066
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1445623863524e-06,
+      "loss": 1.1592,
+      "step": 2067
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1436965035016124e-06,
+      "loss": 1.127,
+      "step": 2068
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.142830273195856e-06,
+      "loss": 1.0947,
+      "step": 2069
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.141963695618239e-06,
+      "loss": 1.0957,
+      "step": 2070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.141096770951945e-06,
+      "loss": 1.0825,
+      "step": 2071
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.140229499380228e-06,
+      "loss": 1.124,
+      "step": 2072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.139361881086419e-06,
+      "loss": 1.0776,
+      "step": 2073
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.138493916253921e-06,
+      "loss": 1.1182,
+      "step": 2074
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.13762560506621e-06,
+      "loss": 1.1367,
+      "step": 2075
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1367569477068335e-06,
+      "loss": 1.1187,
+      "step": 2076
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.135887944359415e-06,
+      "loss": 1.1143,
+      "step": 2077
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.135018595207649e-06,
+      "loss": 1.1123,
+      "step": 2078
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1341489004353045e-06,
+      "loss": 1.1396,
+      "step": 2079
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1332788602262246e-06,
+      "loss": 1.1445,
+      "step": 2080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1324084747643215e-06,
+      "loss": 1.1079,
+      "step": 2081
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1315377442335844e-06,
+      "loss": 1.1182,
+      "step": 2082
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.130666668818073e-06,
+      "loss": 1.0864,
+      "step": 2083
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.129795248701921e-06,
+      "loss": 1.1475,
+      "step": 2084
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.128923484069335e-06,
+      "loss": 1.0879,
+      "step": 2085
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.128051375104593e-06,
+      "loss": 1.1475,
+      "step": 2086
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1271789219920485e-06,
+      "loss": 1.1689,
+      "step": 2087
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.126306124916125e-06,
+      "loss": 1.1553,
+      "step": 2088
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.125432984061321e-06,
+      "loss": 1.1553,
+      "step": 2089
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.124559499612205e-06,
+      "loss": 1.1348,
+      "step": 2090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.123685671753421e-06,
+      "loss": 1.1338,
+      "step": 2091
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.122811500669683e-06,
+      "loss": 1.1631,
+      "step": 2092
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.12193698654578e-06,
+      "loss": 1.1426,
+      "step": 2093
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.121062129566572e-06,
+      "loss": 1.1309,
+      "step": 2094
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.120186929916991e-06,
+      "loss": 1.126,
+      "step": 2095
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.119311387782045e-06,
+      "loss": 1.1338,
+      "step": 2096
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1184355033468074e-06,
+      "loss": 1.1416,
+      "step": 2097
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1175592767964315e-06,
+      "loss": 1.1309,
+      "step": 2098
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1166827083161375e-06,
+      "loss": 1.1123,
+      "step": 2099
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.115805798091222e-06,
+      "loss": 1.1436,
+      "step": 2100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1149285463070505e-06,
+      "loss": 1.1836,
+      "step": 2101
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.114050953149062e-06,
+      "loss": 1.1279,
+      "step": 2102
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.113173018802768e-06,
+      "loss": 1.1338,
+      "step": 2103
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.112294743453752e-06,
+      "loss": 1.1196,
+      "step": 2104
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.111416127287669e-06,
+      "loss": 1.1187,
+      "step": 2105
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.110537170490247e-06,
+      "loss": 1.1387,
+      "step": 2106
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.109657873247286e-06,
+      "loss": 1.1143,
+      "step": 2107
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.108778235744656e-06,
+      "loss": 1.1465,
+      "step": 2108
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1078982581683e-06,
+      "loss": 1.1533,
+      "step": 2109
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.107017940704235e-06,
+      "loss": 1.1108,
+      "step": 2110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.106137283538547e-06,
+      "loss": 1.1592,
+      "step": 2111
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.105256286857395e-06,
+      "loss": 1.0918,
+      "step": 2112
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.10437495084701e-06,
+      "loss": 1.166,
+      "step": 2113
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.103493275693695e-06,
+      "loss": 1.1875,
+      "step": 2114
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.102611261583822e-06,
+      "loss": 1.104,
+      "step": 2115
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.101728908703838e-06,
+      "loss": 1.0996,
+      "step": 2116
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.100846217240261e-06,
+      "loss": 1.1191,
+      "step": 2117
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.099963187379678e-06,
+      "loss": 1.1006,
+      "step": 2118
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.099079819308751e-06,
+      "loss": 1.1113,
+      "step": 2119
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.098196113214213e-06,
+      "loss": 1.125,
+      "step": 2120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0973120692828635e-06,
+      "loss": 1.1445,
+      "step": 2121
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.096427687701581e-06,
+      "loss": 1.1348,
+      "step": 2122
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.095542968657311e-06,
+      "loss": 1.1094,
+      "step": 2123
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.09465791233707e-06,
+      "loss": 1.1045,
+      "step": 2124
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.093772518927947e-06,
+      "loss": 1.1836,
+      "step": 2125
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.092886788617103e-06,
+      "loss": 1.1699,
+      "step": 2126
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.092000721591768e-06,
+      "loss": 1.0996,
+      "step": 2127
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.091114318039244e-06,
+      "loss": 1.124,
+      "step": 2128
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.090227578146907e-06,
+      "loss": 1.1006,
+      "step": 2129
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0893405021022e-06,
+      "loss": 1.0957,
+      "step": 2130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.088453090092639e-06,
+      "loss": 1.1445,
+      "step": 2131
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.087565342305811e-06,
+      "loss": 1.165,
+      "step": 2132
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0866772589293735e-06,
+      "loss": 1.1123,
+      "step": 2133
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.085788840151055e-06,
+      "loss": 1.1621,
+      "step": 2134
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.084900086158655e-06,
+      "loss": 1.106,
+      "step": 2135
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.084010997140045e-06,
+      "loss": 1.0869,
+      "step": 2136
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.083121573283167e-06,
+      "loss": 1.124,
+      "step": 2137
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.08223181477603e-06,
+      "loss": 1.1582,
+      "step": 2138
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.081341721806718e-06,
+      "loss": 1.0918,
+      "step": 2139
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.080451294563384e-06,
+      "loss": 1.1616,
+      "step": 2140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.079560533234255e-06,
+      "loss": 1.1191,
+      "step": 2141
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.078669438007623e-06,
+      "loss": 1.1143,
+      "step": 2142
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.077778009071854e-06,
+      "loss": 1.1201,
+      "step": 2143
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.076886246615385e-06,
+      "loss": 1.0732,
+      "step": 2144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.075994150826721e-06,
+      "loss": 1.1157,
+      "step": 2145
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.075101721894438e-06,
+      "loss": 1.1543,
+      "step": 2146
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.074208960007186e-06,
+      "loss": 1.1279,
+      "step": 2147
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.07331586535368e-06,
+      "loss": 1.1406,
+      "step": 2148
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.07242243812271e-06,
+      "loss": 1.1035,
+      "step": 2149
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.071528678503134e-06,
+      "loss": 1.1016,
+      "step": 2150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0706345866838795e-06,
+      "loss": 1.1133,
+      "step": 2151
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0697401628539456e-06,
+      "loss": 1.1191,
+      "step": 2152
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.068845407202401e-06,
+      "loss": 1.1436,
+      "step": 2153
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.067950319918386e-06,
+      "loss": 1.1133,
+      "step": 2154
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0670549011911095e-06,
+      "loss": 1.166,
+      "step": 2155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.066159151209851e-06,
+      "loss": 1.1426,
+      "step": 2156
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0652630701639576e-06,
+      "loss": 1.1338,
+      "step": 2157
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.064366658242851e-06,
+      "loss": 1.166,
+      "step": 2158
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.063469915636019e-06,
+      "loss": 1.1133,
+      "step": 2159
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.062572842533023e-06,
+      "loss": 1.1377,
+      "step": 2160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.061675439123488e-06,
+      "loss": 1.1865,
+      "step": 2161
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.060777705597117e-06,
+      "loss": 1.1299,
+      "step": 2162
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.059879642143676e-06,
+      "loss": 1.1074,
+      "step": 2163
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0589812489530046e-06,
+      "loss": 1.1196,
+      "step": 2164
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.05808252621501e-06,
+      "loss": 1.1191,
+      "step": 2165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.05718347411967e-06,
+      "loss": 1.1455,
+      "step": 2166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.056284092857032e-06,
+      "loss": 1.1914,
+      "step": 2167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.055384382617214e-06,
+      "loss": 1.1777,
+      "step": 2168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.054484343590401e-06,
+      "loss": 1.1455,
+      "step": 2169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.05358397596685e-06,
+      "loss": 1.0728,
+      "step": 2170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.052683279936885e-06,
+      "loss": 1.0957,
+      "step": 2171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0517822556909015e-06,
+      "loss": 1.1035,
+      "step": 2172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.050880903419365e-06,
+      "loss": 1.1377,
+      "step": 2173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0499792233128075e-06,
+      "loss": 1.0684,
+      "step": 2174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.049077215561832e-06,
+      "loss": 1.1294,
+      "step": 2175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.048174880357111e-06,
+      "loss": 1.0977,
+      "step": 2176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.047272217889385e-06,
+      "loss": 1.1475,
+      "step": 2177
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.046369228349466e-06,
+      "loss": 1.1104,
+      "step": 2178
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.045465911928233e-06,
+      "loss": 1.1348,
+      "step": 2179
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.044562268816635e-06,
+      "loss": 1.1377,
+      "step": 2180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.043658299205688e-06,
+      "loss": 1.1328,
+      "step": 2181
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0427540032864816e-06,
+      "loss": 1.1006,
+      "step": 2182
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.041849381250169e-06,
+      "loss": 1.1104,
+      "step": 2183
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.040944433287977e-06,
+      "loss": 1.165,
+      "step": 2184
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.040039159591198e-06,
+      "loss": 1.1641,
+      "step": 2185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.039133560351196e-06,
+      "loss": 1.124,
+      "step": 2186
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0382276357594e-06,
+      "loss": 1.127,
+      "step": 2187
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.037321386007311e-06,
+      "loss": 1.1328,
+      "step": 2188
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.036414811286498e-06,
+      "loss": 1.1133,
+      "step": 2189
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0355079117886e-06,
+      "loss": 1.127,
+      "step": 2190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.034600687705321e-06,
+      "loss": 1.1387,
+      "step": 2191
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.033693139228436e-06,
+      "loss": 1.0996,
+      "step": 2192
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.032785266549789e-06,
+      "loss": 1.1196,
+      "step": 2193
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.031877069861292e-06,
+      "loss": 1.1299,
+      "step": 2194
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.030968549354925e-06,
+      "loss": 1.0903,
+      "step": 2195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.030059705222736e-06,
+      "loss": 1.1104,
+      "step": 2196
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.029150537656844e-06,
+      "loss": 1.1406,
+      "step": 2197
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.028241046849433e-06,
+      "loss": 1.1104,
+      "step": 2198
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.027331232992758e-06,
+      "loss": 1.1738,
+      "step": 2199
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.026421096279141e-06,
+      "loss": 1.1118,
+      "step": 2200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.025510636900972e-06,
+      "loss": 1.124,
+      "step": 2201
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0245998550507106e-06,
+      "loss": 1.126,
+      "step": 2202
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0236887509208825e-06,
+      "loss": 1.1191,
+      "step": 2203
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.022777324704083e-06,
+      "loss": 1.1582,
+      "step": 2204
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.021865576592975e-06,
+      "loss": 1.0947,
+      "step": 2205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.02095350678029e-06,
+      "loss": 1.1079,
+      "step": 2206
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.020041115458827e-06,
+      "loss": 1.1123,
+      "step": 2207
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.019128402821453e-06,
+      "loss": 1.1064,
+      "step": 2208
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.018215369061103e-06,
+      "loss": 1.1592,
+      "step": 2209
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.017302014370781e-06,
+      "loss": 1.1113,
+      "step": 2210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.016388338943554e-06,
+      "loss": 1.0898,
+      "step": 2211
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.015474342972565e-06,
+      "loss": 1.1182,
+      "step": 2212
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.014560026651018e-06,
+      "loss": 1.1387,
+      "step": 2213
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.013645390172186e-06,
+      "loss": 1.1152,
+      "step": 2214
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.012730433729413e-06,
+      "loss": 1.1445,
+      "step": 2215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0118151575161074e-06,
+      "loss": 1.1514,
+      "step": 2216
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.010899561725745e-06,
+      "loss": 1.1533,
+      "step": 2217
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.009983646551872e-06,
+      "loss": 1.1279,
+      "step": 2218
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.009067412188099e-06,
+      "loss": 1.106,
+      "step": 2219
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.008150858828106e-06,
+      "loss": 1.1787,
+      "step": 2220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.00723398666564e-06,
+      "loss": 1.1367,
+      "step": 2221
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.0063167958945135e-06,
+      "loss": 1.1328,
+      "step": 2222
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.005399286708611e-06,
+      "loss": 1.1826,
+      "step": 2223
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.0044814593018784e-06,
+      "loss": 1.1826,
+      "step": 2224
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.003563313868335e-06,
+      "loss": 1.0908,
+      "step": 2225
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.002644850602062e-06,
+      "loss": 1.0801,
+      "step": 2226
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.00172606969721e-06,
+      "loss": 1.125,
+      "step": 2227
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.000806971347998e-06,
+      "loss": 1.124,
+      "step": 2228
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.99988755574871e-06,
+      "loss": 1.1074,
+      "step": 2229
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.998967823093698e-06,
+      "loss": 1.1572,
+      "step": 2230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.99804777357738e-06,
+      "loss": 1.0781,
+      "step": 2231
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.997127407394244e-06,
+      "loss": 1.1367,
+      "step": 2232
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.996206724738842e-06,
+      "loss": 1.1191,
+      "step": 2233
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.995285725805792e-06,
+      "loss": 1.1357,
+      "step": 2234
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9943644107897825e-06,
+      "loss": 1.1494,
+      "step": 2235
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.993442779885566e-06,
+      "loss": 1.1558,
+      "step": 2236
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.992520833287963e-06,
+      "loss": 1.1416,
+      "step": 2237
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.991598571191861e-06,
+      "loss": 1.0884,
+      "step": 2238
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.990675993792213e-06,
+      "loss": 1.207,
+      "step": 2239
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9897531012840384e-06,
+      "loss": 1.1133,
+      "step": 2240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.988829893862426e-06,
+      "loss": 1.1074,
+      "step": 2241
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9879063717225275e-06,
+      "loss": 1.0791,
+      "step": 2242
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.986982535059563e-06,
+      "loss": 1.1113,
+      "step": 2243
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.98605838406882e-06,
+      "loss": 1.1475,
+      "step": 2244
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.985133918945652e-06,
+      "loss": 1.1514,
+      "step": 2245
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.984209139885476e-06,
+      "loss": 1.1475,
+      "step": 2246
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.983284047083778e-06,
+      "loss": 1.1133,
+      "step": 2247
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.982358640736112e-06,
+      "loss": 1.1582,
+      "step": 2248
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.981432921038095e-06,
+      "loss": 1.1504,
+      "step": 2249
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.98050688818541e-06,
+      "loss": 1.1123,
+      "step": 2250
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.979580542373811e-06,
+      "loss": 1.1289,
+      "step": 2251
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9786538837991114e-06,
+      "loss": 1.1377,
+      "step": 2252
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.977726912657197e-06,
+      "loss": 1.1514,
+      "step": 2253
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.976799629144015e-06,
+      "loss": 1.1416,
+      "step": 2254
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.975872033455582e-06,
+      "loss": 1.126,
+      "step": 2255
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.974944125787977e-06,
+      "loss": 1.1221,
+      "step": 2256
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.974015906337348e-06,
+      "loss": 1.1592,
+      "step": 2257
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.973087375299909e-06,
+      "loss": 1.125,
+      "step": 2258
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.972158532871936e-06,
+      "loss": 1.106,
+      "step": 2259
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.971229379249777e-06,
+      "loss": 1.1465,
+      "step": 2260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9702999146298395e-06,
+      "loss": 1.0615,
+      "step": 2261
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.969370139208601e-06,
+      "loss": 1.1426,
+      "step": 2262
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.968440053182604e-06,
+      "loss": 1.1631,
+      "step": 2263
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.967509656748455e-06,
+      "loss": 1.1099,
+      "step": 2264
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.966578950102827e-06,
+      "loss": 1.0908,
+      "step": 2265
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.965647933442459e-06,
+      "loss": 1.0664,
+      "step": 2266
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9647166069641545e-06,
+      "loss": 1.125,
+      "step": 2267
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.963784970864784e-06,
+      "loss": 1.125,
+      "step": 2268
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.962853025341282e-06,
+      "loss": 1.1055,
+      "step": 2269
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.96192077059065e-06,
+      "loss": 1.1357,
+      "step": 2270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.960988206809954e-06,
+      "loss": 1.125,
+      "step": 2271
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.960055334196323e-06,
+      "loss": 1.1006,
+      "step": 2272
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9591221529469555e-06,
+      "loss": 1.126,
+      "step": 2273
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.958188663259114e-06,
+      "loss": 1.1543,
+      "step": 2274
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.957254865330123e-06,
+      "loss": 1.1338,
+      "step": 2275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.956320759357376e-06,
+      "loss": 1.1377,
+      "step": 2276
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.955386345538331e-06,
+      "loss": 1.1221,
+      "step": 2277
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9544516240705086e-06,
+      "loss": 1.1406,
+      "step": 2278
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9535165951514974e-06,
+      "loss": 1.1416,
+      "step": 2279
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9525812589789495e-06,
+      "loss": 1.1055,
+      "step": 2280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.951645615750581e-06,
+      "loss": 1.1406,
+      "step": 2281
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.950709665664176e-06,
+      "loss": 1.1055,
+      "step": 2282
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.949773408917581e-06,
+      "loss": 1.0684,
+      "step": 2283
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9488368457087065e-06,
+      "loss": 1.1367,
+      "step": 2284
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.947899976235531e-06,
+      "loss": 1.1201,
+      "step": 2285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9469628006960944e-06,
+      "loss": 1.0986,
+      "step": 2286
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.946025319288504e-06,
+      "loss": 1.1641,
+      "step": 2287
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.94508753221093e-06,
+      "loss": 1.1299,
+      "step": 2288
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.944149439661608e-06,
+      "loss": 1.1299,
+      "step": 2289
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9432110418388366e-06,
+      "loss": 1.1504,
+      "step": 2290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.942272338940982e-06,
+      "loss": 1.1162,
+      "step": 2291
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.941333331166472e-06,
+      "loss": 1.1396,
+      "step": 2292
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.940394018713802e-06,
+      "loss": 1.1099,
+      "step": 2293
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.939454401781528e-06,
+      "loss": 1.124,
+      "step": 2294
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9385144805682715e-06,
+      "loss": 1.1406,
+      "step": 2295
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.937574255272721e-06,
+      "loss": 1.1318,
+      "step": 2296
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.936633726093626e-06,
+      "loss": 1.105,
+      "step": 2297
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.935692893229802e-06,
+      "loss": 1.0913,
+      "step": 2298
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.934751756880127e-06,
+      "loss": 1.1318,
+      "step": 2299
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.933810317243547e-06,
+      "loss": 1.1465,
+      "step": 2300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9328685745190675e-06,
+      "loss": 1.1157,
+      "step": 2301
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9319265289057595e-06,
+      "loss": 1.1543,
+      "step": 2302
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9309841806027604e-06,
+      "loss": 1.1172,
+      "step": 2303
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.930041529809269e-06,
+      "loss": 1.0806,
+      "step": 2304
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.929098576724547e-06,
+      "loss": 1.1211,
+      "step": 2305
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.928155321547925e-06,
+      "loss": 1.1631,
+      "step": 2306
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.927211764478792e-06,
+      "loss": 1.1021,
+      "step": 2307
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.926267905716603e-06,
+      "loss": 1.1152,
+      "step": 2308
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.925323745460879e-06,
+      "loss": 1.0967,
+      "step": 2309
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9243792839112e-06,
+      "loss": 1.1221,
+      "step": 2310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.923434521267214e-06,
+      "loss": 1.1279,
+      "step": 2311
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.92248945772863e-06,
+      "loss": 1.1982,
+      "step": 2312
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.921544093495222e-06,
+      "loss": 1.0879,
+      "step": 2313
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9205984287668266e-06,
+      "loss": 1.1211,
+      "step": 2314
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.919652463743345e-06,
+      "loss": 1.1484,
+      "step": 2315
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.918706198624741e-06,
+      "loss": 1.1084,
+      "step": 2316
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9177596336110415e-06,
+      "loss": 1.1011,
+      "step": 2317
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9168127689023375e-06,
+      "loss": 1.0918,
+      "step": 2318
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.915865604698785e-06,
+      "loss": 1.1372,
+      "step": 2319
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9149181412005995e-06,
+      "loss": 1.1621,
+      "step": 2320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.913970378608063e-06,
+      "loss": 1.1172,
+      "step": 2321
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.913022317121519e-06,
+      "loss": 1.1006,
+      "step": 2322
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.912073956941375e-06,
+      "loss": 1.1318,
+      "step": 2323
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9111252982681e-06,
+      "loss": 1.1123,
+      "step": 2324
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9101763413022306e-06,
+      "loss": 1.1377,
+      "step": 2325
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9092270862443605e-06,
+      "loss": 1.1279,
+      "step": 2326
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.90827753329515e-06,
+      "loss": 1.1138,
+      "step": 2327
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.907327682655322e-06,
+      "loss": 1.0928,
+      "step": 2328
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.906377534525662e-06,
+      "loss": 1.1152,
+      "step": 2329
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.905427089107019e-06,
+      "loss": 1.0957,
+      "step": 2330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.904476346600302e-06,
+      "loss": 1.1035,
+      "step": 2331
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.903525307206486e-06,
+      "loss": 1.062,
+      "step": 2332
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.902573971126609e-06,
+      "loss": 1.1299,
+      "step": 2333
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.901622338561769e-06,
+      "loss": 1.1226,
+      "step": 2334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9006704097131275e-06,
+      "loss": 1.1387,
+      "step": 2335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.899718184781911e-06,
+      "loss": 1.1074,
+      "step": 2336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.898765663969407e-06,
+      "loss": 1.1396,
+      "step": 2337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.897812847476963e-06,
+      "loss": 1.1255,
+      "step": 2338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.896859735505994e-06,
+      "loss": 1.165,
+      "step": 2339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.8959063282579735e-06,
+      "loss": 1.1011,
+      "step": 2340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.894952625934439e-06,
+      "loss": 1.1216,
+      "step": 2341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.893998628736988e-06,
+      "loss": 1.1484,
+      "step": 2342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.893044336867287e-06,
+      "loss": 1.1504,
+      "step": 2343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.892089750527057e-06,
+      "loss": 1.1401,
+      "step": 2344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.891134869918085e-06,
+      "loss": 1.1104,
+      "step": 2345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.890179695242221e-06,
+      "loss": 1.1416,
+      "step": 2346
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.889224226701373e-06,
+      "loss": 1.1348,
+      "step": 2347
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.888268464497518e-06,
+      "loss": 1.1064,
+      "step": 2348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.887312408832688e-06,
+      "loss": 1.126,
+      "step": 2349
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.88635605990898e-06,
+      "loss": 1.1191,
+      "step": 2350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.885399417928556e-06,
+      "loss": 1.1709,
+      "step": 2351
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.884442483093634e-06,
+      "loss": 1.1035,
+      "step": 2352
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.883485255606498e-06,
+      "loss": 1.1006,
+      "step": 2353
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.882527735669493e-06,
+      "loss": 1.1455,
+      "step": 2354
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.881569923485026e-06,
+      "loss": 1.1689,
+      "step": 2355
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.880611819255565e-06,
+      "loss": 1.0645,
+      "step": 2356
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.879653423183639e-06,
+      "loss": 1.0967,
+      "step": 2357
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.878694735471842e-06,
+      "loss": 1.167,
+      "step": 2358
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.877735756322825e-06,
+      "loss": 1.1104,
+      "step": 2359
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.876776485939305e-06,
+      "loss": 1.0918,
+      "step": 2360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.875816924524059e-06,
+      "loss": 1.1748,
+      "step": 2361
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.874857072279924e-06,
+      "loss": 1.1123,
+      "step": 2362
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.873896929409799e-06,
+      "loss": 1.1216,
+      "step": 2363
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8729364961166475e-06,
+      "loss": 1.1348,
+      "step": 2364
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.871975772603488e-06,
+      "loss": 1.1538,
+      "step": 2365
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.871014759073408e-06,
+      "loss": 1.0908,
+      "step": 2366
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.870053455729552e-06,
+      "loss": 1.1152,
+      "step": 2367
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.869091862775125e-06,
+      "loss": 1.1475,
+      "step": 2368
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8681299804133955e-06,
+      "loss": 1.166,
+      "step": 2369
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.867167808847693e-06,
+      "loss": 1.1709,
+      "step": 2370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.866205348281405e-06,
+      "loss": 1.1377,
+      "step": 2371
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.865242598917985e-06,
+      "loss": 1.1318,
+      "step": 2372
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8642795609609444e-06,
+      "loss": 1.0894,
+      "step": 2373
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.863316234613855e-06,
+      "loss": 1.1543,
+      "step": 2374
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.862352620080353e-06,
+      "loss": 1.1123,
+      "step": 2375
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.86138871756413e-06,
+      "loss": 1.1592,
+      "step": 2376
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8604245272689466e-06,
+      "loss": 1.082,
+      "step": 2377
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.859460049398615e-06,
+      "loss": 1.0918,
+      "step": 2378
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.858495284157015e-06,
+      "loss": 1.1172,
+      "step": 2379
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.857530231748084e-06,
+      "loss": 1.1104,
+      "step": 2380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8565648923758215e-06,
+      "loss": 1.1113,
+      "step": 2381
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8555992662442865e-06,
+      "loss": 1.1006,
+      "step": 2382
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.854633353557599e-06,
+      "loss": 1.1807,
+      "step": 2383
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.85366715451994e-06,
+      "loss": 1.1206,
+      "step": 2384
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8527006693355514e-06,
+      "loss": 1.1074,
+      "step": 2385
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.851733898208734e-06,
+      "loss": 1.083,
+      "step": 2386
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.85076684134385e-06,
+      "loss": 1.1475,
+      "step": 2387
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.849799498945324e-06,
+      "loss": 1.1768,
+      "step": 2388
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.848831871217635e-06,
+      "loss": 1.1133,
+      "step": 2389
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8478639583653295e-06,
+      "loss": 1.1157,
+      "step": 2390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8468957605930106e-06,
+      "loss": 1.1025,
+      "step": 2391
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.845927278105341e-06,
+      "loss": 1.1123,
+      "step": 2392
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.844958511107045e-06,
+      "loss": 1.1494,
+      "step": 2393
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.843989459802908e-06,
+      "loss": 1.1406,
+      "step": 2394
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.843020124397773e-06,
+      "loss": 1.1807,
+      "step": 2395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.842050505096544e-06,
+      "loss": 1.1318,
+      "step": 2396
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.841080602104186e-06,
+      "loss": 1.1094,
+      "step": 2397
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.840110415625723e-06,
+      "loss": 1.0791,
+      "step": 2398
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.839139945866238e-06,
+      "loss": 1.0781,
+      "step": 2399
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.838169193030877e-06,
+      "loss": 1.1592,
+      "step": 2400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8371981573248425e-06,
+      "loss": 1.1035,
+      "step": 2401
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.836226838953399e-06,
+      "loss": 1.1372,
+      "step": 2402
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.83525523812187e-06,
+      "loss": 1.1123,
+      "step": 2403
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.834283355035637e-06,
+      "loss": 1.1396,
+      "step": 2404
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.833311189900145e-06,
+      "loss": 1.1562,
+      "step": 2405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.832338742920896e-06,
+      "loss": 1.105,
+      "step": 2406
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8313660143034504e-06,
+      "loss": 1.1162,
+      "step": 2407
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.830393004253431e-06,
+      "loss": 1.083,
+      "step": 2408
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8294197129765185e-06,
+      "loss": 1.1025,
+      "step": 2409
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.828446140678454e-06,
+      "loss": 1.1084,
+      "step": 2410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.827472287565036e-06,
+      "loss": 1.1562,
+      "step": 2411
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.826498153842125e-06,
+      "loss": 1.1357,
+      "step": 2412
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.825523739715639e-06,
+      "loss": 1.0703,
+      "step": 2413
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8245490453915566e-06,
+      "loss": 1.1221,
+      "step": 2414
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8235740710759126e-06,
+      "loss": 1.1367,
+      "step": 2415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.822598816974807e-06,
+      "loss": 1.1152,
+      "step": 2416
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.821623283294392e-06,
+      "loss": 1.1016,
+      "step": 2417
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8206474702408825e-06,
+      "loss": 1.1035,
+      "step": 2418
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.819671378020553e-06,
+      "loss": 1.1299,
+      "step": 2419
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.818695006839736e-06,
+      "loss": 1.1055,
+      "step": 2420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.817718356904823e-06,
+      "loss": 1.1562,
+      "step": 2421
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.816741428422263e-06,
+      "loss": 1.126,
+      "step": 2422
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8157642215985665e-06,
+      "loss": 1.1484,
+      "step": 2423
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.814786736640303e-06,
+      "loss": 1.1221,
+      "step": 2424
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.813808973754098e-06,
+      "loss": 1.1299,
+      "step": 2425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8128309331466363e-06,
+      "loss": 1.1133,
+      "step": 2426
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.811852615024664e-06,
+      "loss": 1.1328,
+      "step": 2427
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8108740195949836e-06,
+      "loss": 1.1299,
+      "step": 2428
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8098951470644563e-06,
+      "loss": 1.1294,
+      "step": 2429
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8089159976400024e-06,
+      "loss": 1.1104,
+      "step": 2430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8079365715286017e-06,
+      "loss": 1.1353,
+      "step": 2431
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8069568689372906e-06,
+      "loss": 1.1265,
+      "step": 2432
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.805976890073165e-06,
+      "loss": 1.0889,
+      "step": 2433
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8049966351433787e-06,
+      "loss": 1.125,
+      "step": 2434
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.804016104355146e-06,
+      "loss": 1.1357,
+      "step": 2435
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8030352979157352e-06,
+      "loss": 1.125,
+      "step": 2436
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.802054216032477e-06,
+      "loss": 1.1499,
+      "step": 2437
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8010728589127575e-06,
+      "loss": 1.1279,
+      "step": 2438
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8000912267640233e-06,
+      "loss": 1.1016,
+      "step": 2439
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7991093197937777e-06,
+      "loss": 1.1299,
+      "step": 2440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7981271382095823e-06,
+      "loss": 1.1079,
+      "step": 2441
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7971446822190572e-06,
+      "loss": 1.124,
+      "step": 2442
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7961619520298797e-06,
+      "loss": 1.1641,
+      "step": 2443
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7951789478497846e-06,
+      "loss": 1.1328,
+      "step": 2444
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7941956698865672e-06,
+      "loss": 1.1006,
+      "step": 2445
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.793212118348079e-06,
+      "loss": 1.084,
+      "step": 2446
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.792228293442228e-06,
+      "loss": 1.1846,
+      "step": 2447
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.791244195376983e-06,
+      "loss": 1.1045,
+      "step": 2448
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.790259824360367e-06,
+      "loss": 1.1738,
+      "step": 2449
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7892751806004647e-06,
+      "loss": 1.1182,
+      "step": 2450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.788290264305413e-06,
+      "loss": 1.1387,
+      "step": 2451
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.787305075683414e-06,
+      "loss": 1.1445,
+      "step": 2452
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.78631961494272e-06,
+      "loss": 1.085,
+      "step": 2453
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.785333882291645e-06,
+      "loss": 1.1328,
+      "step": 2454
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.78434787793856e-06,
+      "loss": 1.1074,
+      "step": 2455
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7833616020918915e-06,
+      "loss": 1.0938,
+      "step": 2456
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.782375054960126e-06,
+      "loss": 1.1152,
+      "step": 2457
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7813882367518053e-06,
+      "loss": 1.1455,
+      "step": 2458
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.78040114767553e-06,
+      "loss": 1.1235,
+      "step": 2459
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7794137879399558e-06,
+      "loss": 1.1396,
+      "step": 2460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7784261577537984e-06,
+      "loss": 1.0996,
+      "step": 2461
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.777438257325829e-06,
+      "loss": 1.1436,
+      "step": 2462
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.776450086864876e-06,
+      "loss": 1.1157,
+      "step": 2463
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7754616465798242e-06,
+      "loss": 1.0859,
+      "step": 2464
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.774472936679619e-06,
+      "loss": 1.126,
+      "step": 2465
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.773483957373258e-06,
+      "loss": 1.1113,
+      "step": 2466
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7724947088697984e-06,
+      "loss": 1.1064,
+      "step": 2467
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.771505191378354e-06,
+      "loss": 1.104,
+      "step": 2468
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.770515405108095e-06,
+      "loss": 1.123,
+      "step": 2469
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.769525350268249e-06,
+      "loss": 1.1777,
+      "step": 2470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7685350270681e-06,
+      "loss": 1.1338,
+      "step": 2471
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7675444357169887e-06,
+      "loss": 1.127,
+      "step": 2472
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.766553576424312e-06,
+      "loss": 1.1152,
+      "step": 2473
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.765562449399525e-06,
+      "loss": 1.1689,
+      "step": 2474
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7645710548521373e-06,
+      "loss": 1.1299,
+      "step": 2475
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7635793929917185e-06,
+      "loss": 1.1182,
+      "step": 2476
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.762587464027889e-06,
+      "loss": 1.123,
+      "step": 2477
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7615952681703314e-06,
+      "loss": 1.1108,
+      "step": 2478
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7606028056287817e-06,
+      "loss": 1.1445,
+      "step": 2479
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7596100766130334e-06,
+      "loss": 1.1055,
+      "step": 2480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.758617081332934e-06,
+      "loss": 1.0811,
+      "step": 2481
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.757623819998391e-06,
+      "loss": 1.123,
+      "step": 2482
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7566302928193667e-06,
+      "loss": 1.1318,
+      "step": 2483
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.755636500005877e-06,
+      "loss": 1.127,
+      "step": 2484
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7546424417679973e-06,
+      "loss": 1.1094,
+      "step": 2485
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.753648118315858e-06,
+      "loss": 1.1602,
+      "step": 2486
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7526535298596457e-06,
+      "loss": 1.0791,
+      "step": 2487
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.751658676609602e-06,
+      "loss": 1.1167,
+      "step": 2488
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.750663558776026e-06,
+      "loss": 1.1719,
+      "step": 2489
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.749668176569272e-06,
+      "loss": 1.1338,
+      "step": 2490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7486725301997497e-06,
+      "loss": 1.0898,
+      "step": 2491
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7476766198779256e-06,
+      "loss": 1.0942,
+      "step": 2492
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7466804458143203e-06,
+      "loss": 1.1357,
+      "step": 2493
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.745684008219514e-06,
+      "loss": 1.1436,
+      "step": 2494
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7446873073041375e-06,
+      "loss": 1.1357,
+      "step": 2495
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7436903432788806e-06,
+      "loss": 1.123,
+      "step": 2496
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7426931163544877e-06,
+      "loss": 1.0933,
+      "step": 2497
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7416956267417604e-06,
+      "loss": 1.1045,
+      "step": 2498
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.740697874651551e-06,
+      "loss": 1.1426,
+      "step": 2499
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.739699860294774e-06,
+      "loss": 1.1504,
+      "step": 2500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7387015838823947e-06,
+      "loss": 1.1221,
+      "step": 2501
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7377030456254353e-06,
+      "loss": 1.1377,
+      "step": 2502
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7367042457349724e-06,
+      "loss": 1.1562,
+      "step": 2503
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7357051844221394e-06,
+      "loss": 1.1396,
+      "step": 2504
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7347058618981243e-06,
+      "loss": 1.105,
+      "step": 2505
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7337062783741696e-06,
+      "loss": 1.1162,
+      "step": 2506
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.732706434061574e-06,
+      "loss": 1.1191,
+      "step": 2507
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7317063291716904e-06,
+      "loss": 1.1387,
+      "step": 2508
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7307059639159288e-06,
+      "loss": 1.1289,
+      "step": 2509
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.729705338505751e-06,
+      "loss": 1.0879,
+      "step": 2510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7287044531526755e-06,
+      "loss": 1.0869,
+      "step": 2511
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7277033080682773e-06,
+      "loss": 1.1104,
+      "step": 2512
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7267019034641843e-06,
+      "loss": 1.0889,
+      "step": 2513
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.725700239552079e-06,
+      "loss": 1.1289,
+      "step": 2514
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7246983165436994e-06,
+      "loss": 1.124,
+      "step": 2515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7236961346508396e-06,
+      "loss": 1.1152,
+      "step": 2516
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7226936940853463e-06,
+      "loss": 1.126,
+      "step": 2517
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7216909950591207e-06,
+      "loss": 1.123,
+      "step": 2518
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7206880377841216e-06,
+      "loss": 1.167,
+      "step": 2519
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7196848224723597e-06,
+      "loss": 1.0957,
+      "step": 2520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7186813493359007e-06,
+      "loss": 1.0894,
+      "step": 2521
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.717677618586866e-06,
+      "loss": 1.1152,
+      "step": 2522
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7166736304374283e-06,
+      "loss": 1.1162,
+      "step": 2523
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7156693850998195e-06,
+      "loss": 1.0908,
+      "step": 2524
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7146648827863224e-06,
+      "loss": 1.1484,
+      "step": 2525
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7136601237092745e-06,
+      "loss": 1.1182,
+      "step": 2526
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.712655108081069e-06,
+      "loss": 1.0928,
+      "step": 2527
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.711649836114153e-06,
+      "loss": 1.1279,
+      "step": 2528
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7106443080210254e-06,
+      "loss": 1.1211,
+      "step": 2529
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.709638524014241e-06,
+      "loss": 1.1426,
+      "step": 2530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.708632484306412e-06,
+      "loss": 1.1582,
+      "step": 2531
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7076261891101985e-06,
+      "loss": 1.1211,
+      "step": 2532
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7066196386383178e-06,
+      "loss": 1.1143,
+      "step": 2533
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7056128331035413e-06,
+      "loss": 1.0723,
+      "step": 2534
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7046057727186946e-06,
+      "loss": 1.0947,
+      "step": 2535
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.703598457696656e-06,
+      "loss": 1.1377,
+      "step": 2536
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.702590888250358e-06,
+      "loss": 1.1216,
+      "step": 2537
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.701583064592787e-06,
+      "loss": 1.0923,
+      "step": 2538
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.700574986936984e-06,
+      "loss": 1.1143,
+      "step": 2539
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6995666554960415e-06,
+      "loss": 1.1133,
+      "step": 2540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.698558070483107e-06,
+      "loss": 1.1191,
+      "step": 2541
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6975492321113828e-06,
+      "loss": 1.1309,
+      "step": 2542
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6965401405941225e-06,
+      "loss": 1.1279,
+      "step": 2543
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.695530796144635e-06,
+      "loss": 1.0986,
+      "step": 2544
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6945211989762813e-06,
+      "loss": 1.1221,
+      "step": 2545
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6935113493024766e-06,
+      "loss": 1.1313,
+      "step": 2546
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.69250124733669e-06,
+      "loss": 1.1221,
+      "step": 2547
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.691490893292442e-06,
+      "loss": 1.1055,
+      "step": 2548
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.690480287383308e-06,
+      "loss": 1.1211,
+      "step": 2549
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.689469429822918e-06,
+      "loss": 1.165,
+      "step": 2550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6884583208249507e-06,
+      "loss": 1.1299,
+      "step": 2551
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6874469606031427e-06,
+      "loss": 1.1162,
+      "step": 2552
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6864353493712806e-06,
+      "loss": 1.1128,
+      "step": 2553
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.685423487343206e-06,
+      "loss": 1.1426,
+      "step": 2554
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.684411374732812e-06,
+      "loss": 1.1304,
+      "step": 2555
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6833990117540458e-06,
+      "loss": 1.1113,
+      "step": 2556
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.682386398620907e-06,
+      "loss": 1.1348,
+      "step": 2557
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.681373535547449e-06,
+      "loss": 1.1719,
+      "step": 2558
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6803604227477752e-06,
+      "loss": 1.1494,
+      "step": 2559
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6793470604360455e-06,
+      "loss": 1.1318,
+      "step": 2560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6783334488264706e-06,
+      "loss": 1.0996,
+      "step": 2561
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6773195881333135e-06,
+      "loss": 1.1094,
+      "step": 2562
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6763054785708905e-06,
+      "loss": 1.1445,
+      "step": 2563
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6752911203535716e-06,
+      "loss": 1.1592,
+      "step": 2564
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6742765136957774e-06,
+      "loss": 1.1367,
+      "step": 2565
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.673261658811982e-06,
+      "loss": 1.127,
+      "step": 2566
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.672246555916711e-06,
+      "loss": 1.1089,
+      "step": 2567
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6712312052245447e-06,
+      "loss": 1.1338,
+      "step": 2568
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6702156069501144e-06,
+      "loss": 1.1152,
+      "step": 2569
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6691997613081026e-06,
+      "loss": 1.1572,
+      "step": 2570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6681836685132456e-06,
+      "loss": 1.0908,
+      "step": 2571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.667167328780331e-06,
+      "loss": 1.1299,
+      "step": 2572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6661507423242006e-06,
+      "loss": 1.1465,
+      "step": 2573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.665133909359745e-06,
+      "loss": 1.1211,
+      "step": 2574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6641168301019104e-06,
+      "loss": 1.1401,
+      "step": 2575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6630995047656932e-06,
+      "loss": 1.1079,
+      "step": 2576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.662081933566141e-06,
+      "loss": 1.1426,
+      "step": 2577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.661064116718355e-06,
+      "loss": 1.1167,
+      "step": 2578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6600460544374875e-06,
+      "loss": 1.125,
+      "step": 2579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6590277469387447e-06,
+      "loss": 1.1221,
+      "step": 2580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6580091944373808e-06,
+      "loss": 1.1187,
+      "step": 2581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6569903971487048e-06,
+      "loss": 1.1182,
+      "step": 2582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.655971355288076e-06,
+      "loss": 1.1172,
+      "step": 2583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6549520690709062e-06,
+      "loss": 1.1133,
+      "step": 2584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6539325387126583e-06,
+      "loss": 1.1182,
+      "step": 2585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6529127644288476e-06,
+      "loss": 1.1084,
+      "step": 2586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.651892746435041e-06,
+      "loss": 1.0908,
+      "step": 2587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.650872484946855e-06,
+      "loss": 1.1201,
+      "step": 2588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.649851980179959e-06,
+      "loss": 1.1069,
+      "step": 2589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6488312323500752e-06,
+      "loss": 1.1162,
+      "step": 2590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6478102416729743e-06,
+      "loss": 1.1211,
+      "step": 2591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6467890083644803e-06,
+      "loss": 1.1133,
+      "step": 2592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6457675326404673e-06,
+      "loss": 1.1006,
+      "step": 2593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.644745814716863e-06,
+      "loss": 1.1455,
+      "step": 2594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.643723854809643e-06,
+      "loss": 1.1182,
+      "step": 2595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6427016531348365e-06,
+      "loss": 1.1006,
+      "step": 2596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6416792099085217e-06,
+      "loss": 1.1406,
+      "step": 2597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.640656525346831e-06,
+      "loss": 1.1143,
+      "step": 2598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6396335996659448e-06,
+      "loss": 1.0723,
+      "step": 2599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.638610433082096e-06,
+      "loss": 1.1094,
+      "step": 2600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6375870258115676e-06,
+      "loss": 1.1123,
+      "step": 2601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.636563378070694e-06,
+      "loss": 1.0962,
+      "step": 2602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6355394900758608e-06,
+      "loss": 1.1069,
+      "step": 2603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.634515362043503e-06,
+      "loss": 1.127,
+      "step": 2604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6334909941901087e-06,
+      "loss": 1.1582,
+      "step": 2605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.632466386732214e-06,
+      "loss": 1.1318,
+      "step": 2606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6314415398864074e-06,
+      "loss": 1.1338,
+      "step": 2607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6304164538693276e-06,
+      "loss": 1.1099,
+      "step": 2608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.629391128897663e-06,
+      "loss": 1.1221,
+      "step": 2609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6283655651881546e-06,
+      "loss": 1.1221,
+      "step": 2610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.627339762957591e-06,
+      "loss": 1.1377,
+      "step": 2611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6263137224228138e-06,
+      "loss": 1.1299,
+      "step": 2612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.625287443800715e-06,
+      "loss": 1.166,
+      "step": 2613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6242609273082336e-06,
+      "loss": 1.0977,
+      "step": 2614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6232341731623623e-06,
+      "loss": 1.1299,
+      "step": 2615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6222071815801425e-06,
+      "loss": 1.1196,
+      "step": 2616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6211799527786667e-06,
+      "loss": 1.1357,
+      "step": 2617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.620152486975077e-06,
+      "loss": 1.1328,
+      "step": 2618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.619124784386565e-06,
+      "loss": 1.1572,
+      "step": 2619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6180968452303735e-06,
+      "loss": 1.1201,
+      "step": 2620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6170686697237945e-06,
+      "loss": 1.1162,
+      "step": 2621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.616040258084171e-06,
+      "loss": 1.1035,
+      "step": 2622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6150116105288942e-06,
+      "loss": 1.104,
+      "step": 2623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.613982727275407e-06,
+      "loss": 1.0977,
+      "step": 2624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6129536085411997e-06,
+      "loss": 1.2021,
+      "step": 2625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6119242545438153e-06,
+      "loss": 1.1025,
+      "step": 2626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6108946655008447e-06,
+      "loss": 1.1602,
+      "step": 2627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6098648416299297e-06,
+      "loss": 1.1055,
+      "step": 2628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6088347831487603e-06,
+      "loss": 1.1265,
+      "step": 2629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6078044902750765e-06,
+      "loss": 1.1797,
+      "step": 2630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6067739632266686e-06,
+      "loss": 1.0791,
+      "step": 2631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.605743202221376e-06,
+      "loss": 1.1426,
+      "step": 2632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.604712207477087e-06,
+      "loss": 1.0991,
+      "step": 2633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.60368097921174e-06,
+      "loss": 1.166,
+      "step": 2634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6026495176433236e-06,
+      "loss": 1.1348,
+      "step": 2635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.601617822989873e-06,
+      "loss": 1.1504,
+      "step": 2636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.600585895469475e-06,
+      "loss": 1.1377,
+      "step": 2637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.5995537353002654e-06,
+      "loss": 1.1162,
+      "step": 2638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.5985213427004277e-06,
+      "loss": 1.1318,
+      "step": 2639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.597488717888197e-06,
+      "loss": 1.1201,
+      "step": 2640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.5964558610818544e-06,
+      "loss": 1.1191,
+      "step": 2641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5954227724997337e-06,
+      "loss": 1.127,
+      "step": 2642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5943894523602144e-06,
+      "loss": 1.1172,
+      "step": 2643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5933559008817265e-06,
+      "loss": 1.0874,
+      "step": 2644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.592322118282748e-06,
+      "loss": 1.1592,
+      "step": 2645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5912881047818075e-06,
+      "loss": 1.1152,
+      "step": 2646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5902538605974815e-06,
+      "loss": 1.1348,
+      "step": 2647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5892193859483938e-06,
+      "loss": 1.1494,
+      "step": 2648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5881846810532183e-06,
+      "loss": 1.1113,
+      "step": 2649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.587149746130679e-06,
+      "loss": 1.1035,
+      "step": 2650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5861145813995464e-06,
+      "loss": 1.1328,
+      "step": 2651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5850791870786383e-06,
+      "loss": 1.1309,
+      "step": 2652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.584043563386825e-06,
+      "loss": 1.1328,
+      "step": 2653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.583007710543023e-06,
+      "loss": 1.1011,
+      "step": 2654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.581971628766197e-06,
+      "loss": 1.0972,
+      "step": 2655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.58093531827536e-06,
+      "loss": 1.0811,
+      "step": 2656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.579898779289575e-06,
+      "loss": 1.1211,
+      "step": 2657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5788620120279517e-06,
+      "loss": 1.1348,
+      "step": 2658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.577825016709648e-06,
+      "loss": 1.105,
+      "step": 2659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.576787793553871e-06,
+      "loss": 1.1289,
+      "step": 2660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5757503427798757e-06,
+      "loss": 1.1602,
+      "step": 2661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.574712664606964e-06,
+      "loss": 1.1348,
+      "step": 2662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5736747592544883e-06,
+      "loss": 1.1431,
+      "step": 2663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.572636626941846e-06,
+      "loss": 1.1504,
+      "step": 2664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5715982678884863e-06,
+      "loss": 1.082,
+      "step": 2665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.570559682313902e-06,
+      "loss": 1.1523,
+      "step": 2666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5695208704376373e-06,
+      "loss": 1.1289,
+      "step": 2667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.568481832479282e-06,
+      "loss": 1.1484,
+      "step": 2668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.567442568658475e-06,
+      "loss": 1.1416,
+      "step": 2669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.566403079194901e-06,
+      "loss": 1.1025,
+      "step": 2670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.565363364308296e-06,
+      "loss": 1.1758,
+      "step": 2671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.564323424218441e-06,
+      "loss": 1.0947,
+      "step": 2672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5632832591451643e-06,
+      "loss": 1.0591,
+      "step": 2673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5622428693083432e-06,
+      "loss": 1.1113,
+      "step": 2674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.561202254927901e-06,
+      "loss": 1.1455,
+      "step": 2675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5601614162238107e-06,
+      "loss": 1.0879,
+      "step": 2676
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.55912035341609e-06,
+      "loss": 1.167,
+      "step": 2677
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5580790667248065e-06,
+      "loss": 1.0938,
+      "step": 2678
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5570375563700736e-06,
+      "loss": 1.0947,
+      "step": 2679
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5559958225720524e-06,
+      "loss": 1.1318,
+      "step": 2680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5549538655509514e-06,
+      "loss": 1.1035,
+      "step": 2681
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5539116855270247e-06,
+      "loss": 1.125,
+      "step": 2682
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.552869282720577e-06,
+      "loss": 1.1279,
+      "step": 2683
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5518266573519554e-06,
+      "loss": 1.1318,
+      "step": 2684
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5507838096415593e-06,
+      "loss": 1.1055,
+      "step": 2685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5497407398098314e-06,
+      "loss": 1.1455,
+      "step": 2686
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5486974480772618e-06,
+      "loss": 1.1406,
+      "step": 2687
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.547653934664389e-06,
+      "loss": 1.1123,
+      "step": 2688
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.546610199791797e-06,
+      "loss": 1.123,
+      "step": 2689
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.545566243680117e-06,
+      "loss": 1.0713,
+      "step": 2690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.544522066550028e-06,
+      "loss": 1.1543,
+      "step": 2691
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.543477668622253e-06,
+      "loss": 1.1196,
+      "step": 2692
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.542433050117565e-06,
+      "loss": 1.1104,
+      "step": 2693
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5413882112567816e-06,
+      "loss": 1.1025,
+      "step": 2694
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.540343152260767e-06,
+      "loss": 1.1064,
+      "step": 2695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5392978733504336e-06,
+      "loss": 1.0723,
+      "step": 2696
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5382523747467374e-06,
+      "loss": 1.1299,
+      "step": 2697
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.537206656670683e-06,
+      "loss": 1.125,
+      "step": 2698
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.536160719343323e-06,
+      "loss": 1.082,
+      "step": 2699
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.53511456298575e-06,
+      "loss": 1.1006,
+      "step": 2700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5340681878191114e-06,
+      "loss": 1.1128,
+      "step": 2701
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5330215940645952e-06,
+      "loss": 1.1416,
+      "step": 2702
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5319747819434356e-06,
+      "loss": 1.0972,
+      "step": 2703
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5309277516769156e-06,
+      "loss": 1.1318,
+      "step": 2704
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5298805034863633e-06,
+      "loss": 1.1143,
+      "step": 2705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5288330375931516e-06,
+      "loss": 1.0801,
+      "step": 2706
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5277853542187017e-06,
+      "loss": 1.0938,
+      "step": 2707
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.526737453584479e-06,
+      "loss": 1.0986,
+      "step": 2708
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5256893359119952e-06,
+      "loss": 1.1055,
+      "step": 2709
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.524641001422809e-06,
+      "loss": 1.1592,
+      "step": 2710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.523592450338522e-06,
+      "loss": 1.1367,
+      "step": 2711
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5225436828807847e-06,
+      "loss": 1.1221,
+      "step": 2712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5214946992712932e-06,
+      "loss": 1.085,
+      "step": 2713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.520445499731787e-06,
+      "loss": 1.0977,
+      "step": 2714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5193960844840526e-06,
+      "loss": 1.1323,
+      "step": 2715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5183464537499224e-06,
+      "loss": 1.0845,
+      "step": 2716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.517296607751275e-06,
+      "loss": 1.1113,
+      "step": 2717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.516246546710031e-06,
+      "loss": 1.0996,
+      "step": 2718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.51519627084816e-06,
+      "loss": 1.064,
+      "step": 2719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5141457803876772e-06,
+      "loss": 1.1191,
+      "step": 2720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.513095075550641e-06,
+      "loss": 1.0996,
+      "step": 2721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.512044156559156e-06,
+      "loss": 1.125,
+      "step": 2722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5109930236353723e-06,
+      "loss": 1.1035,
+      "step": 2723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5099416770014848e-06,
+      "loss": 1.1182,
+      "step": 2724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5088901168797336e-06,
+      "loss": 1.0923,
+      "step": 2725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5078383434924048e-06,
+      "loss": 1.0586,
+      "step": 2726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5067863570618283e-06,
+      "loss": 1.1279,
+      "step": 2727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5057341578103804e-06,
+      "loss": 1.1201,
+      "step": 2728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.504681745960481e-06,
+      "loss": 1.0762,
+      "step": 2729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.503629121734595e-06,
+      "loss": 1.1157,
+      "step": 2730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.502576285355234e-06,
+      "loss": 1.0757,
+      "step": 2731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.501523237044953e-06,
+      "loss": 1.1367,
+      "step": 2732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5004699770263512e-06,
+      "loss": 1.1748,
+      "step": 2733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4994165055220742e-06,
+      "loss": 1.062,
+      "step": 2734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4983628227548112e-06,
+      "loss": 1.0996,
+      "step": 2735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.497308928947297e-06,
+      "loss": 1.1157,
+      "step": 2736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.496254824322308e-06,
+      "loss": 1.0884,
+      "step": 2737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4952005091026698e-06,
+      "loss": 1.1514,
+      "step": 2738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.49414598351125e-06,
+      "loss": 1.1514,
+      "step": 2739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.49309124777096e-06,
+      "loss": 1.1113,
+      "step": 2740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4920363021047576e-06,
+      "loss": 1.123,
+      "step": 2741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4909811467356423e-06,
+      "loss": 1.1377,
+      "step": 2742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4899257818866614e-06,
+      "loss": 1.1157,
+      "step": 2743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.488870207780903e-06,
+      "loss": 1.0869,
+      "step": 2744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4878144246415026e-06,
+      "loss": 1.147,
+      "step": 2745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.486758432691637e-06,
+      "loss": 1.1064,
+      "step": 2746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4857022321545293e-06,
+      "loss": 1.1133,
+      "step": 2747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4846458232534445e-06,
+      "loss": 1.1299,
+      "step": 2748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4835892062116943e-06,
+      "loss": 1.1006,
+      "step": 2749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4825323812526337e-06,
+      "loss": 1.1074,
+      "step": 2750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4814753485996586e-06,
+      "loss": 1.0869,
+      "step": 2751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.480418108476214e-06,
+      "loss": 1.0928,
+      "step": 2752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.479360661105784e-06,
+      "loss": 1.1152,
+      "step": 2753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4783030067119e-06,
+      "loss": 1.0977,
+      "step": 2754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4772451455181344e-06,
+      "loss": 1.127,
+      "step": 2755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.476187077748105e-06,
+      "loss": 1.1328,
+      "step": 2756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4751288036254732e-06,
+      "loss": 1.1299,
+      "step": 2757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4740703233739442e-06,
+      "loss": 1.1182,
+      "step": 2758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.473011637217265e-06,
+      "loss": 1.1006,
+      "step": 2759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.471952745379228e-06,
+      "loss": 1.0977,
+      "step": 2760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4708936480836686e-06,
+      "loss": 1.1025,
+      "step": 2761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4698343455544646e-06,
+      "loss": 1.1309,
+      "step": 2762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4687748380155394e-06,
+      "loss": 1.0977,
+      "step": 2763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.467715125690857e-06,
+      "loss": 1.1143,
+      "step": 2764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.466655208804428e-06,
+      "loss": 1.1016,
+      "step": 2765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.465595087580303e-06,
+      "loss": 1.1162,
+      "step": 2766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4645347622425764e-06,
+      "loss": 1.1221,
+      "step": 2767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.463474233015388e-06,
+      "loss": 1.0586,
+      "step": 2768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.462413500122919e-06,
+      "loss": 1.1011,
+      "step": 2769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.461352563789393e-06,
+      "loss": 1.1099,
+      "step": 2770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.460291424239078e-06,
+      "loss": 1.1533,
+      "step": 2771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4592300816962844e-06,
+      "loss": 1.1279,
+      "step": 2772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4581685363853644e-06,
+      "loss": 1.0654,
+      "step": 2773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4571067885307165e-06,
+      "loss": 1.1001,
+      "step": 2774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4560448383567773e-06,
+      "loss": 1.1143,
+      "step": 2775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4549826860880305e-06,
+      "loss": 1.124,
+      "step": 2776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.453920331948999e-06,
+      "loss": 1.1357,
+      "step": 2777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.452857776164251e-06,
+      "loss": 1.126,
+      "step": 2778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.451795018958394e-06,
+      "loss": 1.1719,
+      "step": 2779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4507320605560844e-06,
+      "loss": 1.1396,
+      "step": 2780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4496689011820146e-06,
+      "loss": 1.127,
+      "step": 2781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4486055410609216e-06,
+      "loss": 1.0659,
+      "step": 2782
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.447541980417586e-06,
+      "loss": 1.0776,
+      "step": 2783
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.44647821947683e-06,
+      "loss": 1.0791,
+      "step": 2784
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4454142584635175e-06,
+      "loss": 1.1592,
+      "step": 2785
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4443500976025557e-06,
+      "loss": 1.124,
+      "step": 2786
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4432857371188937e-06,
+      "loss": 1.0791,
+      "step": 2787
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.442221177237523e-06,
+      "loss": 1.1201,
+      "step": 2788
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4411564181834765e-06,
+      "loss": 1.0854,
+      "step": 2789
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4400914601818303e-06,
+      "loss": 1.1123,
+      "step": 2790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.439026303457702e-06,
+      "loss": 1.1123,
+      "step": 2791
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4379609482362495e-06,
+      "loss": 1.124,
+      "step": 2792
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4368953947426765e-06,
+      "loss": 1.1133,
+      "step": 2793
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4358296432022247e-06,
+      "loss": 1.0898,
+      "step": 2794
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4347636938401813e-06,
+      "loss": 1.1074,
+      "step": 2795
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4336975468818723e-06,
+      "loss": 1.1113,
+      "step": 2796
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.432631202552666e-06,
+      "loss": 1.0859,
+      "step": 2797
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.431564661077974e-06,
+      "loss": 1.1055,
+      "step": 2798
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.430497922683248e-06,
+      "loss": 1.1289,
+      "step": 2799
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4294309875939823e-06,
+      "loss": 1.0962,
+      "step": 2800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4283638560357124e-06,
+      "loss": 1.1294,
+      "step": 2801
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.427296528234016e-06,
+      "loss": 1.1406,
+      "step": 2802
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4262290044145096e-06,
+      "loss": 1.0859,
+      "step": 2803
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4251612848028542e-06,
+      "loss": 1.1338,
+      "step": 2804
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.424093369624752e-06,
+      "loss": 1.1436,
+      "step": 2805
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4230252591059455e-06,
+      "loss": 1.1426,
+      "step": 2806
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.421956953472217e-06,
+      "loss": 1.0981,
+      "step": 2807
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.420888452949393e-06,
+      "loss": 1.1182,
+      "step": 2808
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4198197577633396e-06,
+      "loss": 1.1084,
+      "step": 2809
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.418750868139965e-06,
+      "loss": 1.1484,
+      "step": 2810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4176817843052168e-06,
+      "loss": 1.1152,
+      "step": 2811
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.416612506485085e-06,
+      "loss": 1.123,
+      "step": 2812
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4155430349056017e-06,
+      "loss": 1.0986,
+      "step": 2813
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4144733697928368e-06,
+      "loss": 1.1162,
+      "step": 2814
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4134035113729032e-06,
+      "loss": 1.0986,
+      "step": 2815
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.412333459871954e-06,
+      "loss": 1.1216,
+      "step": 2816
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4112632155161856e-06,
+      "loss": 1.1084,
+      "step": 2817
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4101927785318306e-06,
+      "loss": 1.1436,
+      "step": 2818
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.409122149145166e-06,
+      "loss": 1.1226,
+      "step": 2819
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4080513275825077e-06,
+      "loss": 1.1113,
+      "step": 2820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4069803140702136e-06,
+      "loss": 1.1357,
+      "step": 2821
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4059091088346793e-06,
+      "loss": 1.0908,
+      "step": 2822
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4048377121023444e-06,
+      "loss": 1.1172,
+      "step": 2823
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.403766124099689e-06,
+      "loss": 1.1543,
+      "step": 2824
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.402694345053229e-06,
+      "loss": 1.0469,
+      "step": 2825
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4016223751895263e-06,
+      "loss": 1.1084,
+      "step": 2826
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4005502147351786e-06,
+      "loss": 1.1426,
+      "step": 2827
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.399477863916828e-06,
+      "loss": 1.1309,
+      "step": 2828
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3984053229611537e-06,
+      "loss": 1.1035,
+      "step": 2829
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3973325920948763e-06,
+      "loss": 1.125,
+      "step": 2830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3962596715447565e-06,
+      "loss": 1.1436,
+      "step": 2831
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.395186561537596e-06,
+      "loss": 1.1021,
+      "step": 2832
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3941132623002337e-06,
+      "loss": 1.0884,
+      "step": 2833
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.393039774059551e-06,
+      "loss": 1.083,
+      "step": 2834
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3919660970424707e-06,
+      "loss": 1.1621,
+      "step": 2835
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3908922314759502e-06,
+      "loss": 1.1357,
+      "step": 2836
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.389818177586992e-06,
+      "loss": 1.1113,
+      "step": 2837
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3887439356026356e-06,
+      "loss": 1.0879,
+      "step": 2838
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.387669505749962e-06,
+      "loss": 1.1094,
+      "step": 2839
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3865948882560902e-06,
+      "loss": 1.1553,
+      "step": 2840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.38552008334818e-06,
+      "loss": 1.1406,
+      "step": 2841
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.38444509125343e-06,
+      "loss": 1.1162,
+      "step": 2842
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.38336991219908e-06,
+      "loss": 1.0854,
+      "step": 2843
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3822945464124065e-06,
+      "loss": 1.124,
+      "step": 2844
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.381218994120728e-06,
+      "loss": 1.1641,
+      "step": 2845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3801432555514014e-06,
+      "loss": 1.1323,
+      "step": 2846
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3790673309318233e-06,
+      "loss": 1.0957,
+      "step": 2847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.377991220489429e-06,
+      "loss": 1.1021,
+      "step": 2848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.376914924451694e-06,
+      "loss": 1.0928,
+      "step": 2849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3758384430461325e-06,
+      "loss": 1.1348,
+      "step": 2850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3747617765002993e-06,
+      "loss": 1.0957,
+      "step": 2851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.373684925041784e-06,
+      "loss": 1.085,
+      "step": 2852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.37260788889822e-06,
+      "loss": 1.1377,
+      "step": 2853
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.371530668297278e-06,
+      "loss": 1.1562,
+      "step": 2854
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3704532634666683e-06,
+      "loss": 1.1406,
+      "step": 2855
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3693756746341384e-06,
+      "loss": 1.123,
+      "step": 2856
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.368297902027476e-06,
+      "loss": 1.1182,
+      "step": 2857
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3672199458745085e-06,
+      "loss": 1.1377,
+      "step": 2858
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3661418064030993e-06,
+      "loss": 1.1338,
+      "step": 2859
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3650634838411545e-06,
+      "loss": 1.126,
+      "step": 2860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3639849784166147e-06,
+      "loss": 1.1436,
+      "step": 2861
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3629062903574633e-06,
+      "loss": 1.1582,
+      "step": 2862
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3618274198917184e-06,
+      "loss": 1.0879,
+      "step": 2863
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3607483672474396e-06,
+      "loss": 1.1875,
+      "step": 2864
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.359669132652723e-06,
+      "loss": 1.1104,
+      "step": 2865
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3585897163357044e-06,
+      "loss": 1.1001,
+      "step": 2866
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.357510118524558e-06,
+      "loss": 1.1143,
+      "step": 2867
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3564303394474958e-06,
+      "loss": 1.1221,
+      "step": 2868
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3553503793327685e-06,
+      "loss": 1.1514,
+      "step": 2869
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3542702384086646e-06,
+      "loss": 1.1357,
+      "step": 2870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.35318991690351e-06,
+      "loss": 1.1416,
+      "step": 2871
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3521094150456725e-06,
+      "loss": 1.1338,
+      "step": 2872
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.351028733063554e-06,
+      "loss": 1.1558,
+      "step": 2873
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3499478711855947e-06,
+      "loss": 1.1318,
+      "step": 2874
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.348866829640276e-06,
+      "loss": 1.0947,
+      "step": 2875
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.347785608656115e-06,
+      "loss": 1.0918,
+      "step": 2876
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3467042084616665e-06,
+      "loss": 1.0864,
+      "step": 2877
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.345622629285523e-06,
+      "loss": 1.1016,
+      "step": 2878
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3445408713563167e-06,
+      "loss": 1.127,
+      "step": 2879
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.343458934902717e-06,
+      "loss": 1.1074,
+      "step": 2880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3423768201534283e-06,
+      "loss": 1.1011,
+      "step": 2881
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3412945273371964e-06,
+      "loss": 1.1055,
+      "step": 2882
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3402120566828033e-06,
+      "loss": 1.0757,
+      "step": 2883
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3391294084190686e-06,
+      "loss": 1.1279,
+      "step": 2884
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3380465827748482e-06,
+      "loss": 1.0771,
+      "step": 2885
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3369635799790372e-06,
+      "loss": 1.1006,
+      "step": 2886
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.335880400260568e-06,
+      "loss": 1.0986,
+      "step": 2887
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.33479704384841e-06,
+      "loss": 1.1455,
+      "step": 2888
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3337135109715684e-06,
+      "loss": 1.0942,
+      "step": 2889
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3326298018590897e-06,
+      "loss": 1.1836,
+      "step": 2890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.331545916740054e-06,
+      "loss": 1.0776,
+      "step": 2891
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3304618558435787e-06,
+      "loss": 1.1396,
+      "step": 2892
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3293776193988213e-06,
+      "loss": 1.1152,
+      "step": 2893
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.328293207634973e-06,
+      "loss": 1.1016,
+      "step": 2894
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3272086207812655e-06,
+      "loss": 1.1401,
+      "step": 2895
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3261238590669635e-06,
+      "loss": 1.124,
+      "step": 2896
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.325038922721372e-06,
+      "loss": 1.1318,
+      "step": 2897
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3239538119738318e-06,
+      "loss": 1.1152,
+      "step": 2898
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.322868527053721e-06,
+      "loss": 1.1172,
+      "step": 2899
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3217830681904516e-06,
+      "loss": 1.1191,
+      "step": 2900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.320697435613477e-06,
+      "loss": 1.1157,
+      "step": 2901
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.319611629552285e-06,
+      "loss": 1.0977,
+      "step": 2902
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.318525650236399e-06,
+      "loss": 1.123,
+      "step": 2903
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3174394978953805e-06,
+      "loss": 1.1494,
+      "step": 2904
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.316353172758828e-06,
+      "loss": 1.1631,
+      "step": 2905
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3152666750563757e-06,
+      "loss": 1.1523,
+      "step": 2906
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.314180005017693e-06,
+      "loss": 1.1758,
+      "step": 2907
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3130931628724885e-06,
+      "loss": 1.1211,
+      "step": 2908
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.312006148850505e-06,
+      "loss": 1.1553,
+      "step": 2909
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.310918963181524e-06,
+      "loss": 1.0894,
+      "step": 2910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3098316060953595e-06,
+      "loss": 1.1201,
+      "step": 2911
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.308744077821865e-06,
+      "loss": 1.0762,
+      "step": 2912
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3076563785909288e-06,
+      "loss": 1.1982,
+      "step": 2913
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.306568508632476e-06,
+      "loss": 1.1621,
+      "step": 2914
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.305480468176467e-06,
+      "loss": 1.1387,
+      "step": 2915
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.304392257452899e-06,
+      "loss": 1.0996,
+      "step": 2916
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.303303876691805e-06,
+      "loss": 1.0908,
+      "step": 2917
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3022153261232537e-06,
+      "loss": 1.1328,
+      "step": 2918
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3011266059773493e-06,
+      "loss": 1.1113,
+      "step": 2919
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3000377164842324e-06,
+      "loss": 1.0977,
+      "step": 2920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.298948657874081e-06,
+      "loss": 1.1523,
+      "step": 2921
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.297859430377105e-06,
+      "loss": 1.0908,
+      "step": 2922
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.296770034223553e-06,
+      "loss": 1.1025,
+      "step": 2923
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.295680469643709e-06,
+      "loss": 1.1016,
+      "step": 2924
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.294590736867892e-06,
+      "loss": 1.1016,
+      "step": 2925
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.293500836126455e-06,
+      "loss": 1.125,
+      "step": 2926
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.29241076764979e-06,
+      "loss": 1.0957,
+      "step": 2927
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2913205316683227e-06,
+      "loss": 1.1133,
+      "step": 2928
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2902301284125125e-06,
+      "loss": 1.1152,
+      "step": 2929
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.289139558112857e-06,
+      "loss": 1.1689,
+      "step": 2930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2880488209998874e-06,
+      "loss": 1.0957,
+      "step": 2931
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.28695791730417e-06,
+      "loss": 1.1211,
+      "step": 2932
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2858668472563082e-06,
+      "loss": 1.0962,
+      "step": 2933
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.284775611086938e-06,
+      "loss": 1.1533,
+      "step": 2934
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.283684209026733e-06,
+      "loss": 1.1016,
+      "step": 2935
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2825926413064e-06,
+      "loss": 1.1113,
+      "step": 2936
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2815009081566818e-06,
+      "loss": 1.0898,
+      "step": 2937
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.280409009808355e-06,
+      "loss": 1.0977,
+      "step": 2938
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2793169464922332e-06,
+      "loss": 1.1318,
+      "step": 2939
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.278224718439163e-06,
+      "loss": 1.1294,
+      "step": 2940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2771323258800268e-06,
+      "loss": 1.1523,
+      "step": 2941
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.276039769045741e-06,
+      "loss": 1.1328,
+      "step": 2942
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.274947048167258e-06,
+      "loss": 1.1167,
+      "step": 2943
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2738541634755624e-06,
+      "loss": 1.1206,
+      "step": 2944
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2727611152016763e-06,
+      "loss": 1.1758,
+      "step": 2945
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.271667903576656e-06,
+      "loss": 1.1543,
+      "step": 2946
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.27057452883159e-06,
+      "loss": 1.1143,
+      "step": 2947
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2694809911976034e-06,
+      "loss": 1.084,
+      "step": 2948
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2683872909058545e-06,
+      "loss": 1.125,
+      "step": 2949
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.267293428187538e-06,
+      "loss": 1.1299,
+      "step": 2950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.26619940327388e-06,
+      "loss": 1.1553,
+      "step": 2951
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2651052163961426e-06,
+      "loss": 1.0957,
+      "step": 2952
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.264010867785623e-06,
+      "loss": 1.125,
+      "step": 2953
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2629163576736513e-06,
+      "loss": 1.0928,
+      "step": 2954
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2618216862915906e-06,
+      "loss": 1.1064,
+      "step": 2955
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2607268538708404e-06,
+      "loss": 1.1475,
+      "step": 2956
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.259631860642834e-06,
+      "loss": 1.1113,
+      "step": 2957
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.258536706839038e-06,
+      "loss": 1.1582,
+      "step": 2958
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2574413926909508e-06,
+      "loss": 1.1553,
+      "step": 2959
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2563459184301093e-06,
+      "loss": 1.0967,
+      "step": 2960
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2552502842880806e-06,
+      "loss": 1.0933,
+      "step": 2961
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2541544904964668e-06,
+      "loss": 1.0811,
+      "step": 2962
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.253058537286904e-06,
+      "loss": 1.0718,
+      "step": 2963
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2519624248910613e-06,
+      "loss": 1.1367,
+      "step": 2964
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2508661535406434e-06,
+      "loss": 1.126,
+      "step": 2965
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.249769723467385e-06,
+      "loss": 1.1377,
+      "step": 2966
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.248673134903057e-06,
+      "loss": 1.0811,
+      "step": 2967
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2475763880794635e-06,
+      "loss": 1.0781,
+      "step": 2968
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.246479483228443e-06,
+      "loss": 1.1084,
+      "step": 2969
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.245382420581864e-06,
+      "loss": 1.0884,
+      "step": 2970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.244285200371632e-06,
+      "loss": 1.1387,
+      "step": 2971
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.243187822829684e-06,
+      "loss": 1.0698,
+      "step": 2972
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2420902881879903e-06,
+      "loss": 1.1025,
+      "step": 2973
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2409925966785545e-06,
+      "loss": 1.1211,
+      "step": 2974
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2398947485334144e-06,
+      "loss": 1.0669,
+      "step": 2975
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.23879674398464e-06,
+      "loss": 1.1289,
+      "step": 2976
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2376985832643344e-06,
+      "loss": 1.1143,
+      "step": 2977
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2366002666046333e-06,
+      "loss": 1.0938,
+      "step": 2978
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2355017942377055e-06,
+      "loss": 1.1445,
+      "step": 2979
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2344031663957543e-06,
+      "loss": 1.1445,
+      "step": 2980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2333043833110135e-06,
+      "loss": 1.1523,
+      "step": 2981
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.232205445215751e-06,
+      "loss": 1.1338,
+      "step": 2982
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2311063523422674e-06,
+      "loss": 1.1572,
+      "step": 2983
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.230007104922898e-06,
+      "loss": 1.1309,
+      "step": 2984
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.228907703190004e-06,
+      "loss": 1.0962,
+      "step": 2985
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.227808147375988e-06,
+      "loss": 1.1187,
+      "step": 2986
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2267084377132796e-06,
+      "loss": 1.1191,
+      "step": 2987
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2256085744343425e-06,
+      "loss": 1.1191,
+      "step": 2988
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.224508557771672e-06,
+      "loss": 1.168,
+      "step": 2989
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2234083879577977e-06,
+      "loss": 1.0986,
+      "step": 2990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2223080652252802e-06,
+      "loss": 1.1367,
+      "step": 2991
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.221207589806713e-06,
+      "loss": 1.1289,
+      "step": 2992
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2201069619347195e-06,
+      "loss": 1.1216,
+      "step": 2993
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2190061818419604e-06,
+      "loss": 1.1357,
+      "step": 2994
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.217905249761124e-06,
+      "loss": 1.1074,
+      "step": 2995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.216804165924933e-06,
+      "loss": 1.1074,
+      "step": 2996
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2157029305661397e-06,
+      "loss": 1.1221,
+      "step": 2997
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2146015439175314e-06,
+      "loss": 1.1182,
+      "step": 2998
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.213500006211927e-06,
+      "loss": 1.124,
+      "step": 2999
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.212398317682175e-06,
+      "loss": 1.1338,
+      "step": 3000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2112964785611583e-06,
+      "loss": 1.1338,
+      "step": 3001
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2101944890817904e-06,
+      "loss": 1.1128,
+      "step": 3002
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.209092349477017e-06,
+      "loss": 1.1113,
+      "step": 3003
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2079900599798148e-06,
+      "loss": 1.165,
+      "step": 3004
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2068876208231925e-06,
+      "loss": 1.0737,
+      "step": 3005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2057850322401916e-06,
+      "loss": 1.082,
+      "step": 3006
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.204682294463884e-06,
+      "loss": 1.1055,
+      "step": 3007
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.203579407727373e-06,
+      "loss": 1.1445,
+      "step": 3008
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.202476372263794e-06,
+      "loss": 1.0752,
+      "step": 3009
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2013731883063143e-06,
+      "loss": 1.1426,
+      "step": 3010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2002698560881296e-06,
+      "loss": 1.1475,
+      "step": 3011
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.199166375842472e-06,
+      "loss": 1.1328,
+      "step": 3012
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.198062747802601e-06,
+      "loss": 1.1133,
+      "step": 3013
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1969589722018084e-06,
+      "loss": 1.1377,
+      "step": 3014
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1958550492734176e-06,
+      "loss": 1.1631,
+      "step": 3015
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.194750979250782e-06,
+      "loss": 1.1367,
+      "step": 3016
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1936467623672873e-06,
+      "loss": 1.0918,
+      "step": 3017
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1925423988563505e-06,
+      "loss": 1.0996,
+      "step": 3018
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.191437888951418e-06,
+      "loss": 1.1689,
+      "step": 3019
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1903332328859694e-06,
+      "loss": 1.1494,
+      "step": 3020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.189228430893513e-06,
+      "loss": 1.085,
+      "step": 3021
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.188123483207588e-06,
+      "loss": 1.1313,
+      "step": 3022
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1870183900617662e-06,
+      "loss": 1.0938,
+      "step": 3023
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1859131516896502e-06,
+      "loss": 1.1885,
+      "step": 3024
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.18480776832487e-06,
+      "loss": 1.1123,
+      "step": 3025
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.183702240201091e-06,
+      "loss": 1.165,
+      "step": 3026
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1825965675520043e-06,
+      "loss": 1.166,
+      "step": 3027
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.181490750611336e-06,
+      "loss": 1.0811,
+      "step": 3028
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.180384789612839e-06,
+      "loss": 1.1494,
+      "step": 3029
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1792786847902995e-06,
+      "loss": 1.1191,
+      "step": 3030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.178172436377532e-06,
+      "loss": 1.1372,
+      "step": 3031
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1770660446083845e-06,
+      "loss": 1.1045,
+      "step": 3032
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1759595097167306e-06,
+      "loss": 1.1265,
+      "step": 3033
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1748528319364775e-06,
+      "loss": 1.0786,
+      "step": 3034
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1737460115015616e-06,
+      "loss": 1.1035,
+      "step": 3035
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.17263904864595e-06,
+      "loss": 1.1377,
+      "step": 3036
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.171531943603639e-06,
+      "loss": 1.1172,
+      "step": 3037
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1704246966086562e-06,
+      "loss": 1.1309,
+      "step": 3038
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1693173078950577e-06,
+      "loss": 1.0938,
+      "step": 3039
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1682097776969313e-06,
+      "loss": 1.0918,
+      "step": 3040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.167102106248392e-06,
+      "loss": 1.1562,
+      "step": 3041
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1659942937835887e-06,
+      "loss": 1.1074,
+      "step": 3042
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1648863405366966e-06,
+      "loss": 1.1133,
+      "step": 3043
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.163778246741921e-06,
+      "loss": 1.1182,
+      "step": 3044
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.162670012633499e-06,
+      "loss": 1.1289,
+      "step": 3045
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.161561638445696e-06,
+      "loss": 1.1533,
+      "step": 3046
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1604531244128073e-06,
+      "loss": 1.1279,
+      "step": 3047
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.159344470769157e-06,
+      "loss": 1.1328,
+      "step": 3048
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1582356777490997e-06,
+      "loss": 1.1826,
+      "step": 3049
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1571267455870186e-06,
+      "loss": 1.0708,
+      "step": 3050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.156017674517328e-06,
+      "loss": 1.0371,
+      "step": 3051
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1549084647744688e-06,
+      "loss": 1.126,
+      "step": 3052
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1537991165929133e-06,
+      "loss": 1.1108,
+      "step": 3053
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.152689630207164e-06,
+      "loss": 1.1016,
+      "step": 3054
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1515800058517486e-06,
+      "loss": 1.0615,
+      "step": 3055
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.150470243761229e-06,
+      "loss": 1.1074,
+      "step": 3056
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.149360344170192e-06,
+      "loss": 1.125,
+      "step": 3057
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1482503073132563e-06,
+      "loss": 1.1533,
+      "step": 3058
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1471401334250675e-06,
+      "loss": 1.0737,
+      "step": 3059
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.146029822740301e-06,
+      "loss": 1.1055,
+      "step": 3060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1449193754936625e-06,
+      "loss": 1.1392,
+      "step": 3061
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1438087919198855e-06,
+      "loss": 1.1104,
+      "step": 3062
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1426980722537315e-06,
+      "loss": 1.1484,
+      "step": 3063
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.141587216729991e-06,
+      "loss": 1.0811,
+      "step": 3064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.140476225583484e-06,
+      "loss": 1.1533,
+      "step": 3065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1393650990490594e-06,
+      "loss": 1.1367,
+      "step": 3066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1382538373615935e-06,
+      "loss": 1.1099,
+      "step": 3067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.137142440755992e-06,
+      "loss": 1.127,
+      "step": 3068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.13603090946719e-06,
+      "loss": 1.1025,
+      "step": 3069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1349192437301478e-06,
+      "loss": 1.1406,
+      "step": 3070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.133807443779858e-06,
+      "loss": 1.1182,
+      "step": 3071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.132695509851339e-06,
+      "loss": 1.0796,
+      "step": 3072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1315834421796405e-06,
+      "loss": 1.1172,
+      "step": 3073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1304712409998354e-06,
+      "loss": 1.0879,
+      "step": 3074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.12935890654703e-06,
+      "loss": 1.127,
+      "step": 3075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1282464390563548e-06,
+      "loss": 1.126,
+      "step": 3076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1271338387629728e-06,
+      "loss": 1.1094,
+      "step": 3077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1260211059020694e-06,
+      "loss": 1.1387,
+      "step": 3078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.124908240708864e-06,
+      "loss": 1.0723,
+      "step": 3079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1237952434185993e-06,
+      "loss": 1.1299,
+      "step": 3080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.122682114266548e-06,
+      "loss": 1.1211,
+      "step": 3081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1215688534880107e-06,
+      "loss": 1.1035,
+      "step": 3082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.120455461318315e-06,
+      "loss": 1.1133,
+      "step": 3083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1193419379928182e-06,
+      "loss": 1.0928,
+      "step": 3084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.118228283746902e-06,
+      "loss": 1.1113,
+      "step": 3085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1171144988159787e-06,
+      "loss": 1.1416,
+      "step": 3086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1160005834354867e-06,
+      "loss": 1.1094,
+      "step": 3087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.114886537840894e-06,
+      "loss": 1.1152,
+      "step": 3088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1137723622676925e-06,
+      "loss": 1.1504,
+      "step": 3089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.112658056951404e-06,
+      "loss": 1.1807,
+      "step": 3090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.111543622127579e-06,
+      "loss": 1.1343,
+      "step": 3091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.110429058031793e-06,
+      "loss": 1.1191,
+      "step": 3092
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.109314364899649e-06,
+      "loss": 1.0918,
+      "step": 3093
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1081995429667784e-06,
+      "loss": 1.0869,
+      "step": 3094
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.10708459246884e-06,
+      "loss": 1.1582,
+      "step": 3095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1059695136415168e-06,
+      "loss": 1.1396,
+      "step": 3096
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1048543067205232e-06,
+      "loss": 1.1426,
+      "step": 3097
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1037389719415987e-06,
+      "loss": 1.1279,
+      "step": 3098
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.10262350954051e-06,
+      "loss": 1.1255,
+      "step": 3099
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.101507919753049e-06,
+      "loss": 1.166,
+      "step": 3100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.100392202815037e-06,
+      "loss": 1.1182,
+      "step": 3101
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.099276358962322e-06,
+      "loss": 1.1113,
+      "step": 3102
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.098160388430777e-06,
+      "loss": 1.1602,
+      "step": 3103
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.097044291456304e-06,
+      "loss": 1.1235,
+      "step": 3104
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.095928068274829e-06,
+      "loss": 1.1699,
+      "step": 3105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0948117191223093e-06,
+      "loss": 1.1104,
+      "step": 3106
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.093695244234723e-06,
+      "loss": 1.1172,
+      "step": 3107
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.092578643848078e-06,
+      "loss": 1.0796,
+      "step": 3108
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0914619181984095e-06,
+      "loss": 1.1318,
+      "step": 3109
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0903450675217778e-06,
+      "loss": 1.1377,
+      "step": 3110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0892280920542692e-06,
+      "loss": 1.0962,
+      "step": 3111
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0881109920319973e-06,
+      "loss": 1.1562,
+      "step": 3112
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.086993767691102e-06,
+      "loss": 1.1162,
+      "step": 3113
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0858764192677493e-06,
+      "loss": 1.127,
+      "step": 3114
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.084758946998131e-06,
+      "loss": 1.082,
+      "step": 3115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.083641351118466e-06,
+      "loss": 1.1357,
+      "step": 3116
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0825236318649985e-06,
+      "loss": 1.1025,
+      "step": 3117
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.081405789473999e-06,
+      "loss": 1.1211,
+      "step": 3118
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0802878241817634e-06,
+      "loss": 1.1367,
+      "step": 3119
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0791697362246154e-06,
+      "loss": 1.1089,
+      "step": 3120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.078051525838903e-06,
+      "loss": 1.1309,
+      "step": 3121
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0769331932610015e-06,
+      "loss": 1.0903,
+      "step": 3122
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0758147387273097e-06,
+      "loss": 1.1455,
+      "step": 3123
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.074696162474254e-06,
+      "loss": 1.0908,
+      "step": 3124
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.073577464738287e-06,
+      "loss": 1.1338,
+      "step": 3125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0724586457558846e-06,
+      "loss": 1.1729,
+      "step": 3126
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0713397057635507e-06,
+      "loss": 1.1328,
+      "step": 3127
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.070220644997815e-06,
+      "loss": 1.1143,
+      "step": 3128
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0691014636952297e-06,
+      "loss": 1.1045,
+      "step": 3129
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0679821620923757e-06,
+      "loss": 1.1377,
+      "step": 3130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.066862740425857e-06,
+      "loss": 1.1221,
+      "step": 3131
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0657431989323054e-06,
+      "loss": 1.1475,
+      "step": 3132
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.064623537848376e-06,
+      "loss": 1.0723,
+      "step": 3133
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.063503757410749e-06,
+      "loss": 1.1133,
+      "step": 3134
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0623838578561317e-06,
+      "loss": 1.1162,
+      "step": 3135
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0612638394212558e-06,
+      "loss": 1.1016,
+      "step": 3136
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.060143702342877e-06,
+      "loss": 1.1348,
+      "step": 3137
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.059023446857777e-06,
+      "loss": 1.1562,
+      "step": 3138
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0579030732027636e-06,
+      "loss": 1.1406,
+      "step": 3139
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.056782581614668e-06,
+      "loss": 1.1143,
+      "step": 3140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0556619723303453e-06,
+      "loss": 1.1494,
+      "step": 3141
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.054541245586679e-06,
+      "loss": 1.1689,
+      "step": 3142
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.053420401620575e-06,
+      "loss": 1.1416,
+      "step": 3143
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.052299440668964e-06,
+      "loss": 1.1318,
+      "step": 3144
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.051178362968801e-06,
+      "loss": 1.1138,
+      "step": 3145
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0500571687570673e-06,
+      "loss": 1.1553,
+      "step": 3146
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.048935858270769e-06,
+      "loss": 1.0376,
+      "step": 3147
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.047814431746934e-06,
+      "loss": 1.0596,
+      "step": 3148
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.046692889422618e-06,
+      "loss": 1.1445,
+      "step": 3149
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.045571231534899e-06,
+      "loss": 1.0811,
+      "step": 3150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0444494583208807e-06,
+      "loss": 1.1064,
+      "step": 3151
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0433275700176894e-06,
+      "loss": 1.1113,
+      "step": 3152
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0422055668624777e-06,
+      "loss": 1.1377,
+      "step": 3153
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0410834490924223e-06,
+      "loss": 1.0732,
+      "step": 3154
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0399612169447234e-06,
+      "loss": 1.1025,
+      "step": 3155
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0388388706566045e-06,
+      "loss": 1.1143,
+      "step": 3156
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0377164104653143e-06,
+      "loss": 1.1318,
+      "step": 3157
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0365938366081277e-06,
+      "loss": 1.1182,
+      "step": 3158
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0354711493223395e-06,
+      "loss": 1.1387,
+      "step": 3159
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0343483488452702e-06,
+      "loss": 1.061,
+      "step": 3160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0332254354142653e-06,
+      "loss": 1.0996,
+      "step": 3161
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.032102409266694e-06,
+      "loss": 1.0947,
+      "step": 3162
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0309792706399464e-06,
+      "loss": 1.124,
+      "step": 3163
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.029856019771441e-06,
+      "loss": 1.106,
+      "step": 3164
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0287326568986175e-06,
+      "loss": 1.082,
+      "step": 3165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0276091822589376e-06,
+      "loss": 1.1426,
+      "step": 3166
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0264855960898905e-06,
+      "loss": 1.126,
+      "step": 3167
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.025361898628985e-06,
+      "loss": 1.1045,
+      "step": 3168
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0242380901137575e-06,
+      "loss": 1.1377,
+      "step": 3169
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0231141707817642e-06,
+      "loss": 1.1152,
+      "step": 3170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0219901408705867e-06,
+      "loss": 1.1104,
+      "step": 3171
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0208660006178297e-06,
+      "loss": 1.1279,
+      "step": 3172
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0197417502611213e-06,
+      "loss": 1.1279,
+      "step": 3173
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0186173900381126e-06,
+      "loss": 1.1011,
+      "step": 3174
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0174929201864765e-06,
+      "loss": 1.124,
+      "step": 3175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0163683409439136e-06,
+      "loss": 1.0923,
+      "step": 3176
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.015243652548142e-06,
+      "loss": 1.1128,
+      "step": 3177
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.014118855236906e-06,
+      "loss": 1.0864,
+      "step": 3178
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0129939492479733e-06,
+      "loss": 1.1387,
+      "step": 3179
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.011868934819132e-06,
+      "loss": 1.1201,
+      "step": 3180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0107438121881967e-06,
+      "loss": 1.1162,
+      "step": 3181
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0096185815930013e-06,
+      "loss": 1.1045,
+      "step": 3182
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0084932432714053e-06,
+      "loss": 1.1201,
+      "step": 3183
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0073677974612897e-06,
+      "loss": 1.0688,
+      "step": 3184
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0062422444005573e-06,
+      "loss": 1.1226,
+      "step": 3185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.005116584327136e-06,
+      "loss": 1.1357,
+      "step": 3186
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0039908174789734e-06,
+      "loss": 1.0737,
+      "step": 3187
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0028649440940427e-06,
+      "loss": 1.1123,
+      "step": 3188
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.001738964410338e-06,
+      "loss": 1.0928,
+      "step": 3189
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.000612878665875e-06,
+      "loss": 1.1123,
+      "step": 3190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9994866870986937e-06,
+      "loss": 1.1143,
+      "step": 3191
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9983603899468555e-06,
+      "loss": 1.1411,
+      "step": 3192
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9972339874484442e-06,
+      "loss": 1.1279,
+      "step": 3193
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9961074798415646e-06,
+      "loss": 1.1201,
+      "step": 3194
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9949808673643477e-06,
+      "loss": 1.1445,
+      "step": 3195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.993854150254941e-06,
+      "loss": 1.1338,
+      "step": 3196
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9927273287515196e-06,
+      "loss": 1.0977,
+      "step": 3197
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9916004030922768e-06,
+      "loss": 1.1094,
+      "step": 3198
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.99047337351543e-06,
+      "loss": 1.1123,
+      "step": 3199
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.989346240259216e-06,
+      "loss": 1.1001,
+      "step": 3200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9882190035618976e-06,
+      "loss": 1.1396,
+      "step": 3201
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9870916636617564e-06,
+      "loss": 1.0952,
+      "step": 3202
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9859642207970975e-06,
+      "loss": 1.1367,
+      "step": 3203
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.984836675206245e-06,
+      "loss": 1.0312,
+      "step": 3204
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9837090271275476e-06,
+      "loss": 1.1152,
+      "step": 3205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.982581276799375e-06,
+      "loss": 1.1318,
+      "step": 3206
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.981453424460118e-06,
+      "loss": 1.1416,
+      "step": 3207
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.980325470348189e-06,
+      "loss": 1.1357,
+      "step": 3208
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.979197414702022e-06,
+      "loss": 1.1108,
+      "step": 3209
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.978069257760074e-06,
+      "loss": 1.0967,
+      "step": 3210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9769409997608194e-06,
+      "loss": 1.0898,
+      "step": 3211
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9758126409427575e-06,
+      "loss": 1.0801,
+      "step": 3212
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.974684181544409e-06,
+      "loss": 1.0957,
+      "step": 3213
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.973555621804314e-06,
+      "loss": 1.1387,
+      "step": 3214
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9724269619610347e-06,
+      "loss": 1.1411,
+      "step": 3215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.971298202253154e-06,
+      "loss": 1.1689,
+      "step": 3216
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9701693429192767e-06,
+      "loss": 1.1289,
+      "step": 3217
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.969040384198028e-06,
+      "loss": 1.1143,
+      "step": 3218
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9679113263280544e-06,
+      "loss": 1.0938,
+      "step": 3219
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9667821695480237e-06,
+      "loss": 1.126,
+      "step": 3220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.965652914096624e-06,
+      "loss": 1.1006,
+      "step": 3221
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.964523560212564e-06,
+      "loss": 1.1177,
+      "step": 3222
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9633941081345733e-06,
+      "loss": 1.0938,
+      "step": 3223
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9622645581014038e-06,
+      "loss": 1.1016,
+      "step": 3224
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9611349103518267e-06,
+      "loss": 1.0488,
+      "step": 3225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9600051651246337e-06,
+      "loss": 1.1133,
+      "step": 3226
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9588753226586374e-06,
+      "loss": 1.0508,
+      "step": 3227
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9577453831926713e-06,
+      "loss": 1.1216,
+      "step": 3228
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9566153469655894e-06,
+      "loss": 1.0664,
+      "step": 3229
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9554852142162647e-06,
+      "loss": 1.1416,
+      "step": 3230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.954354985183594e-06,
+      "loss": 1.0957,
+      "step": 3231
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9532246601064907e-06,
+      "loss": 1.0898,
+      "step": 3232
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9520942392238903e-06,
+      "loss": 1.1201,
+      "step": 3233
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9509637227747485e-06,
+      "loss": 1.0776,
+      "step": 3234
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9498331109980404e-06,
+      "loss": 1.126,
+      "step": 3235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9487024041327627e-06,
+      "loss": 1.0918,
+      "step": 3236
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.947571602417931e-06,
+      "loss": 1.083,
+      "step": 3237
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9464407060925815e-06,
+      "loss": 1.1309,
+      "step": 3238
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9453097153957704e-06,
+      "loss": 1.1338,
+      "step": 3239
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.944178630566574e-06,
+      "loss": 1.1514,
+      "step": 3240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.943047451844086e-06,
+      "loss": 1.0913,
+      "step": 3241
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9419161794674238e-06,
+      "loss": 1.0967,
+      "step": 3242
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.940784813675724e-06,
+      "loss": 1.0552,
+      "step": 3243
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9396533547081394e-06,
+      "loss": 1.1211,
+      "step": 3244
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.938521802803847e-06,
+      "loss": 1.1152,
+      "step": 3245
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9373901582020403e-06,
+      "loss": 1.1562,
+      "step": 3246
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9362584211419348e-06,
+      "loss": 1.085,
+      "step": 3247
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9351265918627616e-06,
+      "loss": 1.165,
+      "step": 3248
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9339946706037762e-06,
+      "loss": 1.1099,
+      "step": 3249
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9328626576042507e-06,
+      "loss": 1.0996,
+      "step": 3250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9317305531034778e-06,
+      "loss": 1.1406,
+      "step": 3251
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.930598357340768e-06,
+      "loss": 1.1348,
+      "step": 3252
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9294660705554516e-06,
+      "loss": 1.1372,
+      "step": 3253
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.92833369298688e-06,
+      "loss": 1.1172,
+      "step": 3254
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.927201224874421e-06,
+      "loss": 1.1328,
+      "step": 3255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9260686664574633e-06,
+      "loss": 1.1138,
+      "step": 3256
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.924936017975415e-06,
+      "loss": 1.1621,
+      "step": 3257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9238032796677017e-06,
+      "loss": 1.0654,
+      "step": 3258
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9226704517737685e-06,
+      "loss": 1.1494,
+      "step": 3259
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.921537534533081e-06,
+      "loss": 1.0981,
+      "step": 3260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9204045281851203e-06,
+      "loss": 1.1113,
+      "step": 3261
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.919271432969391e-06,
+      "loss": 1.1328,
+      "step": 3262
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9181382491254123e-06,
+      "loss": 1.1514,
+      "step": 3263
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9170049768927245e-06,
+      "loss": 1.1348,
+      "step": 3264
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9158716165108852e-06,
+      "loss": 1.0986,
+      "step": 3265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.914738168219472e-06,
+      "loss": 1.1406,
+      "step": 3266
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9136046322580793e-06,
+      "loss": 1.1162,
+      "step": 3267
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9124710088663224e-06,
+      "loss": 1.1328,
+      "step": 3268
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9113372982838335e-06,
+      "loss": 1.1113,
+      "step": 3269
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9102035007502625e-06,
+      "loss": 1.1514,
+      "step": 3270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9090696165052794e-06,
+      "loss": 1.1621,
+      "step": 3271
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.907935645788572e-06,
+      "loss": 1.1094,
+      "step": 3272
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9068015888398464e-06,
+      "loss": 1.168,
+      "step": 3273
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9056674458988264e-06,
+      "loss": 1.1309,
+      "step": 3274
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.904533217205254e-06,
+      "loss": 1.1143,
+      "step": 3275
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9033989029988895e-06,
+      "loss": 1.145,
+      "step": 3276
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9022645035195135e-06,
+      "loss": 1.0908,
+      "step": 3277
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9011300190069196e-06,
+      "loss": 1.1333,
+      "step": 3278
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8999954497009232e-06,
+      "loss": 1.126,
+      "step": 3279
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8988607958413583e-06,
+      "loss": 1.1533,
+      "step": 3280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8977260576680733e-06,
+      "loss": 1.0942,
+      "step": 3281
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8965912354209373e-06,
+      "loss": 1.1084,
+      "step": 3282
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8954563293398363e-06,
+      "loss": 1.0708,
+      "step": 3283
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.894321339664673e-06,
+      "loss": 1.124,
+      "step": 3284
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.89318626663537e-06,
+      "loss": 1.0791,
+      "step": 3285
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8920511104918654e-06,
+      "loss": 1.1167,
+      "step": 3286
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8909158714741155e-06,
+      "loss": 1.1289,
+      "step": 3287
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8897805498220954e-06,
+      "loss": 1.1592,
+      "step": 3288
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8886451457757953e-06,
+      "loss": 1.1348,
+      "step": 3289
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.887509659575225e-06,
+      "loss": 1.1172,
+      "step": 3290
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.88637409146041e-06,
+      "loss": 1.0957,
+      "step": 3291
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8852384416713942e-06,
+      "loss": 1.1367,
+      "step": 3292
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8841027104482388e-06,
+      "loss": 1.1299,
+      "step": 3293
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.882966898031021e-06,
+      "loss": 1.1143,
+      "step": 3294
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.881831004659837e-06,
+      "loss": 1.0835,
+      "step": 3295
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8806950305747993e-06,
+      "loss": 1.084,
+      "step": 3296
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.879558976016036e-06,
+      "loss": 1.1055,
+      "step": 3297
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8784228412236938e-06,
+      "loss": 1.1084,
+      "step": 3298
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.877286626437937e-06,
+      "loss": 1.1484,
+      "step": 3299
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.876150331898946e-06,
+      "loss": 1.0928,
+      "step": 3300
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.875013957846916e-06,
+      "loss": 1.1279,
+      "step": 3301
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.873877504522063e-06,
+      "loss": 1.1318,
+      "step": 3302
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8727409721646166e-06,
+      "loss": 1.0801,
+      "step": 3303
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8716043610148246e-06,
+      "loss": 1.1108,
+      "step": 3304
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8704676713129507e-06,
+      "loss": 1.1152,
+      "step": 3305
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.869330903299277e-06,
+      "loss": 1.0781,
+      "step": 3306
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.868194057214099e-06,
+      "loss": 1.1074,
+      "step": 3307
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8670571332977303e-06,
+      "loss": 1.126,
+      "step": 3308
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.865920131790501e-06,
+      "loss": 1.1455,
+      "step": 3309
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8647830529327603e-06,
+      "loss": 1.0723,
+      "step": 3310
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8636458969648684e-06,
+      "loss": 1.1113,
+      "step": 3311
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8625086641272047e-06,
+      "loss": 1.168,
+      "step": 3312
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8613713546601664e-06,
+      "loss": 1.1104,
+      "step": 3313
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8602339688041643e-06,
+      "loss": 1.124,
+      "step": 3314
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8590965067996246e-06,
+      "loss": 1.1299,
+      "step": 3315
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8579589688869937e-06,
+      "loss": 1.1035,
+      "step": 3316
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8568213553067305e-06,
+      "loss": 1.1465,
+      "step": 3317
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.855683666299311e-06,
+      "loss": 1.127,
+      "step": 3318
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8545459021052267e-06,
+      "loss": 1.1182,
+      "step": 3319
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8534080629649856e-06,
+      "loss": 1.123,
+      "step": 3320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8522701491191117e-06,
+      "loss": 1.1299,
+      "step": 3321
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8511321608081443e-06,
+      "loss": 1.1074,
+      "step": 3322
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8499940982726385e-06,
+      "loss": 1.1416,
+      "step": 3323
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8488559617531654e-06,
+      "loss": 1.1504,
+      "step": 3324
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8477177514903114e-06,
+      "loss": 1.1016,
+      "step": 3325
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8465794677246783e-06,
+      "loss": 1.1455,
+      "step": 3326
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.845441110696884e-06,
+      "loss": 1.125,
+      "step": 3327
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8443026806475605e-06,
+      "loss": 1.0698,
+      "step": 3328
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.843164177817358e-06,
+      "loss": 1.1162,
+      "step": 3329
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.84202560244694e-06,
+      "loss": 1.0986,
+      "step": 3330
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.840886954776985e-06,
+      "loss": 1.1348,
+      "step": 3331
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.839748235048189e-06,
+      "loss": 1.1621,
+      "step": 3332
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.838609443501261e-06,
+      "loss": 1.1367,
+      "step": 3333
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.837470580376924e-06,
+      "loss": 1.082,
+      "step": 3334
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.836331645915921e-06,
+      "loss": 1.1074,
+      "step": 3335
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.835192640359007e-06,
+      "loss": 1.1328,
+      "step": 3336
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8340535639469505e-06,
+      "loss": 1.0947,
+      "step": 3337
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8329144169205374e-06,
+      "loss": 1.1143,
+      "step": 3338
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8317751995205683e-06,
+      "loss": 1.1465,
+      "step": 3339
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8306359119878572e-06,
+      "loss": 1.0957,
+      "step": 3340
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.829496554563235e-06,
+      "loss": 1.0825,
+      "step": 3341
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.828357127487546e-06,
+      "loss": 1.0918,
+      "step": 3342
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8272176310016487e-06,
+      "loss": 1.0962,
+      "step": 3343
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8260780653464186e-06,
+      "loss": 1.083,
+      "step": 3344
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8249384307627427e-06,
+      "loss": 1.1396,
+      "step": 3345
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.823798727491524e-06,
+      "loss": 1.1309,
+      "step": 3346
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8226589557736824e-06,
+      "loss": 1.1133,
+      "step": 3347
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8215191158501483e-06,
+      "loss": 1.0938,
+      "step": 3348
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.820379207961868e-06,
+      "loss": 1.0991,
+      "step": 3349
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.819239232349804e-06,
+      "loss": 1.1211,
+      "step": 3350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8180991892549303e-06,
+      "loss": 1.0498,
+      "step": 3351
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8169590789182365e-06,
+      "loss": 1.0977,
+      "step": 3352
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8158189015807267e-06,
+      "loss": 1.167,
+      "step": 3353
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.814678657483419e-06,
+      "loss": 1.1299,
+      "step": 3354
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8135383468673454e-06,
+      "loss": 1.0713,
+      "step": 3355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8123979699735514e-06,
+      "loss": 1.0938,
+      "step": 3356
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8112575270430973e-06,
+      "loss": 1.167,
+      "step": 3357
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8101170183170577e-06,
+      "loss": 1.082,
+      "step": 3358
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8089764440365196e-06,
+      "loss": 1.0947,
+      "step": 3359
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8078358044425856e-06,
+      "loss": 1.1465,
+      "step": 3360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8066950997763708e-06,
+      "loss": 1.0713,
+      "step": 3361
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8055543302790055e-06,
+      "loss": 1.0781,
+      "step": 3362
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8044134961916315e-06,
+      "loss": 1.1025,
+      "step": 3363
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8032725977554055e-06,
+      "loss": 1.2051,
+      "step": 3364
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8021316352114985e-06,
+      "loss": 1.125,
+      "step": 3365
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8009906088010956e-06,
+      "loss": 1.1357,
+      "step": 3366
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7998495187653922e-06,
+      "loss": 1.1152,
+      "step": 3367
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7987083653455998e-06,
+      "loss": 1.1211,
+      "step": 3368
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7975671487829426e-06,
+      "loss": 1.0669,
+      "step": 3369
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.796425869318658e-06,
+      "loss": 1.0879,
+      "step": 3370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7952845271939977e-06,
+      "loss": 1.1055,
+      "step": 3371
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.794143122650225e-06,
+      "loss": 1.1084,
+      "step": 3372
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.793001655928618e-06,
+      "loss": 1.167,
+      "step": 3373
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.791860127270466e-06,
+      "loss": 1.1211,
+      "step": 3374
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.790718536917073e-06,
+      "loss": 1.1377,
+      "step": 3375
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7895768851097554e-06,
+      "loss": 1.0562,
+      "step": 3376
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7884351720898443e-06,
+      "loss": 1.0625,
+      "step": 3377
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7872933980986806e-06,
+      "loss": 1.0967,
+      "step": 3378
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7861515633776203e-06,
+      "loss": 1.1118,
+      "step": 3379
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.785009668168031e-06,
+      "loss": 1.0752,
+      "step": 3380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.783867712711296e-06,
+      "loss": 1.1055,
+      "step": 3381
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.782725697248805e-06,
+      "loss": 1.0747,
+      "step": 3382
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7815836220219677e-06,
+      "loss": 1.1074,
+      "step": 3383
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.780441487272203e-06,
+      "loss": 1.0854,
+      "step": 3384
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.779299293240941e-06,
+      "loss": 1.0566,
+      "step": 3385
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.778157040169627e-06,
+      "loss": 1.0835,
+      "step": 3386
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.777014728299718e-06,
+      "loss": 1.1006,
+      "step": 3387
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7758723578726815e-06,
+      "loss": 1.1094,
+      "step": 3388
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7747299291300004e-06,
+      "loss": 1.0928,
+      "step": 3389
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.773587442313169e-06,
+      "loss": 1.0889,
+      "step": 3390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.772444897663692e-06,
+      "loss": 1.1191,
+      "step": 3391
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7713022954230883e-06,
+      "loss": 1.1182,
+      "step": 3392
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7701596358328887e-06,
+      "loss": 1.1099,
+      "step": 3393
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.769016919134635e-06,
+      "loss": 1.1357,
+      "step": 3394
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.767874145569882e-06,
+      "loss": 1.1025,
+      "step": 3395
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.766731315380198e-06,
+      "loss": 1.1089,
+      "step": 3396
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7655884288071595e-06,
+      "loss": 1.082,
+      "step": 3397
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.764445486092358e-06,
+      "loss": 1.0801,
+      "step": 3398
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7633024874773968e-06,
+      "loss": 1.1416,
+      "step": 3399
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7621594332038894e-06,
+      "loss": 1.0889,
+      "step": 3400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7610163235134607e-06,
+      "loss": 1.1133,
+      "step": 3401
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.75987315864775e-06,
+      "loss": 1.124,
+      "step": 3402
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.758729938848407e-06,
+      "loss": 1.1104,
+      "step": 3403
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.757586664357092e-06,
+      "loss": 1.1357,
+      "step": 3404
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7564433354154774e-06,
+      "loss": 1.0474,
+      "step": 3405
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.755299952265248e-06,
+      "loss": 1.1445,
+      "step": 3406
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.754156515148098e-06,
+      "loss": 1.1309,
+      "step": 3407
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.753013024305736e-06,
+      "loss": 1.0718,
+      "step": 3408
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7518694799798797e-06,
+      "loss": 1.1113,
+      "step": 3409
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.750725882412259e-06,
+      "loss": 1.0996,
+      "step": 3410
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.749582231844614e-06,
+      "loss": 1.0889,
+      "step": 3411
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.748438528518697e-06,
+      "loss": 1.1074,
+      "step": 3412
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7472947726762716e-06,
+      "loss": 1.1006,
+      "step": 3413
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.746150964559113e-06,
+      "loss": 1.1523,
+      "step": 3414
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.745007104409005e-06,
+      "loss": 1.0908,
+      "step": 3415
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7438631924677443e-06,
+      "loss": 1.1445,
+      "step": 3416
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.742719228977139e-06,
+      "loss": 1.0938,
+      "step": 3417
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7415752141790075e-06,
+      "loss": 1.0981,
+      "step": 3418
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7404311483151775e-06,
+      "loss": 1.0918,
+      "step": 3419
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.73928703162749e-06,
+      "loss": 1.0928,
+      "step": 3420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7381428643577958e-06,
+      "loss": 1.1104,
+      "step": 3421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7369986467479553e-06,
+      "loss": 1.1348,
+      "step": 3422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7358543790398408e-06,
+      "loss": 1.1055,
+      "step": 3423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7347100614753352e-06,
+      "loss": 1.1406,
+      "step": 3424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7335656942963313e-06,
+      "loss": 1.0918,
+      "step": 3425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.732421277744733e-06,
+      "loss": 1.1387,
+      "step": 3426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.731276812062453e-06,
+      "loss": 1.0869,
+      "step": 3427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.730132297491418e-06,
+      "loss": 1.1572,
+      "step": 3428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7289877342735614e-06,
+      "loss": 1.1436,
+      "step": 3429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7278431226508282e-06,
+      "loss": 1.1025,
+      "step": 3430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7266984628651732e-06,
+      "loss": 1.1353,
+      "step": 3431
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.725553755158563e-06,
+      "loss": 1.1338,
+      "step": 3432
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.724408999772973e-06,
+      "loss": 1.1094,
+      "step": 3433
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7232641969503875e-06,
+      "loss": 1.1367,
+      "step": 3434
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7221193469328033e-06,
+      "loss": 1.0801,
+      "step": 3435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.720974449962226e-06,
+      "loss": 1.1021,
+      "step": 3436
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7198295062806706e-06,
+      "loss": 1.1455,
+      "step": 3437
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7186845161301628e-06,
+      "loss": 1.1357,
+      "step": 3438
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7175394797527383e-06,
+      "loss": 1.1318,
+      "step": 3439
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7163943973904422e-06,
+      "loss": 1.1406,
+      "step": 3440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7152492692853283e-06,
+      "loss": 1.1362,
+      "step": 3441
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7141040956794616e-06,
+      "loss": 1.1416,
+      "step": 3442
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.712958876814916e-06,
+      "loss": 1.0928,
+      "step": 3443
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7118136129337757e-06,
+      "loss": 1.0654,
+      "step": 3444
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7106683042781334e-06,
+      "loss": 1.0498,
+      "step": 3445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7095229510900917e-06,
+      "loss": 1.1582,
+      "step": 3446
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7083775536117623e-06,
+      "loss": 1.1348,
+      "step": 3447
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7072321120852675e-06,
+      "loss": 1.1157,
+      "step": 3448
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.706086626752736e-06,
+      "loss": 1.0986,
+      "step": 3449
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7049410978563097e-06,
+      "loss": 1.1445,
+      "step": 3450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7037955256381373e-06,
+      "loss": 1.1191,
+      "step": 3451
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.702649910340377e-06,
+      "loss": 1.1426,
+      "step": 3452
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7015042522051955e-06,
+      "loss": 1.1094,
+      "step": 3453
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7003585514747705e-06,
+      "loss": 1.0752,
+      "step": 3454
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6992128083912867e-06,
+      "loss": 1.0874,
+      "step": 3455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6980670231969385e-06,
+      "loss": 1.1143,
+      "step": 3456
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6969211961339292e-06,
+      "loss": 1.1123,
+      "step": 3457
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.695775327444472e-06,
+      "loss": 1.1113,
+      "step": 3458
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6946294173707864e-06,
+      "loss": 1.1216,
+      "step": 3459
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6934834661551037e-06,
+      "loss": 1.1348,
+      "step": 3460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.69233747403966e-06,
+      "loss": 1.127,
+      "step": 3461
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6911914412667043e-06,
+      "loss": 1.126,
+      "step": 3462
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6900453680784923e-06,
+      "loss": 1.0991,
+      "step": 3463
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6888992547172872e-06,
+      "loss": 1.1138,
+      "step": 3464
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.687753101425362e-06,
+      "loss": 1.1094,
+      "step": 3465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.686606908444999e-06,
+      "loss": 1.1309,
+      "step": 3466
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6854606760184853e-06,
+      "loss": 1.1553,
+      "step": 3467
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6843144043881203e-06,
+      "loss": 1.1084,
+      "step": 3468
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6831680937962107e-06,
+      "loss": 1.1396,
+      "step": 3469
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.68202174448507e-06,
+      "loss": 1.1553,
+      "step": 3470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.680875356697021e-06,
+      "loss": 1.1348,
+      "step": 3471
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.679728930674394e-06,
+      "loss": 1.0562,
+      "step": 3472
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6785824666595284e-06,
+      "loss": 1.1768,
+      "step": 3473
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6774359648947704e-06,
+      "loss": 1.1338,
+      "step": 3474
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.676289425622476e-06,
+      "loss": 1.1177,
+      "step": 3475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.675142849085007e-06,
+      "loss": 1.0684,
+      "step": 3476
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.673996235524734e-06,
+      "loss": 1.0054,
+      "step": 3477
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.672849585184035e-06,
+      "loss": 1.1123,
+      "step": 3478
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6717028983052973e-06,
+      "loss": 1.1484,
+      "step": 3479
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6705561751309143e-06,
+      "loss": 1.0815,
+      "step": 3480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6694094159032875e-06,
+      "loss": 1.0957,
+      "step": 3481
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.668262620864826e-06,
+      "loss": 1.0747,
+      "step": 3482
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.667115790257947e-06,
+      "loss": 1.1104,
+      "step": 3483
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.665968924325074e-06,
+      "loss": 1.1113,
+      "step": 3484
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.664822023308641e-06,
+      "loss": 1.0742,
+      "step": 3485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6636750874510835e-06,
+      "loss": 1.0908,
+      "step": 3486
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6625281169948515e-06,
+      "loss": 1.1045,
+      "step": 3487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6613811121823973e-06,
+      "loss": 1.0869,
+      "step": 3488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6602340732561828e-06,
+      "loss": 1.1016,
+      "step": 3489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.659087000458675e-06,
+      "loss": 1.165,
+      "step": 3490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.65793989403235e-06,
+      "loss": 1.0737,
+      "step": 3491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.656792754219691e-06,
+      "loss": 1.0898,
+      "step": 3492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6556455812631872e-06,
+      "loss": 1.125,
+      "step": 3493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6544983754053356e-06,
+      "loss": 1.1279,
+      "step": 3494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6533511368886393e-06,
+      "loss": 1.1006,
+      "step": 3495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.65220386595561e-06,
+      "loss": 1.0986,
+      "step": 3496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.651056562848764e-06,
+      "loss": 1.0991,
+      "step": 3497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.649909227810624e-06,
+      "loss": 1.0967,
+      "step": 3498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.648761861083725e-06,
+      "loss": 1.1025,
+      "step": 3499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6476144629106015e-06,
+      "loss": 1.0918,
+      "step": 3500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.646467033533798e-06,
+      "loss": 1.1426,
+      "step": 3501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6453195731958663e-06,
+      "loss": 1.0806,
+      "step": 3502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6441720821393646e-06,
+      "loss": 1.124,
+      "step": 3503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6430245606068543e-06,
+      "loss": 1.1064,
+      "step": 3504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6418770088409073e-06,
+      "loss": 1.1348,
+      "step": 3505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.640729427084101e-06,
+      "loss": 1.1201,
+      "step": 3506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.639581815579018e-06,
+      "loss": 1.1309,
+      "step": 3507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.638434174568247e-06,
+      "loss": 1.1182,
+      "step": 3508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6372865042943837e-06,
+      "loss": 1.1094,
+      "step": 3509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.63613880500003e-06,
+      "loss": 1.0796,
+      "step": 3510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6349910769277954e-06,
+      "loss": 1.1167,
+      "step": 3511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6338433203202916e-06,
+      "loss": 1.1348,
+      "step": 3512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.63269553542014e-06,
+      "loss": 1.1191,
+      "step": 3513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6315477224699663e-06,
+      "loss": 1.1367,
+      "step": 3514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6303998817124027e-06,
+      "loss": 1.1104,
+      "step": 3515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.629252013390086e-06,
+      "loss": 1.0938,
+      "step": 3516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.628104117745661e-06,
+      "loss": 1.0791,
+      "step": 3517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.626956195021778e-06,
+      "loss": 1.1484,
+      "step": 3518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.62580824546109e-06,
+      "loss": 1.1504,
+      "step": 3519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.624660269306259e-06,
+      "loss": 1.0791,
+      "step": 3520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6235122667999512e-06,
+      "loss": 1.1348,
+      "step": 3521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6223642381848393e-06,
+      "loss": 1.0869,
+      "step": 3522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6212161837035993e-06,
+      "loss": 1.0933,
+      "step": 3523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.620068103598916e-06,
+      "loss": 1.1494,
+      "step": 3524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6189199981134774e-06,
+      "loss": 1.1309,
+      "step": 3525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6177718674899767e-06,
+      "loss": 1.0674,
+      "step": 3526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.616623711971113e-06,
+      "loss": 1.0903,
+      "step": 3527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.615475531799591e-06,
+      "loss": 1.1211,
+      "step": 3528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.61432732721812e-06,
+      "loss": 1.0908,
+      "step": 3529
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6131790984694155e-06,
+      "loss": 1.0996,
+      "step": 3530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6120308457961967e-06,
+      "loss": 1.1338,
+      "step": 3531
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6108825694411886e-06,
+      "loss": 1.1201,
+      "step": 3532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6097342696471225e-06,
+      "loss": 1.1167,
+      "step": 3533
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6085859466567304e-06,
+      "loss": 1.1396,
+      "step": 3534
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6074376007127535e-06,
+      "loss": 1.1562,
+      "step": 3535
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6062892320579373e-06,
+      "loss": 1.168,
+      "step": 3536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.605140840935031e-06,
+      "loss": 1.126,
+      "step": 3537
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6039924275867877e-06,
+      "loss": 1.0918,
+      "step": 3538
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.602843992255967e-06,
+      "loss": 1.127,
+      "step": 3539
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.601695535185333e-06,
+      "loss": 1.1699,
+      "step": 3540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6005470566176517e-06,
+      "loss": 1.0947,
+      "step": 3541
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.599398556795698e-06,
+      "loss": 1.1387,
+      "step": 3542
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5982500359622488e-06,
+      "loss": 1.1279,
+      "step": 3543
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.597101494360085e-06,
+      "loss": 1.1582,
+      "step": 3544
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5959529322319927e-06,
+      "loss": 1.1089,
+      "step": 3545
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.594804349820762e-06,
+      "loss": 1.1475,
+      "step": 3546
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5936557473691877e-06,
+      "loss": 1.1689,
+      "step": 3547
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.592507125120069e-06,
+      "loss": 1.1436,
+      "step": 3548
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5913584833162086e-06,
+      "loss": 1.0732,
+      "step": 3549
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.590209822200414e-06,
+      "loss": 1.1348,
+      "step": 3550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.589061142015497e-06,
+      "loss": 1.0693,
+      "step": 3551
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5879124430042713e-06,
+      "loss": 1.123,
+      "step": 3552
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.586763725409557e-06,
+      "loss": 1.127,
+      "step": 3553
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.585614989474177e-06,
+      "loss": 1.1318,
+      "step": 3554
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.58446623544096e-06,
+      "loss": 1.1191,
+      "step": 3555
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5833174635527346e-06,
+      "loss": 1.1475,
+      "step": 3556
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5821686740523366e-06,
+      "loss": 1.1201,
+      "step": 3557
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5810198671826043e-06,
+      "loss": 1.0869,
+      "step": 3558
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.579871043186379e-06,
+      "loss": 1.0879,
+      "step": 3559
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5787222023065084e-06,
+      "loss": 1.085,
+      "step": 3560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.57757334478584e-06,
+      "loss": 1.1064,
+      "step": 3561
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.576424470867227e-06,
+      "loss": 1.0791,
+      "step": 3562
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5752755807935255e-06,
+      "loss": 1.1602,
+      "step": 3563
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5741266748075956e-06,
+      "loss": 1.1221,
+      "step": 3564
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5729777531522993e-06,
+      "loss": 1.1074,
+      "step": 3565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5718288160705045e-06,
+      "loss": 1.1094,
+      "step": 3566
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5706798638050795e-06,
+      "loss": 1.1475,
+      "step": 3567
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.569530896598898e-06,
+      "loss": 1.1196,
+      "step": 3568
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.568381914694835e-06,
+      "loss": 1.1201,
+      "step": 3569
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.567232918335771e-06,
+      "loss": 1.1406,
+      "step": 3570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.566083907764586e-06,
+      "loss": 1.0928,
+      "step": 3571
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5649348832241666e-06,
+      "loss": 1.1533,
+      "step": 3572
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5637858449574016e-06,
+      "loss": 1.1035,
+      "step": 3573
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5626367932071806e-06,
+      "loss": 1.0947,
+      "step": 3574
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5614877282163973e-06,
+      "loss": 1.0693,
+      "step": 3575
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5603386502279494e-06,
+      "loss": 1.062,
+      "step": 3576
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5591895594847353e-06,
+      "loss": 1.1426,
+      "step": 3577
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5580404562296574e-06,
+      "loss": 1.0874,
+      "step": 3578
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.556891340705621e-06,
+      "loss": 1.0649,
+      "step": 3579
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5557422131555327e-06,
+      "loss": 1.0679,
+      "step": 3580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5545930738223036e-06,
+      "loss": 1.1406,
+      "step": 3581
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.553443922948844e-06,
+      "loss": 1.1221,
+      "step": 3582
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.552294760778069e-06,
+      "loss": 1.1387,
+      "step": 3583
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5511455875528977e-06,
+      "loss": 1.1387,
+      "step": 3584
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5499964035162483e-06,
+      "loss": 1.1494,
+      "step": 3585
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5488472089110426e-06,
+      "loss": 1.1357,
+      "step": 3586
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5476980039802046e-06,
+      "loss": 1.1426,
+      "step": 3587
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.546548788966661e-06,
+      "loss": 1.125,
+      "step": 3588
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5453995641133393e-06,
+      "loss": 1.1006,
+      "step": 3589
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5442503296631705e-06,
+      "loss": 1.123,
+      "step": 3590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.543101085859087e-06,
+      "loss": 1.0645,
+      "step": 3591
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.541951832944024e-06,
+      "loss": 1.145,
+      "step": 3592
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.540802571160916e-06,
+      "loss": 1.105,
+      "step": 3593
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5396533007527023e-06,
+      "loss": 1.1426,
+      "step": 3594
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5385040219623235e-06,
+      "loss": 1.0786,
+      "step": 3595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5373547350327203e-06,
+      "loss": 1.1387,
+      "step": 3596
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.536205440206837e-06,
+      "loss": 1.084,
+      "step": 3597
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5350561377276188e-06,
+      "loss": 1.1387,
+      "step": 3598
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.533906827838013e-06,
+      "loss": 1.1221,
+      "step": 3599
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5327575107809664e-06,
+      "loss": 1.1362,
+      "step": 3600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5316081867994304e-06,
+      "loss": 1.1055,
+      "step": 3601
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5304588561363545e-06,
+      "loss": 1.1289,
+      "step": 3602
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5293095190346945e-06,
+      "loss": 1.0928,
+      "step": 3603
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5281601757374017e-06,
+      "loss": 1.1221,
+      "step": 3604
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5270108264874334e-06,
+      "loss": 1.0298,
+      "step": 3605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.525861471527745e-06,
+      "loss": 1.123,
+      "step": 3606
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5247121111012964e-06,
+      "loss": 1.127,
+      "step": 3607
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5235627454510435e-06,
+      "loss": 1.085,
+      "step": 3608
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5224133748199497e-06,
+      "loss": 1.127,
+      "step": 3609
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5212639994509746e-06,
+      "loss": 1.1055,
+      "step": 3610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5201146195870807e-06,
+      "loss": 1.1035,
+      "step": 3611
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5189652354712313e-06,
+      "loss": 1.127,
+      "step": 3612
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5178158473463906e-06,
+      "loss": 1.1143,
+      "step": 3613
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5166664554555224e-06,
+      "loss": 1.1357,
+      "step": 3614
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.515517060041594e-06,
+      "loss": 1.1182,
+      "step": 3615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.514367661347571e-06,
+      "loss": 1.0557,
+      "step": 3616
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5132182596164212e-06,
+      "loss": 1.1123,
+      "step": 3617
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.512068855091112e-06,
+      "loss": 1.1021,
+      "step": 3618
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.510919448014612e-06,
+      "loss": 1.1182,
+      "step": 3619
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.509770038629889e-06,
+      "loss": 1.1279,
+      "step": 3620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5086206271799144e-06,
+      "loss": 1.124,
+      "step": 3621
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.507471213907657e-06,
+      "loss": 1.1758,
+      "step": 3622
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5063217990560863e-06,
+      "loss": 1.123,
+      "step": 3623
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5051723828681746e-06,
+      "loss": 1.1123,
+      "step": 3624
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.504022965586891e-06,
+      "loss": 1.1328,
+      "step": 3625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.502873547455208e-06,
+      "loss": 1.1123,
+      "step": 3626
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5017241287160965e-06,
+      "loss": 1.1172,
+      "step": 3627
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5005747096125275e-06,
+      "loss": 1.0576,
+      "step": 3628
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.499425290387473e-06,
+      "loss": 1.1274,
+      "step": 3629
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4982758712839035e-06,
+      "loss": 1.1377,
+      "step": 3630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4971264525447917e-06,
+      "loss": 1.1787,
+      "step": 3631
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4959770344131096e-06,
+      "loss": 1.1328,
+      "step": 3632
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4948276171318262e-06,
+      "loss": 1.0879,
+      "step": 3633
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.493678200943914e-06,
+      "loss": 1.126,
+      "step": 3634
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4925287860923443e-06,
+      "loss": 1.083,
+      "step": 3635
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4913793728200864e-06,
+      "loss": 1.1152,
+      "step": 3636
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4902299613701115e-06,
+      "loss": 1.1172,
+      "step": 3637
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4890805519853893e-06,
+      "loss": 1.1094,
+      "step": 3638
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.487931144908889e-06,
+      "loss": 1.0693,
+      "step": 3639
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4867817403835796e-06,
+      "loss": 1.1128,
+      "step": 3640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.485632338652429e-06,
+      "loss": 1.125,
+      "step": 3641
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.484482939958406e-06,
+      "loss": 1.1016,
+      "step": 3642
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.483333544544478e-06,
+      "loss": 1.0825,
+      "step": 3643
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4821841526536107e-06,
+      "loss": 1.1113,
+      "step": 3644
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4810347645287696e-06,
+      "loss": 1.0796,
+      "step": 3645
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4798853804129198e-06,
+      "loss": 1.1152,
+      "step": 3646
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.478736000549026e-06,
+      "loss": 1.1396,
+      "step": 3647
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.477586625180051e-06,
+      "loss": 1.1045,
+      "step": 3648
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.476437254548957e-06,
+      "loss": 1.1455,
+      "step": 3649
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.475287888898705e-06,
+      "loss": 1.1177,
+      "step": 3650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4741385284722553e-06,
+      "loss": 1.1123,
+      "step": 3651
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.472989173512568e-06,
+      "loss": 1.1367,
+      "step": 3652
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4718398242625982e-06,
+      "loss": 1.1045,
+      "step": 3653
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4706904809653064e-06,
+      "loss": 1.1025,
+      "step": 3654
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.469541143863646e-06,
+      "loss": 1.0947,
+      "step": 3655
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.468391813200571e-06,
+      "loss": 1.1846,
+      "step": 3656
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4672424892190344e-06,
+      "loss": 1.1289,
+      "step": 3657
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.466093172161988e-06,
+      "loss": 1.1396,
+      "step": 3658
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4649438622723816e-06,
+      "loss": 1.1123,
+      "step": 3659
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4637945597931632e-06,
+      "loss": 1.1553,
+      "step": 3660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.46264526496728e-06,
+      "loss": 1.1201,
+      "step": 3661
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4614959780376774e-06,
+      "loss": 1.0889,
+      "step": 3662
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.460346699247299e-06,
+      "loss": 1.127,
+      "step": 3663
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4591974288390853e-06,
+      "loss": 1.0938,
+      "step": 3664
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4580481670559774e-06,
+      "loss": 1.083,
+      "step": 3665
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4568989141409135e-06,
+      "loss": 1.1396,
+      "step": 3666
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.45574967033683e-06,
+      "loss": 1.083,
+      "step": 3667
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.454600435886661e-06,
+      "loss": 1.0601,
+      "step": 3668
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4534512110333393e-06,
+      "loss": 1.106,
+      "step": 3669
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4523019960197963e-06,
+      "loss": 1.1133,
+      "step": 3670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4511527910889582e-06,
+      "loss": 1.0977,
+      "step": 3671
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4500035964837525e-06,
+      "loss": 1.0889,
+      "step": 3672
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4488544124471027e-06,
+      "loss": 1.1572,
+      "step": 3673
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4477052392219315e-06,
+      "loss": 1.1182,
+      "step": 3674
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4465560770511572e-06,
+      "loss": 1.1338,
+      "step": 3675
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.445406926177698e-06,
+      "loss": 1.0703,
+      "step": 3676
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4442577868444677e-06,
+      "loss": 1.1006,
+      "step": 3677
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4431086592943792e-06,
+      "loss": 1.0996,
+      "step": 3678
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4419595437703426e-06,
+      "loss": 1.1553,
+      "step": 3679
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4408104405152656e-06,
+      "loss": 1.124,
+      "step": 3680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4396613497720515e-06,
+      "loss": 1.0986,
+      "step": 3681
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4385122717836035e-06,
+      "loss": 1.0996,
+      "step": 3682
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4373632067928206e-06,
+      "loss": 1.0894,
+      "step": 3683
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4362141550425993e-06,
+      "loss": 1.1553,
+      "step": 3684
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.435065116775834e-06,
+      "loss": 1.0986,
+      "step": 3685
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.433916092235415e-06,
+      "loss": 1.166,
+      "step": 3686
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4327670816642303e-06,
+      "loss": 1.0889,
+      "step": 3687
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4316180853051656e-06,
+      "loss": 1.124,
+      "step": 3688
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4304691034011037e-06,
+      "loss": 1.1348,
+      "step": 3689
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4293201361949205e-06,
+      "loss": 1.0938,
+      "step": 3690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4281711839294963e-06,
+      "loss": 1.1299,
+      "step": 3691
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.427022246847701e-06,
+      "loss": 1.1318,
+      "step": 3692
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4258733251924053e-06,
+      "loss": 1.1177,
+      "step": 3693
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.424724419206475e-06,
+      "loss": 1.1523,
+      "step": 3694
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4235755291327737e-06,
+      "loss": 1.1289,
+      "step": 3695
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.422426655214161e-06,
+      "loss": 1.1353,
+      "step": 3696
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4212777976934924e-06,
+      "loss": 1.103,
+      "step": 3697
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4201289568136212e-06,
+      "loss": 1.0859,
+      "step": 3698
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4189801328173965e-06,
+      "loss": 1.1113,
+      "step": 3699
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4178313259476647e-06,
+      "loss": 1.0977,
+      "step": 3700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4166825364472667e-06,
+      "loss": 1.082,
+      "step": 3701
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4155337645590406e-06,
+      "loss": 1.1543,
+      "step": 3702
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.414385010525823e-06,
+      "loss": 1.1104,
+      "step": 3703
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4132362745904435e-06,
+      "loss": 1.0576,
+      "step": 3704
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.412087556995729e-06,
+      "loss": 1.0908,
+      "step": 3705
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4109388579845037e-06,
+      "loss": 1.1318,
+      "step": 3706
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4097901777995864e-06,
+      "loss": 1.0815,
+      "step": 3707
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.408641516683792e-06,
+      "loss": 1.0698,
+      "step": 3708
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4074928748799313e-06,
+      "loss": 1.0947,
+      "step": 3709
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4063442526308127e-06,
+      "loss": 1.166,
+      "step": 3710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4051956501792393e-06,
+      "loss": 1.1162,
+      "step": 3711
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.404047067768009e-06,
+      "loss": 1.1406,
+      "step": 3712
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4028985056399163e-06,
+      "loss": 1.1572,
+      "step": 3713
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4017499640377516e-06,
+      "loss": 1.1387,
+      "step": 3714
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.400601443204302e-06,
+      "loss": 1.1387,
+      "step": 3715
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3994529433823483e-06,
+      "loss": 1.1011,
+      "step": 3716
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.398304464814668e-06,
+      "loss": 1.0928,
+      "step": 3717
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3971560077440336e-06,
+      "loss": 1.1211,
+      "step": 3718
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.396007572413213e-06,
+      "loss": 1.0981,
+      "step": 3719
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.39485915906497e-06,
+      "loss": 1.1328,
+      "step": 3720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.393710767942063e-06,
+      "loss": 1.1152,
+      "step": 3721
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.392562399287247e-06,
+      "loss": 1.1309,
+      "step": 3722
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.391414053343271e-06,
+      "loss": 1.1553,
+      "step": 3723
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3902657303528792e-06,
+      "loss": 1.1553,
+      "step": 3724
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.389117430558812e-06,
+      "loss": 1.0903,
+      "step": 3725
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3879691542038033e-06,
+      "loss": 1.1182,
+      "step": 3726
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3868209015305845e-06,
+      "loss": 1.0967,
+      "step": 3727
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.38567267278188e-06,
+      "loss": 1.1465,
+      "step": 3728
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3845244682004097e-06,
+      "loss": 1.0854,
+      "step": 3729
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3833762880288875e-06,
+      "loss": 1.1084,
+      "step": 3730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.382228132510024e-06,
+      "loss": 1.085,
+      "step": 3731
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.381080001886523e-06,
+      "loss": 1.0693,
+      "step": 3732
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.379931896401085e-06,
+      "loss": 1.1396,
+      "step": 3733
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.378783816296401e-06,
+      "loss": 1.1191,
+      "step": 3734
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.377635761815162e-06,
+      "loss": 1.1543,
+      "step": 3735
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.376487733200049e-06,
+      "loss": 1.0996,
+      "step": 3736
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3753397306937423e-06,
+      "loss": 1.1338,
+      "step": 3737
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.374191754538911e-06,
+      "loss": 1.1357,
+      "step": 3738
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3730438049782223e-06,
+      "loss": 1.1172,
+      "step": 3739
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.371895882254339e-06,
+      "loss": 1.1094,
+      "step": 3740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3707479866099144e-06,
+      "loss": 1.1357,
+      "step": 3741
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.369600118287598e-06,
+      "loss": 1.1221,
+      "step": 3742
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.368452277530034e-06,
+      "loss": 1.1406,
+      "step": 3743
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.367304464579861e-06,
+      "loss": 1.124,
+      "step": 3744
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.366156679679709e-06,
+      "loss": 1.0698,
+      "step": 3745
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.365008923072206e-06,
+      "loss": 1.1445,
+      "step": 3746
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3638611949999702e-06,
+      "loss": 1.0732,
+      "step": 3747
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3627134957056176e-06,
+      "loss": 1.084,
+      "step": 3748
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3615658254317543e-06,
+      "loss": 1.1162,
+      "step": 3749
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.360418184420983e-06,
+      "loss": 1.1104,
+      "step": 3750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3592705729158993e-06,
+      "loss": 1.1406,
+      "step": 3751
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3581229911590927e-06,
+      "loss": 1.1055,
+      "step": 3752
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.356975439393146e-06,
+      "loss": 1.1152,
+      "step": 3753
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.355827917860636e-06,
+      "loss": 1.1152,
+      "step": 3754
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.354680426804134e-06,
+      "loss": 1.1201,
+      "step": 3755
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3535329664662022e-06,
+      "loss": 1.1914,
+      "step": 3756
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3523855370893993e-06,
+      "loss": 1.1211,
+      "step": 3757
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.351238138916276e-06,
+      "loss": 1.1079,
+      "step": 3758
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3500907721893764e-06,
+      "loss": 1.042,
+      "step": 3759
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3489434371512377e-06,
+      "loss": 1.0947,
+      "step": 3760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3477961340443912e-06,
+      "loss": 1.1309,
+      "step": 3761
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.346648863111361e-06,
+      "loss": 1.1216,
+      "step": 3762
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3455016245946648e-06,
+      "loss": 1.1348,
+      "step": 3763
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3443544187368128e-06,
+      "loss": 1.1729,
+      "step": 3764
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.343207245780309e-06,
+      "loss": 1.1035,
+      "step": 3765
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3420601059676504e-06,
+      "loss": 1.1289,
+      "step": 3766
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.340912999541326e-06,
+      "loss": 1.1484,
+      "step": 3767
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3397659267438185e-06,
+      "loss": 1.1104,
+      "step": 3768
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.338618887817603e-06,
+      "loss": 1.1318,
+      "step": 3769
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3374718830051494e-06,
+      "loss": 1.0996,
+      "step": 3770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.336324912548917e-06,
+      "loss": 1.0752,
+      "step": 3771
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3351779766913604e-06,
+      "loss": 1.1104,
+      "step": 3772
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.334031075674926e-06,
+      "loss": 1.1172,
+      "step": 3773
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3328842097420546e-06,
+      "loss": 1.1421,
+      "step": 3774
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.331737379135174e-06,
+      "loss": 1.124,
+      "step": 3775
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.330590584096713e-06,
+      "loss": 1.0967,
+      "step": 3776
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3294438248690865e-06,
+      "loss": 1.1157,
+      "step": 3777
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.328297101694703e-06,
+      "loss": 1.1162,
+      "step": 3778
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3271504148159653e-06,
+      "loss": 1.0874,
+      "step": 3779
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3260037644752667e-06,
+      "loss": 1.1323,
+      "step": 3780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.324857150914994e-06,
+      "loss": 1.0562,
+      "step": 3781
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3237105743775246e-06,
+      "loss": 1.1289,
+      "step": 3782
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.32256403510523e-06,
+      "loss": 1.0835,
+      "step": 3783
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3214175333404724e-06,
+      "loss": 1.0898,
+      "step": 3784
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.320271069325607e-06,
+      "loss": 1.1289,
+      "step": 3785
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3191246433029804e-06,
+      "loss": 1.0869,
+      "step": 3786
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.31797825551493e-06,
+      "loss": 1.1396,
+      "step": 3787
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.31683190620379e-06,
+      "loss": 1.1299,
+      "step": 3788
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.31568559561188e-06,
+      "loss": 1.085,
+      "step": 3789
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.314539323981515e-06,
+      "loss": 1.0835,
+      "step": 3790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.313393091555002e-06,
+      "loss": 1.1206,
+      "step": 3791
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3122468985746387e-06,
+      "loss": 1.1455,
+      "step": 3792
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3111007452827136e-06,
+      "loss": 1.1348,
+      "step": 3793
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3099546319215085e-06,
+      "loss": 1.0942,
+      "step": 3794
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.308808558733296e-06,
+      "loss": 1.1338,
+      "step": 3795
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3076625259603412e-06,
+      "loss": 1.1265,
+      "step": 3796
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.306516533844898e-06,
+      "loss": 1.1338,
+      "step": 3797
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.305370582629215e-06,
+      "loss": 1.0898,
+      "step": 3798
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.304224672555529e-06,
+      "loss": 1.0806,
+      "step": 3799
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3030788038660708e-06,
+      "loss": 1.1118,
+      "step": 3800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.301932976803062e-06,
+      "loss": 1.0845,
+      "step": 3801
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3007871916087137e-06,
+      "loss": 1.1123,
+      "step": 3802
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2996414485252307e-06,
+      "loss": 1.1475,
+      "step": 3803
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.298495747794805e-06,
+      "loss": 1.1094,
+      "step": 3804
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.297350089659624e-06,
+      "loss": 1.1172,
+      "step": 3805
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.296204474361863e-06,
+      "loss": 1.1328,
+      "step": 3806
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2950589021436916e-06,
+      "loss": 1.1084,
+      "step": 3807
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2939133732472653e-06,
+      "loss": 1.0957,
+      "step": 3808
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.292767887914734e-06,
+      "loss": 1.1553,
+      "step": 3809
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2916224463882385e-06,
+      "loss": 1.1191,
+      "step": 3810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.29047704890991e-06,
+      "loss": 1.1299,
+      "step": 3811
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2893316957218666e-06,
+      "loss": 1.0884,
+      "step": 3812
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2881863870662242e-06,
+      "loss": 1.1045,
+      "step": 3813
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2870411231850844e-06,
+      "loss": 1.1108,
+      "step": 3814
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.285895904320539e-06,
+      "loss": 1.0811,
+      "step": 3815
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.284750730714672e-06,
+      "loss": 1.126,
+      "step": 3816
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2836056026095586e-06,
+      "loss": 1.126,
+      "step": 3817
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2824605202472626e-06,
+      "loss": 1.083,
+      "step": 3818
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.281315483869838e-06,
+      "loss": 1.0527,
+      "step": 3819
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2801704937193302e-06,
+      "loss": 1.126,
+      "step": 3820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.279025550037775e-06,
+      "loss": 1.1064,
+      "step": 3821
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.277880653067198e-06,
+      "loss": 1.1006,
+      "step": 3822
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.276735803049614e-06,
+      "loss": 1.1562,
+      "step": 3823
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2755910002270275e-06,
+      "loss": 1.1318,
+      "step": 3824
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2744462448414373e-06,
+      "loss": 1.1172,
+      "step": 3825
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2733015371348268e-06,
+      "loss": 1.1367,
+      "step": 3826
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.272156877349172e-06,
+      "loss": 1.1084,
+      "step": 3827
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.271012265726439e-06,
+      "loss": 1.1016,
+      "step": 3828
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2698677025085823e-06,
+      "loss": 1.125,
+      "step": 3829
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2687231879375472e-06,
+      "loss": 1.1396,
+      "step": 3830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.267578722255268e-06,
+      "loss": 1.1104,
+      "step": 3831
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2664343057036695e-06,
+      "loss": 1.1348,
+      "step": 3832
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2652899385246665e-06,
+      "loss": 1.1338,
+      "step": 3833
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2641456209601605e-06,
+      "loss": 1.1074,
+      "step": 3834
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.263001353252046e-06,
+      "loss": 1.1094,
+      "step": 3835
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.261857135642205e-06,
+      "loss": 1.0791,
+      "step": 3836
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2607129683725105e-06,
+      "loss": 1.1094,
+      "step": 3837
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.259568851684823e-06,
+      "loss": 1.1338,
+      "step": 3838
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2584247858209933e-06,
+      "loss": 1.1416,
+      "step": 3839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2572807710228616e-06,
+      "loss": 1.125,
+      "step": 3840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.256136807532256e-06,
+      "loss": 1.103,
+      "step": 3841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2549928955909962e-06,
+      "loss": 1.1104,
+      "step": 3842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.253849035440888e-06,
+      "loss": 1.1514,
+      "step": 3843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2527052273237293e-06,
+      "loss": 1.1367,
+      "step": 3844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.251561471481304e-06,
+      "loss": 1.1318,
+      "step": 3845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2504177681553875e-06,
+      "loss": 1.145,
+      "step": 3846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2492741175877424e-06,
+      "loss": 1.1045,
+      "step": 3847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2481305200201207e-06,
+      "loss": 1.124,
+      "step": 3848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2469869756942643e-06,
+      "loss": 1.1094,
+      "step": 3849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.245843484851902e-06,
+      "loss": 1.1279,
+      "step": 3850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.244700047734753e-06,
+      "loss": 1.0903,
+      "step": 3851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2435566645845234e-06,
+      "loss": 1.0986,
+      "step": 3852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2424133356429087e-06,
+      "loss": 1.123,
+      "step": 3853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2412700611515932e-06,
+      "loss": 1.1338,
+      "step": 3854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2401268413522507e-06,
+      "loss": 1.1123,
+      "step": 3855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.23898367648654e-06,
+      "loss": 1.1299,
+      "step": 3856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2378405667961123e-06,
+      "loss": 1.1094,
+      "step": 3857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.236697512522604e-06,
+      "loss": 1.064,
+      "step": 3858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2355545139076427e-06,
+      "loss": 1.1416,
+      "step": 3859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.234411571192841e-06,
+      "loss": 1.082,
+      "step": 3860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2332686846198026e-06,
+      "loss": 1.1377,
+      "step": 3861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2321258544301184e-06,
+      "loss": 1.1455,
+      "step": 3862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.230983080865366e-06,
+      "loss": 1.0786,
+      "step": 3863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2298403641671125e-06,
+      "loss": 1.1069,
+      "step": 3864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.228697704576912e-06,
+      "loss": 1.0859,
+      "step": 3865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.227555102336309e-06,
+      "loss": 1.1348,
+      "step": 3866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.226412557686832e-06,
+      "loss": 1.0674,
+      "step": 3867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.22527007087e-06,
+      "loss": 1.1221,
+      "step": 3868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.224127642127319e-06,
+      "loss": 1.0693,
+      "step": 3869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.222985271700283e-06,
+      "loss": 1.0996,
+      "step": 3870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.221842959830374e-06,
+      "loss": 1.0879,
+      "step": 3871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.22070070675906e-06,
+      "loss": 1.1426,
+      "step": 3872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2195585127277976e-06,
+      "loss": 1.0933,
+      "step": 3873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2184163779780323e-06,
+      "loss": 1.1621,
+      "step": 3874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2172743027511955e-06,
+      "loss": 1.0957,
+      "step": 3875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2161322872887054e-06,
+      "loss": 1.1484,
+      "step": 3876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2149903318319695e-06,
+      "loss": 1.1328,
+      "step": 3877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2138484366223806e-06,
+      "loss": 1.0845,
+      "step": 3878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.21270660190132e-06,
+      "loss": 1.0771,
+      "step": 3879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.211564827910156e-06,
+      "loss": 1.0854,
+      "step": 3880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.210423114890245e-06,
+      "loss": 1.1191,
+      "step": 3881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2092814630829283e-06,
+      "loss": 1.1113,
+      "step": 3882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2081398727295354e-06,
+      "loss": 1.1035,
+      "step": 3883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2069983440713834e-06,
+      "loss": 1.0981,
+      "step": 3884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.205856877349775e-06,
+      "loss": 1.1133,
+      "step": 3885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2047154728060023e-06,
+      "loss": 1.0781,
+      "step": 3886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.203574130681342e-06,
+      "loss": 1.0957,
+      "step": 3887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2024328512170583e-06,
+      "loss": 1.0977,
+      "step": 3888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.201291634654401e-06,
+      "loss": 1.1152,
+      "step": 3889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2001504812346086e-06,
+      "loss": 1.1104,
+      "step": 3890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.199009391198905e-06,
+      "loss": 1.0713,
+      "step": 3891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.197868364788502e-06,
+      "loss": 1.0703,
+      "step": 3892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1967274022445953e-06,
+      "loss": 1.1279,
+      "step": 3893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1955865038083698e-06,
+      "loss": 1.1406,
+      "step": 3894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.194445669720996e-06,
+      "loss": 1.0474,
+      "step": 3895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.19330490022363e-06,
+      "loss": 1.106,
+      "step": 3896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1921641955574148e-06,
+      "loss": 1.1006,
+      "step": 3897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1910235559634804e-06,
+      "loss": 1.1338,
+      "step": 3898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.189882981682943e-06,
+      "loss": 1.0732,
+      "step": 3899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.188742472956903e-06,
+      "loss": 1.0986,
+      "step": 3900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.187602030026449e-06,
+      "loss": 1.0957,
+      "step": 3901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.186461653132655e-06,
+      "loss": 1.0967,
+      "step": 3902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.185321342516582e-06,
+      "loss": 1.0786,
+      "step": 3903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1841810984192737e-06,
+      "loss": 1.1191,
+      "step": 3904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1830409210817643e-06,
+      "loss": 1.1382,
+      "step": 3905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1819008107450705e-06,
+      "loss": 1.0957,
+      "step": 3906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.180760767650197e-06,
+      "loss": 1.0835,
+      "step": 3907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.179620792038133e-06,
+      "loss": 1.1465,
+      "step": 3908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.178480884149852e-06,
+      "loss": 1.1279,
+      "step": 3909
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.177341044226318e-06,
+      "loss": 1.1396,
+      "step": 3910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.176201272508476e-06,
+      "loss": 1.0635,
+      "step": 3911
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.175061569237258e-06,
+      "loss": 1.0669,
+      "step": 3912
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1739219346535823e-06,
+      "loss": 1.0972,
+      "step": 3913
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1727823689983517e-06,
+      "loss": 1.1309,
+      "step": 3914
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.171642872512455e-06,
+      "loss": 1.1245,
+      "step": 3915
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1705034454367653e-06,
+      "loss": 1.1816,
+      "step": 3916
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.169364088012143e-06,
+      "loss": 1.1055,
+      "step": 3917
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1682248004794325e-06,
+      "loss": 1.1353,
+      "step": 3918
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1670855830794634e-06,
+      "loss": 1.1147,
+      "step": 3919
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1659464360530503e-06,
+      "loss": 1.0977,
+      "step": 3920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1648073596409936e-06,
+      "loss": 1.1689,
+      "step": 3921
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.163668354084079e-06,
+      "loss": 1.103,
+      "step": 3922
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.162529419623076e-06,
+      "loss": 1.124,
+      "step": 3923
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.16139055649874e-06,
+      "loss": 1.1289,
+      "step": 3924
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1602517649518116e-06,
+      "loss": 1.1304,
+      "step": 3925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1591130452230153e-06,
+      "loss": 1.1172,
+      "step": 3926
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1579743975530605e-06,
+      "loss": 1.1055,
+      "step": 3927
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1568358221826423e-06,
+      "loss": 1.1152,
+      "step": 3928
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.15569731935244e-06,
+      "loss": 1.1006,
+      "step": 3929
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.154558889303118e-06,
+      "loss": 1.0898,
+      "step": 3930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1534205322753233e-06,
+      "loss": 1.1348,
+      "step": 3931
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.15228224850969e-06,
+      "loss": 1.1133,
+      "step": 3932
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1511440382468355e-06,
+      "loss": 1.0952,
+      "step": 3933
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.150005901727362e-06,
+      "loss": 1.1172,
+      "step": 3934
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1488678391918557e-06,
+      "loss": 1.1328,
+      "step": 3935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1477298508808887e-06,
+      "loss": 1.0957,
+      "step": 3936
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.146591937035015e-06,
+      "loss": 1.1631,
+      "step": 3937
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.145454097894774e-06,
+      "loss": 1.1006,
+      "step": 3938
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.14431633370069e-06,
+      "loss": 1.0923,
+      "step": 3939
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.14317864469327e-06,
+      "loss": 1.125,
+      "step": 3940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.142041031113007e-06,
+      "loss": 1.1392,
+      "step": 3941
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.140903493200376e-06,
+      "loss": 1.1045,
+      "step": 3942
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.139766031195837e-06,
+      "loss": 1.1069,
+      "step": 3943
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1386286453398345e-06,
+      "loss": 1.1069,
+      "step": 3944
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.137491335872796e-06,
+      "loss": 1.1172,
+      "step": 3945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.136354103035132e-06,
+      "loss": 1.1338,
+      "step": 3946
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1352169470672405e-06,
+      "loss": 1.0854,
+      "step": 3947
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.134079868209499e-06,
+      "loss": 1.1367,
+      "step": 3948
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1329428667022706e-06,
+      "loss": 1.1064,
+      "step": 3949
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1318059427859024e-06,
+      "loss": 1.0938,
+      "step": 3950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1306690967007236e-06,
+      "loss": 1.1201,
+      "step": 3951
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1295323286870497e-06,
+      "loss": 1.1738,
+      "step": 3952
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.128395638985176e-06,
+      "loss": 1.0879,
+      "step": 3953
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1272590278353838e-06,
+      "loss": 1.1328,
+      "step": 3954
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1261224954779373e-06,
+      "loss": 1.1494,
+      "step": 3955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.124986042153085e-06,
+      "loss": 1.1133,
+      "step": 3956
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1238496681010553e-06,
+      "loss": 1.1133,
+      "step": 3957
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1227133735620633e-06,
+      "loss": 1.1631,
+      "step": 3958
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.121577158776306e-06,
+      "loss": 1.0977,
+      "step": 3959
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1204410239839646e-06,
+      "loss": 1.1064,
+      "step": 3960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.119304969425201e-06,
+      "loss": 1.1465,
+      "step": 3961
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1181689953401635e-06,
+      "loss": 1.1182,
+      "step": 3962
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1170331019689798e-06,
+      "loss": 1.1299,
+      "step": 3963
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.115897289551762e-06,
+      "loss": 1.1553,
+      "step": 3964
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1147615583286066e-06,
+      "loss": 1.0923,
+      "step": 3965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.113625908539591e-06,
+      "loss": 1.1084,
+      "step": 3966
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1124903404247766e-06,
+      "loss": 1.1357,
+      "step": 3967
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1113548542242055e-06,
+      "loss": 1.103,
+      "step": 3968
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.110219450177906e-06,
+      "loss": 1.1475,
+      "step": 3969
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1090841285258845e-06,
+      "loss": 1.0898,
+      "step": 3970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.107948889508135e-06,
+      "loss": 1.1133,
+      "step": 3971
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1068137333646304e-06,
+      "loss": 1.1294,
+      "step": 3972
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1056786603353273e-06,
+      "loss": 1.0835,
+      "step": 3973
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1045436706601645e-06,
+      "loss": 1.084,
+      "step": 3974
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.103408764579063e-06,
+      "loss": 1.0718,
+      "step": 3975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.102273942331927e-06,
+      "loss": 1.1152,
+      "step": 3976
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1011392041586425e-06,
+      "loss": 1.1191,
+      "step": 3977
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1000045502990776e-06,
+      "loss": 1.1543,
+      "step": 3978
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.0988699809930816e-06,
+      "loss": 1.1367,
+      "step": 3979
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.097735496480488e-06,
+      "loss": 1.1543,
+      "step": 3980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.096601097001111e-06,
+      "loss": 1.1299,
+      "step": 3981
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.095466782794746e-06,
+      "loss": 1.0928,
+      "step": 3982
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.094332554101174e-06,
+      "loss": 1.0801,
+      "step": 3983
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.093198411160154e-06,
+      "loss": 1.0547,
+      "step": 3984
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0920643542114282e-06,
+      "loss": 1.1572,
+      "step": 3985
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.090930383494721e-06,
+      "loss": 1.1216,
+      "step": 3986
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0897964992497383e-06,
+      "loss": 1.0806,
+      "step": 3987
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0886627017161678e-06,
+      "loss": 1.1387,
+      "step": 3988
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0875289911336785e-06,
+      "loss": 1.1191,
+      "step": 3989
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0863953677419215e-06,
+      "loss": 1.1113,
+      "step": 3990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0852618317805293e-06,
+      "loss": 1.1484,
+      "step": 3991
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0841283834891156e-06,
+      "loss": 1.0601,
+      "step": 3992
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.082995023107277e-06,
+      "loss": 1.1035,
+      "step": 3993
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0818617508745876e-06,
+      "loss": 1.1387,
+      "step": 3994
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0807285670306095e-06,
+      "loss": 1.0991,
+      "step": 3995
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0795954718148797e-06,
+      "loss": 1.1323,
+      "step": 3996
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.07846246546692e-06,
+      "loss": 1.1265,
+      "step": 3997
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.077329548226232e-06,
+      "loss": 1.1299,
+      "step": 3998
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.076196720332299e-06,
+      "loss": 1.0518,
+      "step": 3999
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0750639820245862e-06,
+      "loss": 1.0684,
+      "step": 4000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0739313335425375e-06,
+      "loss": 1.1084,
+      "step": 4001
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.07279877512558e-06,
+      "loss": 1.0811,
+      "step": 4002
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.071666307013121e-06,
+      "loss": 1.1523,
+      "step": 4003
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0705339294445496e-06,
+      "loss": 1.1357,
+      "step": 4004
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0694016426592333e-06,
+      "loss": 1.0879,
+      "step": 4005
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.068269446896523e-06,
+      "loss": 1.0752,
+      "step": 4006
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0671373423957493e-06,
+      "loss": 1.125,
+      "step": 4007
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0660053293962238e-06,
+      "loss": 1.0938,
+      "step": 4008
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0648734081372384e-06,
+      "loss": 1.1592,
+      "step": 4009
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.063741578858066e-06,
+      "loss": 1.0908,
+      "step": 4010
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.06260984179796e-06,
+      "loss": 1.1309,
+      "step": 4011
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0614781971961534e-06,
+      "loss": 1.0815,
+      "step": 4012
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.060346645291861e-06,
+      "loss": 1.0967,
+      "step": 4013
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0592151863242768e-06,
+      "loss": 1.1025,
+      "step": 4014
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0580838205325766e-06,
+      "loss": 1.1431,
+      "step": 4015
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0569525481559152e-06,
+      "loss": 1.0972,
+      "step": 4016
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.055821369433428e-06,
+      "loss": 1.1396,
+      "step": 4017
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0546902846042304e-06,
+      "loss": 1.106,
+      "step": 4018
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0535592939074185e-06,
+      "loss": 1.126,
+      "step": 4019
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0524283975820688e-06,
+      "loss": 1.085,
+      "step": 4020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0512975958672372e-06,
+      "loss": 1.1729,
+      "step": 4021
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0501668890019604e-06,
+      "loss": 1.0957,
+      "step": 4022
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0490362772252523e-06,
+      "loss": 1.1543,
+      "step": 4023
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.04790576077611e-06,
+      "loss": 1.1304,
+      "step": 4024
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0467753398935097e-06,
+      "loss": 1.1348,
+      "step": 4025
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0456450148164067e-06,
+      "loss": 1.1021,
+      "step": 4026
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0445147857837357e-06,
+      "loss": 1.0957,
+      "step": 4027
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0433846530344115e-06,
+      "loss": 1.1436,
+      "step": 4028
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.042254616807329e-06,
+      "loss": 1.1055,
+      "step": 4029
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.041124677341364e-06,
+      "loss": 1.1084,
+      "step": 4030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0399948348753667e-06,
+      "loss": 1.0884,
+      "step": 4031
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0388650896481733e-06,
+      "loss": 1.1406,
+      "step": 4032
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0377354418985966e-06,
+      "loss": 1.1143,
+      "step": 4033
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.036605891865427e-06,
+      "loss": 1.1543,
+      "step": 4034
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.035476439787437e-06,
+      "loss": 1.0859,
+      "step": 4035
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.034347085903377e-06,
+      "loss": 1.1021,
+      "step": 4036
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.033217830451977e-06,
+      "loss": 1.0957,
+      "step": 4037
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.032088673671946e-06,
+      "loss": 1.1172,
+      "step": 4038
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0309596158019728e-06,
+      "loss": 1.0649,
+      "step": 4039
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0298306570807237e-06,
+      "loss": 1.1445,
+      "step": 4040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.028701797746847e-06,
+      "loss": 1.1519,
+      "step": 4041
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.027573038038966e-06,
+      "loss": 1.1094,
+      "step": 4042
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0264443781956862e-06,
+      "loss": 1.1055,
+      "step": 4043
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0253158184555914e-06,
+      "loss": 1.1201,
+      "step": 4044
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0241873590572425e-06,
+      "loss": 1.0791,
+      "step": 4045
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0230590002391814e-06,
+      "loss": 1.0977,
+      "step": 4046
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.021930742239927e-06,
+      "loss": 1.0947,
+      "step": 4047
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0208025852979783e-06,
+      "loss": 1.125,
+      "step": 4048
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0196745296518115e-06,
+      "loss": 1.126,
+      "step": 4049
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.018546575539883e-06,
+      "loss": 1.0957,
+      "step": 4050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0174187232006255e-06,
+      "loss": 1.1016,
+      "step": 4051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0162909728724536e-06,
+      "loss": 1.0591,
+      "step": 4052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0151633247937564e-06,
+      "loss": 1.1377,
+      "step": 4053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.014035779202904e-06,
+      "loss": 1.0859,
+      "step": 4054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0129083363382435e-06,
+      "loss": 1.124,
+      "step": 4055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0117809964381024e-06,
+      "loss": 1.0908,
+      "step": 4056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0106537597407843e-06,
+      "loss": 1.1455,
+      "step": 4057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.009526626484571e-06,
+      "loss": 1.0801,
+      "step": 4058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.008399596907724e-06,
+      "loss": 1.1289,
+      "step": 4059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0072726712484812e-06,
+      "loss": 1.1602,
+      "step": 4060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0061458497450594e-06,
+      "loss": 1.0737,
+      "step": 4061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.005019132635653e-06,
+      "loss": 1.1367,
+      "step": 4062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.003892520158436e-06,
+      "loss": 1.0947,
+      "step": 4063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.002766012551557e-06,
+      "loss": 1.1226,
+      "step": 4064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0016396100531453e-06,
+      "loss": 1.1387,
+      "step": 4065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.000513312901307e-06,
+      "loss": 1.1172,
+      "step": 4066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.999387121334125e-06,
+      "loss": 1.0933,
+      "step": 4067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.998261035589662e-06,
+      "loss": 1.1289,
+      "step": 4068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9971350559059573e-06,
+      "loss": 1.1182,
+      "step": 4069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.996009182521027e-06,
+      "loss": 1.1221,
+      "step": 4070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.994883415672865e-06,
+      "loss": 1.0938,
+      "step": 4071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.993757755599443e-06,
+      "loss": 1.1152,
+      "step": 4072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.992632202538711e-06,
+      "loss": 1.1895,
+      "step": 4073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9915067567285956e-06,
+      "loss": 1.1025,
+      "step": 4074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9903814184069995e-06,
+      "loss": 1.1123,
+      "step": 4075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9892561878118046e-06,
+      "loss": 1.1035,
+      "step": 4076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9881310651808686e-06,
+      "loss": 1.084,
+      "step": 4077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9870060507520284e-06,
+      "loss": 1.1289,
+      "step": 4078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.985881144763095e-06,
+      "loss": 1.1025,
+      "step": 4079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9847563474518584e-06,
+      "loss": 1.0967,
+      "step": 4080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.983631659056087e-06,
+      "loss": 1.104,
+      "step": 4081
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.982507079813523e-06,
+      "loss": 1.1064,
+      "step": 4082
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.981382609961888e-06,
+      "loss": 1.1572,
+      "step": 4083
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.980258249738879e-06,
+      "loss": 1.1289,
+      "step": 4084
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.979133999382171e-06,
+      "loss": 1.0986,
+      "step": 4085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.978009859129414e-06,
+      "loss": 1.1436,
+      "step": 4086
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.976885829218237e-06,
+      "loss": 1.1133,
+      "step": 4087
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9757619098862434e-06,
+      "loss": 1.1162,
+      "step": 4088
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9746381013710163e-06,
+      "loss": 1.123,
+      "step": 4089
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.973514403910111e-06,
+      "loss": 1.1064,
+      "step": 4090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.972390817741063e-06,
+      "loss": 1.0996,
+      "step": 4091
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9712673431013834e-06,
+      "loss": 1.1182,
+      "step": 4092
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.970143980228559e-06,
+      "loss": 1.1133,
+      "step": 4093
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9690207293600536e-06,
+      "loss": 1.0894,
+      "step": 4094
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9678975907333067e-06,
+      "loss": 1.1401,
+      "step": 4095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.966774564585735e-06,
+      "loss": 1.0767,
+      "step": 4096
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9656516511547306e-06,
+      "loss": 1.1387,
+      "step": 4097
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9645288506776618e-06,
+      "loss": 1.1416,
+      "step": 4098
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.963406163391873e-06,
+      "loss": 1.0879,
+      "step": 4099
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.962283589534686e-06,
+      "loss": 1.0967,
+      "step": 4100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9611611293433968e-06,
+      "loss": 1.124,
+      "step": 4101
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.960038783055278e-06,
+      "loss": 1.1719,
+      "step": 4102
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.958916550907578e-06,
+      "loss": 1.0747,
+      "step": 4103
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.957794433137522e-06,
+      "loss": 1.0977,
+      "step": 4104
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.956672429982311e-06,
+      "loss": 1.1533,
+      "step": 4105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9555505416791197e-06,
+      "loss": 1.1377,
+      "step": 4106
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9544287684651016e-06,
+      "loss": 1.1211,
+      "step": 4107
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9533071105773823e-06,
+      "loss": 1.127,
+      "step": 4108
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.952185568253066e-06,
+      "loss": 1.0967,
+      "step": 4109
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9510641417292315e-06,
+      "loss": 1.1562,
+      "step": 4110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9499428312429335e-06,
+      "loss": 1.1768,
+      "step": 4111
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9488216370312004e-06,
+      "loss": 1.105,
+      "step": 4112
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9477005593310374e-06,
+      "loss": 1.1123,
+      "step": 4113
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9465795983794263e-06,
+      "loss": 1.0791,
+      "step": 4114
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.945458754413322e-06,
+      "loss": 1.1387,
+      "step": 4115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9443380276696542e-06,
+      "loss": 1.1348,
+      "step": 4116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9432174183853324e-06,
+      "loss": 1.1279,
+      "step": 4117
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.942096926797237e-06,
+      "loss": 1.1123,
+      "step": 4118
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9409765531422233e-06,
+      "loss": 1.0894,
+      "step": 4119
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9398562976571235e-06,
+      "loss": 1.0669,
+      "step": 4120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9387361605787446e-06,
+      "loss": 1.0972,
+      "step": 4121
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.937616142143869e-06,
+      "loss": 1.1055,
+      "step": 4122
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9364962425892517e-06,
+      "loss": 1.103,
+      "step": 4123
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9353764621516253e-06,
+      "loss": 1.1172,
+      "step": 4124
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9342568010676955e-06,
+      "loss": 1.1562,
+      "step": 4125
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9331372595741443e-06,
+      "loss": 1.1553,
+      "step": 4126
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9320178379076256e-06,
+      "loss": 1.0776,
+      "step": 4127
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9308985363047703e-06,
+      "loss": 1.1719,
+      "step": 4128
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9297793550021857e-06,
+      "loss": 1.1211,
+      "step": 4129
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9286602942364493e-06,
+      "loss": 1.0693,
+      "step": 4130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.927541354244116e-06,
+      "loss": 1.0889,
+      "step": 4131
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9264225352617137e-06,
+      "loss": 1.0977,
+      "step": 4132
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.925303837525747e-06,
+      "loss": 1.1143,
+      "step": 4133
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.924185261272691e-06,
+      "loss": 1.1089,
+      "step": 4134
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9230668067389993e-06,
+      "loss": 1.0679,
+      "step": 4135
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9219484741610972e-06,
+      "loss": 1.082,
+      "step": 4136
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9208302637753855e-06,
+      "loss": 1.167,
+      "step": 4137
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9197121758182375e-06,
+      "loss": 1.123,
+      "step": 4138
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9185942105260025e-06,
+      "loss": 1.1025,
+      "step": 4139
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9174763681350023e-06,
+      "loss": 1.0938,
+      "step": 4140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.916358648881534e-06,
+      "loss": 1.106,
+      "step": 4141
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.915241053001869e-06,
+      "loss": 1.1445,
+      "step": 4142
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9141235807322507e-06,
+      "loss": 1.0801,
+      "step": 4143
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9130062323088984e-06,
+      "loss": 1.1035,
+      "step": 4144
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.911889007968003e-06,
+      "loss": 1.1006,
+      "step": 4145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.910771907945731e-06,
+      "loss": 1.1152,
+      "step": 4146
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9096549324782226e-06,
+      "loss": 1.1211,
+      "step": 4147
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9085380818015913e-06,
+      "loss": 1.126,
+      "step": 4148
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9074213561519227e-06,
+      "loss": 1.123,
+      "step": 4149
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.906304755765278e-06,
+      "loss": 1.1025,
+      "step": 4150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9051882808776918e-06,
+      "loss": 1.126,
+      "step": 4151
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.904071931725171e-06,
+      "loss": 1.1138,
+      "step": 4152
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9029557085436964e-06,
+      "loss": 1.0703,
+      "step": 4153
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.901839611569223e-06,
+      "loss": 1.0928,
+      "step": 4154
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.900723641037679e-06,
+      "loss": 1.0908,
+      "step": 4155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8996077971849636e-06,
+      "loss": 1.0947,
+      "step": 4156
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.898492080246952e-06,
+      "loss": 1.1475,
+      "step": 4157
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8973764904594911e-06,
+      "loss": 1.1377,
+      "step": 4158
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8962610280584021e-06,
+      "loss": 1.1201,
+      "step": 4159
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8951456932794774e-06,
+      "loss": 1.0879,
+      "step": 4160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.894030486358484e-06,
+      "loss": 1.1064,
+      "step": 4161
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8929154075311619e-06,
+      "loss": 1.1621,
+      "step": 4162
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8918004570332229e-06,
+      "loss": 1.1064,
+      "step": 4163
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8906856351003523e-06,
+      "loss": 1.0996,
+      "step": 4164
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8895709419682074e-06,
+      "loss": 1.0645,
+      "step": 4165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8884563778724213e-06,
+      "loss": 1.1504,
+      "step": 4166
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.887341943048596e-06,
+      "loss": 1.1279,
+      "step": 4167
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8862276377323083e-06,
+      "loss": 1.1221,
+      "step": 4168
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.885113462159107e-06,
+      "loss": 1.127,
+      "step": 4169
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8839994165645139e-06,
+      "loss": 1.1045,
+      "step": 4170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8828855011840222e-06,
+      "loss": 1.1123,
+      "step": 4171
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.881771716253099e-06,
+      "loss": 1.1162,
+      "step": 4172
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.880658062007183e-06,
+      "loss": 1.0898,
+      "step": 4173
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8795445386816857e-06,
+      "loss": 1.1094,
+      "step": 4174
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8784311465119903e-06,
+      "loss": 1.1216,
+      "step": 4175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8773178857334528e-06,
+      "loss": 1.1162,
+      "step": 4176
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8762047565814013e-06,
+      "loss": 1.0928,
+      "step": 4177
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8750917592911364e-06,
+      "loss": 1.1016,
+      "step": 4178
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8739788940979304e-06,
+      "loss": 1.1318,
+      "step": 4179
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8728661612370278e-06,
+      "loss": 1.1123,
+      "step": 4180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8717535609436454e-06,
+      "loss": 1.1299,
+      "step": 4181
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8706410934529712e-06,
+      "loss": 1.0996,
+      "step": 4182
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8695287590001655e-06,
+      "loss": 1.1426,
+      "step": 4183
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8684165578203605e-06,
+      "loss": 1.0664,
+      "step": 4184
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8673044901486614e-06,
+      "loss": 1.1455,
+      "step": 4185
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.866192556220143e-06,
+      "loss": 1.1689,
+      "step": 4186
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.865080756269853e-06,
+      "loss": 1.0977,
+      "step": 4187
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8639690905328114e-06,
+      "loss": 1.0947,
+      "step": 4188
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8628575592440078e-06,
+      "loss": 1.082,
+      "step": 4189
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8617461626384067e-06,
+      "loss": 1.1089,
+      "step": 4190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8606349009509408e-06,
+      "loss": 1.1455,
+      "step": 4191
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8595237744165168e-06,
+      "loss": 1.1514,
+      "step": 4192
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.85841278327001e-06,
+      "loss": 1.1118,
+      "step": 4193
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8573019277462696e-06,
+      "loss": 1.0771,
+      "step": 4194
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8561912080801147e-06,
+      "loss": 1.1396,
+      "step": 4195
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.855080624506338e-06,
+      "loss": 1.1279,
+      "step": 4196
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8539701772596995e-06,
+      "loss": 1.166,
+      "step": 4197
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8528598665749338e-06,
+      "loss": 1.1338,
+      "step": 4198
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8517496926867451e-06,
+      "loss": 1.1133,
+      "step": 4199
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.850639655829809e-06,
+      "loss": 1.1162,
+      "step": 4200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8495297562387712e-06,
+      "loss": 1.0972,
+      "step": 4201
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8484199941482512e-06,
+      "loss": 1.1152,
+      "step": 4202
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.847310369792837e-06,
+      "loss": 1.0938,
+      "step": 4203
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8462008834070865e-06,
+      "loss": 1.0957,
+      "step": 4204
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8450915352255317e-06,
+      "loss": 1.0947,
+      "step": 4205
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8439823254826728e-06,
+      "loss": 1.124,
+      "step": 4206
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.842873254412982e-06,
+      "loss": 1.1357,
+      "step": 4207
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8417643222509014e-06,
+      "loss": 1.0762,
+      "step": 4208
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8406555292308437e-06,
+      "loss": 1.1201,
+      "step": 4209
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8395468755871938e-06,
+      "loss": 1.1465,
+      "step": 4210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8384383615543045e-06,
+      "loss": 1.1582,
+      "step": 4211
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8373299873665018e-06,
+      "loss": 1.1348,
+      "step": 4212
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.83622175325808e-06,
+      "loss": 1.126,
+      "step": 4213
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8351136594633045e-06,
+      "loss": 1.1426,
+      "step": 4214
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.834005706216412e-06,
+      "loss": 1.062,
+      "step": 4215
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8328978937516078e-06,
+      "loss": 1.0786,
+      "step": 4216
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8317902223030693e-06,
+      "loss": 1.1069,
+      "step": 4217
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8306826921049432e-06,
+      "loss": 1.0869,
+      "step": 4218
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8295753033913446e-06,
+      "loss": 1.0889,
+      "step": 4219
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8284680563963616e-06,
+      "loss": 1.1484,
+      "step": 4220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8273609513540509e-06,
+      "loss": 1.1211,
+      "step": 4221
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8262539884984392e-06,
+      "loss": 1.0996,
+      "step": 4222
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8251471680635238e-06,
+      "loss": 1.1602,
+      "step": 4223
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8240404902832706e-06,
+      "loss": 1.1436,
+      "step": 4224
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8229339553916168e-06,
+      "loss": 1.1025,
+      "step": 4225
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8218275636224676e-06,
+      "loss": 1.0811,
+      "step": 4226
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.820721315209701e-06,
+      "loss": 1.0605,
+      "step": 4227
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8196152103871612e-06,
+      "loss": 1.1279,
+      "step": 4228
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8185092493886653e-06,
+      "loss": 1.0898,
+      "step": 4229
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8174034324479965e-06,
+      "loss": 1.1079,
+      "step": 4230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8162977597989101e-06,
+      "loss": 1.1338,
+      "step": 4231
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8151922316751302e-06,
+      "loss": 1.1348,
+      "step": 4232
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8140868483103504e-06,
+      "loss": 1.0454,
+      "step": 4233
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8129816099382344e-06,
+      "loss": 1.1094,
+      "step": 4234
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811876516792413e-06,
+      "loss": 1.0947,
+      "step": 4235
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8107715691064883e-06,
+      "loss": 1.0908,
+      "step": 4236
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8096667671140317e-06,
+      "loss": 1.1299,
+      "step": 4237
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8085621110485818e-06,
+      "loss": 1.0938,
+      "step": 4238
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.80745760114365e-06,
+      "loss": 1.0908,
+      "step": 4239
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8063532376327134e-06,
+      "loss": 1.1338,
+      "step": 4240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.805249020749219e-06,
+      "loss": 1.126,
+      "step": 4241
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8041449507265834e-06,
+      "loss": 1.0874,
+      "step": 4242
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8030410277981924e-06,
+      "loss": 1.0615,
+      "step": 4243
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8019372521973994e-06,
+      "loss": 1.0859,
+      "step": 4244
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8008336241575286e-06,
+      "loss": 1.1094,
+      "step": 4245
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7997301439118709e-06,
+      "loss": 1.127,
+      "step": 4246
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7986268116936874e-06,
+      "loss": 1.125,
+      "step": 4247
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7975236277362067e-06,
+      "loss": 1.0444,
+      "step": 4248
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7964205922726284e-06,
+      "loss": 1.0923,
+      "step": 4249
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7953177055361163e-06,
+      "loss": 1.1279,
+      "step": 4250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.794214967759809e-06,
+      "loss": 1.0596,
+      "step": 4251
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.793112379176808e-06,
+      "loss": 1.0918,
+      "step": 4252
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7920099400201859e-06,
+      "loss": 1.1123,
+      "step": 4253
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7909076505229839e-06,
+      "loss": 1.126,
+      "step": 4254
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7898055109182105e-06,
+      "loss": 1.1016,
+      "step": 4255
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7887035214388424e-06,
+      "loss": 1.1562,
+      "step": 4256
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7876016823178254e-06,
+      "loss": 1.0928,
+      "step": 4257
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.786499993788074e-06,
+      "loss": 1.123,
+      "step": 4258
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7853984560824692e-06,
+      "loss": 1.1221,
+      "step": 4259
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7842970694338618e-06,
+      "loss": 1.1748,
+      "step": 4260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7831958340750688e-06,
+      "loss": 1.1162,
+      "step": 4261
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7820947502388763e-06,
+      "loss": 1.0957,
+      "step": 4262
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.78099381815804e-06,
+      "loss": 1.1128,
+      "step": 4263
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.77989303806528e-06,
+      "loss": 1.085,
+      "step": 4264
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.778792410193288e-06,
+      "loss": 1.1387,
+      "step": 4265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7776919347747206e-06,
+      "loss": 1.0942,
+      "step": 4266
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7765916120422031e-06,
+      "loss": 1.1113,
+      "step": 4267
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7754914422283286e-06,
+      "loss": 1.0513,
+      "step": 4268
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7743914255656586e-06,
+      "loss": 1.124,
+      "step": 4269
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7732915622867214e-06,
+      "loss": 1.1201,
+      "step": 4270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.772191852624013e-06,
+      "loss": 1.0713,
+      "step": 4271
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.771092296809997e-06,
+      "loss": 1.0972,
+      "step": 4272
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7699928950771039e-06,
+      "loss": 1.0737,
+      "step": 4273
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7688936476577328e-06,
+      "loss": 1.126,
+      "step": 4274
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7677945547842494e-06,
+      "loss": 1.1367,
+      "step": 4275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7666956166889871e-06,
+      "loss": 1.0884,
+      "step": 4276
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7655968336042468e-06,
+      "loss": 1.1367,
+      "step": 4277
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.764498205762295e-06,
+      "loss": 1.1191,
+      "step": 4278
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7633997333953678e-06,
+      "loss": 1.0879,
+      "step": 4279
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7623014167356664e-06,
+      "loss": 1.0781,
+      "step": 4280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7612032560153602e-06,
+      "loss": 1.1025,
+      "step": 4281
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7601052514665862e-06,
+      "loss": 1.1123,
+      "step": 4282
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7590074033214461e-06,
+      "loss": 1.1113,
+      "step": 4283
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7579097118120108e-06,
+      "loss": 1.123,
+      "step": 4284
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7568121771703172e-06,
+      "loss": 1.1455,
+      "step": 4285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7557147996283695e-06,
+      "loss": 1.1055,
+      "step": 4286
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7546175794181362e-06,
+      "loss": 1.1123,
+      "step": 4287
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.753520516771558e-06,
+      "loss": 1.1357,
+      "step": 4288
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7524236119205367e-06,
+      "loss": 1.1104,
+      "step": 4289
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7513268650969434e-06,
+      "loss": 1.0977,
+      "step": 4290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7502302765326157e-06,
+      "loss": 1.0835,
+      "step": 4291
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7491338464593577e-06,
+      "loss": 1.085,
+      "step": 4292
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7480375751089393e-06,
+      "loss": 1.125,
+      "step": 4293
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7469414627130971e-06,
+      "loss": 1.1157,
+      "step": 4294
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.745845509503534e-06,
+      "loss": 1.0771,
+      "step": 4295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7447497157119202e-06,
+      "loss": 1.1221,
+      "step": 4296
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7436540815698921e-06,
+      "loss": 1.1406,
+      "step": 4297
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.74255860730905e-06,
+      "loss": 1.0864,
+      "step": 4298
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7414632931609633e-06,
+      "loss": 1.1318,
+      "step": 4299
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.740368139357166e-06,
+      "loss": 1.1299,
+      "step": 4300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7392731461291596e-06,
+      "loss": 1.1108,
+      "step": 4301
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7381783137084098e-06,
+      "loss": 1.084,
+      "step": 4302
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7370836423263495e-06,
+      "loss": 1.1206,
+      "step": 4303
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7359891322143778e-06,
+      "loss": 1.0845,
+      "step": 4304
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7348947836038582e-06,
+      "loss": 1.0898,
+      "step": 4305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7338005967261212e-06,
+      "loss": 1.1045,
+      "step": 4306
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.732706571812463e-06,
+      "loss": 1.0889,
+      "step": 4307
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7316127090941466e-06,
+      "loss": 1.1172,
+      "step": 4308
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.730519008802398e-06,
+      "loss": 1.124,
+      "step": 4309
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.729425471168411e-06,
+      "loss": 1.1055,
+      "step": 4310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7283320964233447e-06,
+      "loss": 1.1113,
+      "step": 4311
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7272388847983237e-06,
+      "loss": 1.1006,
+      "step": 4312
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.726145836524438e-06,
+      "loss": 1.0439,
+      "step": 4313
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.725052951832743e-06,
+      "loss": 1.0947,
+      "step": 4314
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7239602309542598e-06,
+      "loss": 1.0947,
+      "step": 4315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.722867674119974e-06,
+      "loss": 1.127,
+      "step": 4316
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7217752815608377e-06,
+      "loss": 1.1367,
+      "step": 4317
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7206830535077674e-06,
+      "loss": 1.0879,
+      "step": 4318
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.719590990191646e-06,
+      "loss": 1.1196,
+      "step": 4319
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7184990918433197e-06,
+      "loss": 1.0996,
+      "step": 4320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.717407358693601e-06,
+      "loss": 1.126,
+      "step": 4321
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.716315790973268e-06,
+      "loss": 1.1001,
+      "step": 4322
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.715224388913062e-06,
+      "loss": 1.127,
+      "step": 4323
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7141331527436922e-06,
+      "loss": 1.1338,
+      "step": 4324
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7130420826958303e-06,
+      "loss": 1.1152,
+      "step": 4325
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7119511790001136e-06,
+      "loss": 1.124,
+      "step": 4326
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7108604418871437e-06,
+      "loss": 1.1094,
+      "step": 4327
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7097698715874877e-06,
+      "loss": 1.1333,
+      "step": 4328
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7086794683316777e-06,
+      "loss": 1.0747,
+      "step": 4329
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7075892323502103e-06,
+      "loss": 1.0908,
+      "step": 4330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7064991638735454e-06,
+      "loss": 1.1533,
+      "step": 4331
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.705409263132109e-06,
+      "loss": 1.1045,
+      "step": 4332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7043195303562915e-06,
+      "loss": 1.1045,
+      "step": 4333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7032299657764478e-06,
+      "loss": 1.1055,
+      "step": 4334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7021405696228952e-06,
+      "loss": 1.082,
+      "step": 4335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7010513421259201e-06,
+      "loss": 1.0923,
+      "step": 4336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6999622835157676e-06,
+      "loss": 1.1118,
+      "step": 4337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6988733940226511e-06,
+      "loss": 1.1333,
+      "step": 4338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.697784673876747e-06,
+      "loss": 1.0977,
+      "step": 4339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6966961233081953e-06,
+      "loss": 1.0742,
+      "step": 4340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6956077425471018e-06,
+      "loss": 1.0801,
+      "step": 4341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6945195318235336e-06,
+      "loss": 1.1162,
+      "step": 4342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6934314913675248e-06,
+      "loss": 1.1011,
+      "step": 4343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6923436214090718e-06,
+      "loss": 1.0488,
+      "step": 4344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6912559221781363e-06,
+      "loss": 1.0977,
+      "step": 4345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6901683939046415e-06,
+      "loss": 1.0977,
+      "step": 4346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6890810368184774e-06,
+      "loss": 1.1162,
+      "step": 4347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6879938511494948e-06,
+      "loss": 1.1289,
+      "step": 4348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6869068371275115e-06,
+      "loss": 1.0933,
+      "step": 4349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6858199949823072e-06,
+      "loss": 1.1465,
+      "step": 4350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6847333249436249e-06,
+      "loss": 1.1416,
+      "step": 4351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6836468272411727e-06,
+      "loss": 1.1016,
+      "step": 4352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6825605021046199e-06,
+      "loss": 1.041,
+      "step": 4353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6814743497636018e-06,
+      "loss": 1.1172,
+      "step": 4354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6803883704477158e-06,
+      "loss": 1.123,
+      "step": 4355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6793025643865242e-06,
+      "loss": 1.1182,
+      "step": 4356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6782169318095492e-06,
+      "loss": 1.1021,
+      "step": 4357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6771314729462807e-06,
+      "loss": 1.1162,
+      "step": 4358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6760461880261688e-06,
+      "loss": 1.0986,
+      "step": 4359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6749610772786279e-06,
+      "loss": 1.0879,
+      "step": 4360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6738761409330367e-06,
+      "loss": 1.1143,
+      "step": 4361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.672791379218735e-06,
+      "loss": 1.0918,
+      "step": 4362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6717067923650275e-06,
+      "loss": 1.1064,
+      "step": 4363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6706223806011795e-06,
+      "loss": 1.1416,
+      "step": 4364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6695381441564217e-06,
+      "loss": 1.0908,
+      "step": 4365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6684540832599472e-06,
+      "loss": 1.1133,
+      "step": 4366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6673701981409114e-06,
+      "loss": 1.126,
+      "step": 4367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.666286489028432e-06,
+      "loss": 1.0908,
+      "step": 4368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6652029561515915e-06,
+      "loss": 1.165,
+      "step": 4369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6641195997394328e-06,
+      "loss": 1.1357,
+      "step": 4370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6630364200209643e-06,
+      "loss": 1.1143,
+      "step": 4371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6619534172251522e-06,
+      "loss": 1.1572,
+      "step": 4372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.660870591580932e-06,
+      "loss": 1.1094,
+      "step": 4373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.659787943317197e-06,
+      "loss": 1.105,
+      "step": 4374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6587054726628038e-06,
+      "loss": 1.1309,
+      "step": 4375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6576231798465719e-06,
+      "loss": 1.1494,
+      "step": 4376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6565410650972838e-06,
+      "loss": 1.1118,
+      "step": 4377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6554591286436839e-06,
+      "loss": 1.125,
+      "step": 4378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6543773707144779e-06,
+      "loss": 1.104,
+      "step": 4379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6532957915383345e-06,
+      "loss": 1.0928,
+      "step": 4380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6522143913438859e-06,
+      "loss": 1.1045,
+      "step": 4381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.651133170359725e-06,
+      "loss": 1.0938,
+      "step": 4382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6500521288144062e-06,
+      "loss": 1.1045,
+      "step": 4383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6489712669364466e-06,
+      "loss": 1.1094,
+      "step": 4384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6478905849543282e-06,
+      "loss": 1.0718,
+      "step": 4385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.64681008309649e-06,
+      "loss": 1.1191,
+      "step": 4386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6457297615913364e-06,
+      "loss": 1.085,
+      "step": 4387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6446496206672325e-06,
+      "loss": 1.1064,
+      "step": 4388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6435696605525053e-06,
+      "loss": 1.0693,
+      "step": 4389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6424898814754427e-06,
+      "loss": 1.1016,
+      "step": 4390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6414102836642964e-06,
+      "loss": 1.1162,
+      "step": 4391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6403308673472778e-06,
+      "loss": 1.1162,
+      "step": 4392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6392516327525617e-06,
+      "loss": 1.1094,
+      "step": 4393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6381725801082827e-06,
+      "loss": 1.0908,
+      "step": 4394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6370937096425377e-06,
+      "loss": 1.1309,
+      "step": 4395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6360150215833853e-06,
+      "loss": 1.1064,
+      "step": 4396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.634936516158846e-06,
+      "loss": 1.1348,
+      "step": 4397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6338581935969005e-06,
+      "loss": 1.1016,
+      "step": 4398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6327800541254923e-06,
+      "loss": 1.1011,
+      "step": 4399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6317020979725245e-06,
+      "loss": 1.1074,
+      "step": 4400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6306243253658627e-06,
+      "loss": 1.1445,
+      "step": 4401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6295467365333323e-06,
+      "loss": 1.1162,
+      "step": 4402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6284693317027222e-06,
+      "loss": 1.0566,
+      "step": 4403
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.627392111101781e-06,
+      "loss": 1.1377,
+      "step": 4404
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.626315074958217e-06,
+      "loss": 1.0776,
+      "step": 4405
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6252382234997022e-06,
+      "loss": 1.0933,
+      "step": 4406
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6241615569538677e-06,
+      "loss": 1.0981,
+      "step": 4407
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6230850755483068e-06,
+      "loss": 1.1416,
+      "step": 4408
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.622008779510571e-06,
+      "loss": 1.1094,
+      "step": 4409
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.620932669068177e-06,
+      "loss": 1.125,
+      "step": 4410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6198567444485996e-06,
+      "loss": 1.1309,
+      "step": 4411
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.618781005879273e-06,
+      "loss": 1.0786,
+      "step": 4412
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6177054535875943e-06,
+      "loss": 1.1045,
+      "step": 4413
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6166300878009212e-06,
+      "loss": 1.1299,
+      "step": 4414
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6155549087465708e-06,
+      "loss": 1.1025,
+      "step": 4415
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.614479916651821e-06,
+      "loss": 1.1396,
+      "step": 4416
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6134051117439104e-06,
+      "loss": 1.0781,
+      "step": 4417
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6123304942500385e-06,
+      "loss": 1.1216,
+      "step": 4418
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6112560643973652e-06,
+      "loss": 1.1348,
+      "step": 4419
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6101818224130092e-06,
+      "loss": 1.1025,
+      "step": 4420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6091077685240497e-06,
+      "loss": 1.0806,
+      "step": 4421
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6080339029575304e-06,
+      "loss": 1.1143,
+      "step": 4422
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6069602259404487e-06,
+      "loss": 1.0903,
+      "step": 4423
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6058867376997668e-06,
+      "loss": 1.1377,
+      "step": 4424
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6048134384624047e-06,
+      "loss": 1.0996,
+      "step": 4425
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.603740328455244e-06,
+      "loss": 1.1084,
+      "step": 4426
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6026674079051243e-06,
+      "loss": 1.1143,
+      "step": 4427
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.601594677038847e-06,
+      "loss": 1.0767,
+      "step": 4428
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6005221360831727e-06,
+      "loss": 1.1221,
+      "step": 4429
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5994497852648223e-06,
+      "loss": 1.0728,
+      "step": 4430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5983776248104754e-06,
+      "loss": 1.1221,
+      "step": 4431
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5973056549467718e-06,
+      "loss": 1.123,
+      "step": 4432
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5962338759003122e-06,
+      "loss": 1.0967,
+      "step": 4433
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5951622878976555e-06,
+      "loss": 1.1309,
+      "step": 4434
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5940908911653211e-06,
+      "loss": 1.1323,
+      "step": 4435
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5930196859297874e-06,
+      "loss": 1.124,
+      "step": 4436
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5919486724174931e-06,
+      "loss": 1.1396,
+      "step": 4437
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5908778508548348e-06,
+      "loss": 1.1455,
+      "step": 4438
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5898072214681698e-06,
+      "loss": 1.0938,
+      "step": 4439
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.588736784483815e-06,
+      "loss": 1.0693,
+      "step": 4440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5876665401280464e-06,
+      "loss": 1.1201,
+      "step": 4441
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5865964886270978e-06,
+      "loss": 1.1338,
+      "step": 4442
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5855266302071645e-06,
+      "loss": 1.084,
+      "step": 4443
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5844569650943994e-06,
+      "loss": 1.1289,
+      "step": 4444
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5833874935149145e-06,
+      "loss": 1.0977,
+      "step": 4445
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5823182156947834e-06,
+      "loss": 1.1104,
+      "step": 4446
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5812491318600354e-06,
+      "loss": 1.0957,
+      "step": 4447
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5801802422366608e-06,
+      "loss": 1.0942,
+      "step": 4448
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5791115470506078e-06,
+      "loss": 1.0986,
+      "step": 4449
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5780430465277837e-06,
+      "loss": 1.1514,
+      "step": 4450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5769747408940555e-06,
+      "loss": 1.1035,
+      "step": 4451
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5759066303752489e-06,
+      "loss": 1.126,
+      "step": 4452
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5748387151971464e-06,
+      "loss": 1.1123,
+      "step": 4453
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5737709955854912e-06,
+      "loss": 1.123,
+      "step": 4454
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5727034717659856e-06,
+      "loss": 1.0684,
+      "step": 4455
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5716361439642886e-06,
+      "loss": 1.064,
+      "step": 4456
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5705690124060177e-06,
+      "loss": 1.1348,
+      "step": 4457
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5695020773167518e-06,
+      "loss": 1.0957,
+      "step": 4458
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5684353389220269e-06,
+      "loss": 1.1021,
+      "step": 4459
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5673687974473344e-06,
+      "loss": 1.0952,
+      "step": 4460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5663024531181287e-06,
+      "loss": 1.0894,
+      "step": 4461
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.565236306159819e-06,
+      "loss": 1.0361,
+      "step": 4462
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.564170356797776e-06,
+      "loss": 1.0894,
+      "step": 4463
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5631046052573246e-06,
+      "loss": 1.0977,
+      "step": 4464
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5620390517637513e-06,
+      "loss": 1.0972,
+      "step": 4465
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5609736965422996e-06,
+      "loss": 1.1172,
+      "step": 4466
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.559908539818171e-06,
+      "loss": 1.0601,
+      "step": 4467
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5588435818165245e-06,
+      "loss": 1.1357,
+      "step": 4468
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.557778822762477e-06,
+      "loss": 1.0879,
+      "step": 4469
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5567142628811067e-06,
+      "loss": 1.085,
+      "step": 4470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5556499023974447e-06,
+      "loss": 1.126,
+      "step": 4471
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.554585741536483e-06,
+      "loss": 1.0552,
+      "step": 4472
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5535217805231706e-06,
+      "loss": 1.0791,
+      "step": 4473
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5524580195824152e-06,
+      "loss": 1.0986,
+      "step": 4474
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5513944589390795e-06,
+      "loss": 1.0962,
+      "step": 4475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5503310988179864e-06,
+      "loss": 1.1445,
+      "step": 4476
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5492679394439162e-06,
+      "loss": 1.1138,
+      "step": 4477
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.548204981041606e-06,
+      "loss": 1.0996,
+      "step": 4478
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5471422238357508e-06,
+      "loss": 1.1289,
+      "step": 4479
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5460796680510024e-06,
+      "loss": 1.1079,
+      "step": 4480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.545017313911971e-06,
+      "loss": 1.1377,
+      "step": 4481
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.543955161643223e-06,
+      "loss": 1.1074,
+      "step": 4482
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5428932114692841e-06,
+      "loss": 1.084,
+      "step": 4483
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5418314636146356e-06,
+      "loss": 1.1211,
+      "step": 4484
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5407699183037167e-06,
+      "loss": 1.1299,
+      "step": 4485
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5397085757609229e-06,
+      "loss": 1.0879,
+      "step": 4486
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.538647436210608e-06,
+      "loss": 1.1089,
+      "step": 4487
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5375864998770818e-06,
+      "loss": 1.0645,
+      "step": 4488
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5365257669846128e-06,
+      "loss": 1.0962,
+      "step": 4489
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5354652377574244e-06,
+      "loss": 1.1045,
+      "step": 4490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5344049124196984e-06,
+      "loss": 1.0967,
+      "step": 4491
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5333447911955729e-06,
+      "loss": 1.1465,
+      "step": 4492
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5322848743091434e-06,
+      "loss": 1.1553,
+      "step": 4493
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5312251619844608e-06,
+      "loss": 1.1021,
+      "step": 4494
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5301656544455356e-06,
+      "loss": 1.125,
+      "step": 4495
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5291063519163325e-06,
+      "loss": 1.1445,
+      "step": 4496
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5280472546207728e-06,
+      "loss": 1.0869,
+      "step": 4497
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5269883627827358e-06,
+      "loss": 1.1758,
+      "step": 4498
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5259296766260564e-06,
+      "loss": 1.1094,
+      "step": 4499
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5248711963745272e-06,
+      "loss": 1.1123,
+      "step": 4500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5238129222518955e-06,
+      "loss": 1.0928,
+      "step": 4501
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5227548544818666e-06,
+      "loss": 1.1631,
+      "step": 4502
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5216969932881009e-06,
+      "loss": 1.1162,
+      "step": 4503
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5206393388942164e-06,
+      "loss": 1.1104,
+      "step": 4504
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5195818915237875e-06,
+      "loss": 1.1123,
+      "step": 4505
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5185246514003413e-06,
+      "loss": 1.1152,
+      "step": 4506
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5174676187473674e-06,
+      "loss": 1.1089,
+      "step": 4507
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5164107937883061e-06,
+      "loss": 1.0928,
+      "step": 4508
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.515354176746556e-06,
+      "loss": 1.1216,
+      "step": 4509
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5142977678454718e-06,
+      "loss": 1.0752,
+      "step": 4510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.513241567308364e-06,
+      "loss": 1.1309,
+      "step": 4511
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5121855753584982e-06,
+      "loss": 1.0713,
+      "step": 4512
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5111297922190972e-06,
+      "loss": 1.1279,
+      "step": 4513
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5100742181133394e-06,
+      "loss": 1.1221,
+      "step": 4514
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5090188532643579e-06,
+      "loss": 1.0986,
+      "step": 4515
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5079636978952439e-06,
+      "loss": 1.124,
+      "step": 4516
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.506908752229041e-06,
+      "loss": 1.084,
+      "step": 4517
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5058540164887506e-06,
+      "loss": 1.1396,
+      "step": 4518
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5047994908973304e-06,
+      "loss": 1.1543,
+      "step": 4519
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5037451756776924e-06,
+      "loss": 1.124,
+      "step": 4520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5026910710527042e-06,
+      "loss": 1.1221,
+      "step": 4521
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5016371772451896e-06,
+      "loss": 1.1396,
+      "step": 4522
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5005834944779266e-06,
+      "loss": 1.1016,
+      "step": 4523
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4995300229736492e-06,
+      "loss": 1.1426,
+      "step": 4524
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4984767629550478e-06,
+      "loss": 1.0869,
+      "step": 4525
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4974237146447665e-06,
+      "loss": 1.1074,
+      "step": 4526
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4963708782654057e-06,
+      "loss": 1.1035,
+      "step": 4527
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4953182540395203e-06,
+      "loss": 1.127,
+      "step": 4528
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4942658421896206e-06,
+      "loss": 1.0908,
+      "step": 4529
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4932136429381719e-06,
+      "loss": 1.0498,
+      "step": 4530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4921616565075956e-06,
+      "loss": 1.0962,
+      "step": 4531
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4911098831202664e-06,
+      "loss": 1.1074,
+      "step": 4532
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.490058322998516e-06,
+      "loss": 1.0864,
+      "step": 4533
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4890069763646284e-06,
+      "loss": 1.1064,
+      "step": 4534
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4879558434408444e-06,
+      "loss": 1.1406,
+      "step": 4535
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4869049244493594e-06,
+      "loss": 1.1377,
+      "step": 4536
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4858542196123236e-06,
+      "loss": 1.0908,
+      "step": 4537
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4848037291518404e-06,
+      "loss": 1.1191,
+      "step": 4538
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4837534532899702e-06,
+      "loss": 1.1123,
+      "step": 4539
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4827033922487266e-06,
+      "loss": 1.0781,
+      "step": 4540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.481653546250078e-06,
+      "loss": 1.0728,
+      "step": 4541
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4806039155159485e-06,
+      "loss": 1.1455,
+      "step": 4542
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4795545002682133e-06,
+      "loss": 1.1074,
+      "step": 4543
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4785053007287076e-06,
+      "loss": 1.1108,
+      "step": 4544
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4774563171192153e-06,
+      "loss": 1.1279,
+      "step": 4545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4764075496614784e-06,
+      "loss": 1.1523,
+      "step": 4546
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4753589985771922e-06,
+      "loss": 1.1436,
+      "step": 4547
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4743106640880056e-06,
+      "loss": 1.1089,
+      "step": 4548
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4732625464155218e-06,
+      "loss": 1.1367,
+      "step": 4549
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4722146457812989e-06,
+      "loss": 1.123,
+      "step": 4550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4711669624068486e-06,
+      "loss": 1.104,
+      "step": 4551
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4701194965136375e-06,
+      "loss": 1.124,
+      "step": 4552
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4690722483230856e-06,
+      "loss": 1.1167,
+      "step": 4553
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4680252180565657e-06,
+      "loss": 1.1094,
+      "step": 4554
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4669784059354058e-06,
+      "loss": 1.0957,
+      "step": 4555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4659318121808888e-06,
+      "loss": 1.0684,
+      "step": 4556
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4648854370142496e-06,
+      "loss": 1.0459,
+      "step": 4557
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4638392806566782e-06,
+      "loss": 1.1001,
+      "step": 4558
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4627933433293174e-06,
+      "loss": 1.1133,
+      "step": 4559
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4617476252532637e-06,
+      "loss": 1.1182,
+      "step": 4560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4607021266495675e-06,
+      "loss": 1.124,
+      "step": 4561
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4596568477392334e-06,
+      "loss": 1.1475,
+      "step": 4562
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4586117887432188e-06,
+      "loss": 1.083,
+      "step": 4563
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4575669498824358e-06,
+      "loss": 1.1172,
+      "step": 4564
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4565223313777472e-06,
+      "loss": 1.1113,
+      "step": 4565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.455477933449973e-06,
+      "loss": 1.1211,
+      "step": 4566
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4544337563198829e-06,
+      "loss": 1.0796,
+      "step": 4567
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4533898002082036e-06,
+      "loss": 1.0547,
+      "step": 4568
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4523460653356114e-06,
+      "loss": 1.0566,
+      "step": 4569
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4513025519227387e-06,
+      "loss": 1.1191,
+      "step": 4570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4502592601901688e-06,
+      "loss": 1.0928,
+      "step": 4571
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4492161903584412e-06,
+      "loss": 1.1035,
+      "step": 4572
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4481733426480455e-06,
+      "loss": 1.0674,
+      "step": 4573
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4471307172794241e-06,
+      "loss": 1.1309,
+      "step": 4574
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4460883144729764e-06,
+      "loss": 1.0703,
+      "step": 4575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4450461344490496e-06,
+      "loss": 1.1045,
+      "step": 4576
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4440041774279484e-06,
+      "loss": 1.125,
+      "step": 4577
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4429624436299274e-06,
+      "loss": 1.0962,
+      "step": 4578
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4419209332751937e-06,
+      "loss": 1.085,
+      "step": 4579
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4408796465839097e-06,
+      "loss": 1.1807,
+      "step": 4580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4398385837761901e-06,
+      "loss": 1.063,
+      "step": 4581
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.438797745072099e-06,
+      "loss": 1.1641,
+      "step": 4582
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4377571306916576e-06,
+      "loss": 1.1084,
+      "step": 4583
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4367167408548366e-06,
+      "loss": 1.0894,
+      "step": 4584
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4356765757815593e-06,
+      "loss": 1.0591,
+      "step": 4585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4346366356917047e-06,
+      "loss": 1.0947,
+      "step": 4586
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4335969208050986e-06,
+      "loss": 1.0977,
+      "step": 4587
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4325574313415263e-06,
+      "loss": 1.1152,
+      "step": 4588
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4315181675207193e-06,
+      "loss": 1.1357,
+      "step": 4589
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4304791295623633e-06,
+      "loss": 1.0908,
+      "step": 4590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4294403176860978e-06,
+      "loss": 1.0869,
+      "step": 4591
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.428401732111514e-06,
+      "loss": 1.1006,
+      "step": 4592
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4273633730581532e-06,
+      "loss": 1.1035,
+      "step": 4593
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4263252407455125e-06,
+      "loss": 1.1504,
+      "step": 4594
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4252873353930369e-06,
+      "loss": 1.1172,
+      "step": 4595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4242496572201251e-06,
+      "loss": 1.1172,
+      "step": 4596
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4232122064461301e-06,
+      "loss": 1.1055,
+      "step": 4597
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4221749832903526e-06,
+      "loss": 1.1084,
+      "step": 4598
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4211379879720494e-06,
+      "loss": 1.1191,
+      "step": 4599
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.420101220710426e-06,
+      "loss": 1.085,
+      "step": 4600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.41906468172464e-06,
+      "loss": 1.1084,
+      "step": 4601
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4180283712338039e-06,
+      "loss": 1.1094,
+      "step": 4602
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4169922894569773e-06,
+      "loss": 1.125,
+      "step": 4603
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4159564366131746e-06,
+      "loss": 1.0977,
+      "step": 4604
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.414920812921362e-06,
+      "loss": 1.1191,
+      "step": 4605
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4138854186004553e-06,
+      "loss": 1.1089,
+      "step": 4606
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4128502538693213e-06,
+      "loss": 1.0806,
+      "step": 4607
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.411815318946782e-06,
+      "loss": 1.0889,
+      "step": 4608
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4107806140516066e-06,
+      "loss": 1.127,
+      "step": 4609
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4097461394025198e-06,
+      "loss": 1.1089,
+      "step": 4610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4087118952181934e-06,
+      "loss": 1.1538,
+      "step": 4611
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4076778817172523e-06,
+      "loss": 1.0762,
+      "step": 4612
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.406644099118275e-06,
+      "loss": 1.1201,
+      "step": 4613
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4056105476397875e-06,
+      "loss": 1.123,
+      "step": 4614
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4045772275002673e-06,
+      "loss": 1.1152,
+      "step": 4615
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4035441389181456e-06,
+      "loss": 1.1426,
+      "step": 4616
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4025112821118043e-06,
+      "loss": 1.1396,
+      "step": 4617
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4014786572995725e-06,
+      "loss": 1.1348,
+      "step": 4618
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4004462646997357e-06,
+      "loss": 1.0713,
+      "step": 4619
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3994141045305253e-06,
+      "loss": 1.0996,
+      "step": 4620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.398382177010128e-06,
+      "loss": 1.124,
+      "step": 4621
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3973504823566778e-06,
+      "loss": 1.1055,
+      "step": 4622
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.39631902078826e-06,
+      "loss": 1.1108,
+      "step": 4623
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3952877925229137e-06,
+      "loss": 1.082,
+      "step": 4624
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3942567977786253e-06,
+      "loss": 1.0938,
+      "step": 4625
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3932260367733318e-06,
+      "loss": 1.083,
+      "step": 4626
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3921955097249246e-06,
+      "loss": 1.1182,
+      "step": 4627
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3911652168512405e-06,
+      "loss": 1.0894,
+      "step": 4628
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3901351583700703e-06,
+      "loss": 1.0679,
+      "step": 4629
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3891053344991557e-06,
+      "loss": 1.1304,
+      "step": 4630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3880757454561847e-06,
+      "loss": 1.1172,
+      "step": 4631
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.387046391458801e-06,
+      "loss": 1.1094,
+      "step": 4632
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3860172727245947e-06,
+      "loss": 1.0879,
+      "step": 4633
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3849883894711064e-06,
+      "loss": 1.0947,
+      "step": 4634
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3839597419158298e-06,
+      "loss": 1.1406,
+      "step": 4635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3829313302762065e-06,
+      "loss": 1.1011,
+      "step": 4636
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3819031547696271e-06,
+      "loss": 1.1143,
+      "step": 4637
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.380875215613436e-06,
+      "loss": 1.1226,
+      "step": 4638
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.379847513024924e-06,
+      "loss": 1.1006,
+      "step": 4639
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3788200472213331e-06,
+      "loss": 1.1133,
+      "step": 4640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.377792818419858e-06,
+      "loss": 1.1348,
+      "step": 4641
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3767658268376377e-06,
+      "loss": 1.1001,
+      "step": 4642
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3757390726917668e-06,
+      "loss": 1.1064,
+      "step": 4643
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3747125561992863e-06,
+      "loss": 1.1172,
+      "step": 4644
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3736862775771858e-06,
+      "loss": 1.0977,
+      "step": 4645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3726602370424093e-06,
+      "loss": 1.1133,
+      "step": 4646
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3716344348118469e-06,
+      "loss": 1.1006,
+      "step": 4647
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3706088711023374e-06,
+      "loss": 1.0718,
+      "step": 4648
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.369583546130674e-06,
+      "loss": 1.1465,
+      "step": 4649
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3685584601135933e-06,
+      "loss": 1.1191,
+      "step": 4650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3675336132677868e-06,
+      "loss": 1.0947,
+      "step": 4651
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3665090058098917e-06,
+      "loss": 1.1177,
+      "step": 4652
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3654846379564965e-06,
+      "loss": 1.085,
+      "step": 4653
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3644605099241394e-06,
+      "loss": 1.1465,
+      "step": 4654
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3634366219293066e-06,
+      "loss": 1.1006,
+      "step": 4655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3624129741884326e-06,
+      "loss": 1.084,
+      "step": 4656
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3613895669179045e-06,
+      "loss": 1.1025,
+      "step": 4657
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.360366400334056e-06,
+      "loss": 1.0908,
+      "step": 4658
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3593434746531691e-06,
+      "loss": 1.0947,
+      "step": 4659
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3583207900914787e-06,
+      "loss": 1.0781,
+      "step": 4660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.357298346865164e-06,
+      "loss": 1.0786,
+      "step": 4661
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3562761451903577e-06,
+      "loss": 1.1074,
+      "step": 4662
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3552541852831382e-06,
+      "loss": 1.0762,
+      "step": 4663
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3542324673595331e-06,
+      "loss": 1.0718,
+      "step": 4664
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3532109916355201e-06,
+      "loss": 1.1709,
+      "step": 4665
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3521897583270266e-06,
+      "loss": 1.1318,
+      "step": 4666
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3511687676499252e-06,
+      "loss": 1.0674,
+      "step": 4667
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3501480198200412e-06,
+      "loss": 1.1201,
+      "step": 4668
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3491275150531463e-06,
+      "loss": 1.124,
+      "step": 4669
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3481072535649598e-06,
+      "loss": 1.1133,
+      "step": 4670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3470872355711528e-06,
+      "loss": 1.1187,
+      "step": 4671
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3460674612873417e-06,
+      "loss": 1.1104,
+      "step": 4672
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3450479309290948e-06,
+      "loss": 1.0996,
+      "step": 4673
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3440286447119256e-06,
+      "loss": 1.0801,
+      "step": 4674
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3430096028512963e-06,
+      "loss": 1.1196,
+      "step": 4675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3419908055626194e-06,
+      "loss": 1.1338,
+      "step": 4676
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.340972253061256e-06,
+      "loss": 1.0942,
+      "step": 4677
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3399539455625118e-06,
+      "loss": 1.0967,
+      "step": 4678
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3389358832816455e-06,
+      "loss": 1.0991,
+      "step": 4679
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3379180664338603e-06,
+      "loss": 1.0947,
+      "step": 4680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3369004952343076e-06,
+      "loss": 1.123,
+      "step": 4681
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3358831698980904e-06,
+      "loss": 1.0591,
+      "step": 4682
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.334866090640255e-06,
+      "loss": 1.1279,
+      "step": 4683
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3338492576758005e-06,
+      "loss": 1.0889,
+      "step": 4684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.33283267121967e-06,
+      "loss": 1.1182,
+      "step": 4685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3318163314867555e-06,
+      "loss": 1.0737,
+      "step": 4686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.330800238691899e-06,
+      "loss": 1.1289,
+      "step": 4687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.329784393049887e-06,
+      "loss": 1.1211,
+      "step": 4688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3287687947754548e-06,
+      "loss": 1.1338,
+      "step": 4689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.327753444083289e-06,
+      "loss": 1.1099,
+      "step": 4690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.326738341188019e-06,
+      "loss": 1.1328,
+      "step": 4691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3257234863042228e-06,
+      "loss": 1.1074,
+      "step": 4692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3247088796464292e-06,
+      "loss": 1.1104,
+      "step": 4693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.323694521429109e-06,
+      "loss": 1.1064,
+      "step": 4694
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3226804118666871e-06,
+      "loss": 1.1152,
+      "step": 4695
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3216665511735305e-06,
+      "loss": 1.1191,
+      "step": 4696
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.320652939563955e-06,
+      "loss": 1.1035,
+      "step": 4697
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3196395772522258e-06,
+      "loss": 1.1465,
+      "step": 4698
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3186264644525528e-06,
+      "loss": 1.0977,
+      "step": 4699
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3176136013790936e-06,
+      "loss": 1.0811,
+      "step": 4700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3166009882459547e-06,
+      "loss": 1.0947,
+      "step": 4701
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.315588625267189e-06,
+      "loss": 1.1387,
+      "step": 4702
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3145765126567944e-06,
+      "loss": 1.0977,
+      "step": 4703
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3135646506287203e-06,
+      "loss": 1.127,
+      "step": 4704
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3125530393968577e-06,
+      "loss": 1.1367,
+      "step": 4705
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.31154167917505e-06,
+      "loss": 1.0972,
+      "step": 4706
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3105305701770834e-06,
+      "loss": 1.1455,
+      "step": 4707
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.309519712616692e-06,
+      "loss": 1.1201,
+      "step": 4708
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3085091067075589e-06,
+      "loss": 1.1279,
+      "step": 4709
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3074987526633114e-06,
+      "loss": 1.0801,
+      "step": 4710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3064886506975238e-06,
+      "loss": 1.1748,
+      "step": 4711
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3054788010237202e-06,
+      "loss": 1.1279,
+      "step": 4712
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3044692038553658e-06,
+      "loss": 1.1055,
+      "step": 4713
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3034598594058775e-06,
+      "loss": 1.106,
+      "step": 4714
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3024507678886179e-06,
+      "loss": 1.0742,
+      "step": 4715
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3014419295168929e-06,
+      "loss": 1.1064,
+      "step": 4716
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3004333445039597e-06,
+      "loss": 1.1025,
+      "step": 4717
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2994250130630173e-06,
+      "loss": 1.126,
+      "step": 4718
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2984169354072133e-06,
+      "loss": 1.1455,
+      "step": 4719
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2974091117496426e-06,
+      "loss": 1.1016,
+      "step": 4720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.296401542303345e-06,
+      "loss": 1.1294,
+      "step": 4721
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2953942272813058e-06,
+      "loss": 1.123,
+      "step": 4722
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2943871668964597e-06,
+      "loss": 1.1104,
+      "step": 4723
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.293380361361683e-06,
+      "loss": 1.0791,
+      "step": 4724
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2923738108898022e-06,
+      "loss": 1.1387,
+      "step": 4725
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.291367515693589e-06,
+      "loss": 1.1182,
+      "step": 4726
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2903614759857584e-06,
+      "loss": 1.124,
+      "step": 4727
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2893556919789757e-06,
+      "loss": 1.1396,
+      "step": 4728
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2883501638858486e-06,
+      "loss": 1.0898,
+      "step": 4729
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.287344891918931e-06,
+      "loss": 1.1182,
+      "step": 4730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.286339876290726e-06,
+      "loss": 1.1094,
+      "step": 4731
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2853351172136791e-06,
+      "loss": 1.1133,
+      "step": 4732
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.284330614900181e-06,
+      "loss": 1.0708,
+      "step": 4733
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2833263695625725e-06,
+      "loss": 1.1152,
+      "step": 4734
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2823223814131352e-06,
+      "loss": 1.1123,
+      "step": 4735
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2813186506641e-06,
+      "loss": 1.1318,
+      "step": 4736
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2803151775276407e-06,
+      "loss": 1.0991,
+      "step": 4737
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2793119622158782e-06,
+      "loss": 1.0801,
+      "step": 4738
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2783090049408793e-06,
+      "loss": 1.1357,
+      "step": 4739
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2773063059146551e-06,
+      "loss": 1.1021,
+      "step": 4740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2763038653491609e-06,
+      "loss": 1.1123,
+      "step": 4741
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2753016834563008e-06,
+      "loss": 1.1377,
+      "step": 4742
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2742997604479225e-06,
+      "loss": 1.1562,
+      "step": 4743
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2732980965358166e-06,
+      "loss": 1.1475,
+      "step": 4744
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2722966919317234e-06,
+      "loss": 1.0596,
+      "step": 4745
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2712955468473245e-06,
+      "loss": 1.0874,
+      "step": 4746
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2702946614942502e-06,
+      "loss": 1.0815,
+      "step": 4747
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2692940360840727e-06,
+      "loss": 1.1396,
+      "step": 4748
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.26829367082831e-06,
+      "loss": 1.1182,
+      "step": 4749
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2672935659384263e-06,
+      "loss": 1.0874,
+      "step": 4750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2662937216258312e-06,
+      "loss": 1.1172,
+      "step": 4751
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.265294138101876e-06,
+      "loss": 1.1064,
+      "step": 4752
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2642948155778612e-06,
+      "loss": 1.062,
+      "step": 4753
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2632957542650287e-06,
+      "loss": 1.0864,
+      "step": 4754
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2622969543745655e-06,
+      "loss": 1.084,
+      "step": 4755
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2612984161176062e-06,
+      "loss": 1.0752,
+      "step": 4756
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.260300139705226e-06,
+      "loss": 1.1221,
+      "step": 4757
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2593021253484494e-06,
+      "loss": 1.0889,
+      "step": 4758
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2583043732582417e-06,
+      "loss": 1.0806,
+      "step": 4759
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2573068836455127e-06,
+      "loss": 1.1309,
+      "step": 4760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2563096567211209e-06,
+      "loss": 1.1025,
+      "step": 4761
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2553126926958631e-06,
+      "loss": 1.0889,
+      "step": 4762
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2543159917804865e-06,
+      "loss": 1.1035,
+      "step": 4763
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2533195541856797e-06,
+      "loss": 1.0859,
+      "step": 4764
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2523233801220758e-06,
+      "loss": 1.1328,
+      "step": 4765
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2513274698002509e-06,
+      "loss": 1.1357,
+      "step": 4766
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2503318234307288e-06,
+      "loss": 1.1318,
+      "step": 4767
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2493364412239741e-06,
+      "loss": 1.1152,
+      "step": 4768
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2483413233903988e-06,
+      "loss": 1.126,
+      "step": 4769
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2473464701403556e-06,
+      "loss": 1.0879,
+      "step": 4770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2463518816841427e-06,
+      "loss": 1.126,
+      "step": 4771
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2453575582320038e-06,
+      "loss": 1.1338,
+      "step": 4772
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2443634999941238e-06,
+      "loss": 1.1318,
+      "step": 4773
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.243369707180634e-06,
+      "loss": 1.1299,
+      "step": 4774
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.242376180001609e-06,
+      "loss": 1.123,
+      "step": 4775
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2413829186670664e-06,
+      "loss": 1.125,
+      "step": 4776
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2403899233869674e-06,
+      "loss": 1.1123,
+      "step": 4777
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2393971943712189e-06,
+      "loss": 1.1143,
+      "step": 4778
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2384047318296685e-06,
+      "loss": 1.1455,
+      "step": 4779
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2374125359721112e-06,
+      "loss": 1.0879,
+      "step": 4780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.236420607008283e-06,
+      "loss": 1.1309,
+      "step": 4781
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2354289451478624e-06,
+      "loss": 1.1572,
+      "step": 4782
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2344375506004757e-06,
+      "loss": 1.062,
+      "step": 4783
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2334464235756894e-06,
+      "loss": 1.1455,
+      "step": 4784
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2324555642830124e-06,
+      "loss": 1.1191,
+      "step": 4785
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2314649729319005e-06,
+      "loss": 1.1016,
+      "step": 4786
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.230474649731752e-06,
+      "loss": 1.1309,
+      "step": 4787
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2294845948919052e-06,
+      "loss": 1.1055,
+      "step": 4788
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2284948086216466e-06,
+      "loss": 1.0767,
+      "step": 4789
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2275052911302018e-06,
+      "loss": 1.1846,
+      "step": 4790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2265160426267427e-06,
+      "loss": 1.1445,
+      "step": 4791
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.225527063320382e-06,
+      "loss": 1.1221,
+      "step": 4792
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2245383534201758e-06,
+      "loss": 1.0942,
+      "step": 4793
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2235499131351252e-06,
+      "loss": 1.1094,
+      "step": 4794
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2225617426741729e-06,
+      "loss": 1.0869,
+      "step": 4795
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2215738422462025e-06,
+      "loss": 1.1309,
+      "step": 4796
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2205862120600457e-06,
+      "loss": 1.1172,
+      "step": 4797
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2195988523244712e-06,
+      "loss": 1.1143,
+      "step": 4798
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2186117632481947e-06,
+      "loss": 1.0977,
+      "step": 4799
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2176249450398745e-06,
+      "loss": 1.0962,
+      "step": 4800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.216638397908108e-06,
+      "loss": 1.0718,
+      "step": 4801
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2156521220614404e-06,
+      "loss": 1.125,
+      "step": 4802
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2146661177083554e-06,
+      "loss": 1.0835,
+      "step": 4803
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2136803850572804e-06,
+      "loss": 1.0791,
+      "step": 4804
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2126949243165869e-06,
+      "loss": 1.1016,
+      "step": 4805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2117097356945878e-06,
+      "loss": 1.0854,
+      "step": 4806
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2107248193995368e-06,
+      "loss": 1.0522,
+      "step": 4807
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2097401756396344e-06,
+      "loss": 1.1221,
+      "step": 4808
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.208755804623018e-06,
+      "loss": 1.0928,
+      "step": 4809
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2077717065577718e-06,
+      "loss": 1.064,
+      "step": 4810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2067878816519217e-06,
+      "loss": 1.083,
+      "step": 4811
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2058043301134325e-06,
+      "loss": 1.0669,
+      "step": 4812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2048210521502158e-06,
+      "loss": 1.0801,
+      "step": 4813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2038380479701217e-06,
+      "loss": 1.1104,
+      "step": 4814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2028553177809436e-06,
+      "loss": 1.0981,
+      "step": 4815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2018728617904183e-06,
+      "loss": 1.1289,
+      "step": 4816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2008906802062232e-06,
+      "loss": 1.1133,
+      "step": 4817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1999087732359771e-06,
+      "loss": 1.1201,
+      "step": 4818
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1989271410872433e-06,
+      "loss": 1.1484,
+      "step": 4819
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1979457839675238e-06,
+      "loss": 1.0967,
+      "step": 4820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1969647020842658e-06,
+      "loss": 1.0913,
+      "step": 4821
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1959838956448557e-06,
+      "loss": 1.0723,
+      "step": 4822
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1950033648566206e-06,
+      "loss": 1.1064,
+      "step": 4823
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1940231099268353e-06,
+      "loss": 1.127,
+      "step": 4824
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1930431310627102e-06,
+      "loss": 1.0439,
+      "step": 4825
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.192063428471399e-06,
+      "loss": 1.0649,
+      "step": 4826
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1910840023599984e-06,
+      "loss": 1.0898,
+      "step": 4827
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1901048529355452e-06,
+      "loss": 1.1016,
+      "step": 4828
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1891259804050177e-06,
+      "loss": 1.1069,
+      "step": 4829
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1881473849753373e-06,
+      "loss": 1.1265,
+      "step": 4830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1871690668533642e-06,
+      "loss": 1.1152,
+      "step": 4831
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1861910262459034e-06,
+      "loss": 1.1094,
+      "step": 4832
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.185213263359698e-06,
+      "loss": 1.0791,
+      "step": 4833
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1842357784014335e-06,
+      "loss": 1.1401,
+      "step": 4834
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.183258571577737e-06,
+      "loss": 1.0811,
+      "step": 4835
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1822816430951784e-06,
+      "loss": 1.0649,
+      "step": 4836
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1813049931602644e-06,
+      "loss": 1.1089,
+      "step": 4837
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1803286219794474e-06,
+      "loss": 1.0933,
+      "step": 4838
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1793525297591185e-06,
+      "loss": 1.1152,
+      "step": 4839
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.178376716705609e-06,
+      "loss": 1.1533,
+      "step": 4840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1774011830251945e-06,
+      "loss": 1.1318,
+      "step": 4841
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.176425928924087e-06,
+      "loss": 1.043,
+      "step": 4842
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1754509546084445e-06,
+      "loss": 1.1089,
+      "step": 4843
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.174476260284362e-06,
+      "loss": 1.1152,
+      "step": 4844
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1735018461578757e-06,
+      "loss": 1.1118,
+      "step": 4845
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.172527712434965e-06,
+      "loss": 1.1582,
+      "step": 4846
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.171553859321547e-06,
+      "loss": 1.1206,
+      "step": 4847
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1705802870234815e-06,
+      "loss": 1.1035,
+      "step": 4848
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1696069957465694e-06,
+      "loss": 1.1182,
+      "step": 4849
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1686339856965506e-06,
+      "loss": 1.1191,
+      "step": 4850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1676612570791046e-06,
+      "loss": 1.1182,
+      "step": 4851
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1666888100998555e-06,
+      "loss": 1.1631,
+      "step": 4852
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1657166449643628e-06,
+      "loss": 1.0933,
+      "step": 4853
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.164744761878131e-06,
+      "loss": 1.1309,
+      "step": 4854
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1637731610466018e-06,
+      "loss": 1.0669,
+      "step": 4855
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1628018426751577e-06,
+      "loss": 1.1543,
+      "step": 4856
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1618308069691243e-06,
+      "loss": 1.0513,
+      "step": 4857
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1608600541337625e-06,
+      "loss": 1.0928,
+      "step": 4858
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1598895843742778e-06,
+      "loss": 1.0537,
+      "step": 4859
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1589193978958148e-06,
+      "loss": 1.1104,
+      "step": 4860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.157949494903457e-06,
+      "loss": 1.0942,
+      "step": 4861
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1569798756022278e-06,
+      "loss": 1.0825,
+      "step": 4862
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1560105401970924e-06,
+      "loss": 1.0645,
+      "step": 4863
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1550414888929545e-06,
+      "loss": 1.1582,
+      "step": 4864
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1540727218946596e-06,
+      "loss": 1.1167,
+      "step": 4865
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1531042394069907e-06,
+      "loss": 1.1221,
+      "step": 4866
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1521360416346707e-06,
+      "loss": 1.1309,
+      "step": 4867
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1511681287823659e-06,
+      "loss": 1.1436,
+      "step": 4868
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1502005010546775e-06,
+      "loss": 1.1211,
+      "step": 4869
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1492331586561506e-06,
+      "loss": 1.1143,
+      "step": 4870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1482661017912665e-06,
+      "loss": 1.1426,
+      "step": 4871
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1472993306644494e-06,
+      "loss": 1.127,
+      "step": 4872
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.14633284548006e-06,
+      "loss": 1.1011,
+      "step": 4873
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1453666464424015e-06,
+      "loss": 1.04,
+      "step": 4874
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.144400733755714e-06,
+      "loss": 1.1113,
+      "step": 4875
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1434351076241793e-06,
+      "loss": 1.1357,
+      "step": 4876
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1424697682519168e-06,
+      "loss": 1.1016,
+      "step": 4877
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1415047158429855e-06,
+      "loss": 1.1025,
+      "step": 4878
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.140539950601386e-06,
+      "loss": 1.106,
+      "step": 4879
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1395754727310543e-06,
+      "loss": 1.1226,
+      "step": 4880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.13861128243587e-06,
+      "loss": 1.1182,
+      "step": 4881
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1376473799196488e-06,
+      "loss": 1.0918,
+      "step": 4882
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1366837653861457e-06,
+      "loss": 1.085,
+      "step": 4883
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1357204390390562e-06,
+      "loss": 1.1035,
+      "step": 4884
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1347574010820155e-06,
+      "loss": 1.0952,
+      "step": 4885
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1337946517185948e-06,
+      "loss": 1.0664,
+      "step": 4886
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.132832191152308e-06,
+      "loss": 1.1211,
+      "step": 4887
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.131870019586605e-06,
+      "loss": 1.1465,
+      "step": 4888
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.130908137224875e-06,
+      "loss": 1.1484,
+      "step": 4889
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1299465442704488e-06,
+      "loss": 1.1182,
+      "step": 4890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1289852409265917e-06,
+      "loss": 1.0977,
+      "step": 4891
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1280242273965122e-06,
+      "loss": 1.1045,
+      "step": 4892
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1270635038833546e-06,
+      "loss": 1.1455,
+      "step": 4893
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1261030705902016e-06,
+      "loss": 1.1055,
+      "step": 4894
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1251429277200773e-06,
+      "loss": 1.1021,
+      "step": 4895
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1241830754759413e-06,
+      "loss": 1.124,
+      "step": 4896
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1232235140606942e-06,
+      "loss": 1.124,
+      "step": 4897
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.122264243677175e-06,
+      "loss": 1.0908,
+      "step": 4898
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1213052645281591e-06,
+      "loss": 1.1035,
+      "step": 4899
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1203465768163613e-06,
+      "loss": 1.1006,
+      "step": 4900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1193881807444362e-06,
+      "loss": 1.0986,
+      "step": 4901
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1184300765149744e-06,
+      "loss": 1.1152,
+      "step": 4902
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1174722643305075e-06,
+      "loss": 1.1475,
+      "step": 4903
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1165147443935029e-06,
+      "loss": 1.1572,
+      "step": 4904
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1155575169063668e-06,
+      "loss": 1.1172,
+      "step": 4905
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1146005820714454e-06,
+      "loss": 1.0679,
+      "step": 4906
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1136439400910206e-06,
+      "loss": 1.1475,
+      "step": 4907
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1126875911673124e-06,
+      "loss": 1.083,
+      "step": 4908
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1117315355024826e-06,
+      "loss": 1.1104,
+      "step": 4909
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.110775773298627e-06,
+      "loss": 1.1279,
+      "step": 4910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1098203047577797e-06,
+      "loss": 1.064,
+      "step": 4911
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1088651300819153e-06,
+      "loss": 1.1025,
+      "step": 4912
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1079102494729432e-06,
+      "loss": 1.0771,
+      "step": 4913
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1069556631327135e-06,
+      "loss": 1.0337,
+      "step": 4914
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.106001371263012e-06,
+      "loss": 1.1201,
+      "step": 4915
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1050473740655624e-06,
+      "loss": 1.0635,
+      "step": 4916
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1040936717420281e-06,
+      "loss": 1.0918,
+      "step": 4917
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1031402644940076e-06,
+      "loss": 1.1196,
+      "step": 4918
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1021871525230373e-06,
+      "loss": 1.1182,
+      "step": 4919
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1012343360305935e-06,
+      "loss": 1.1279,
+      "step": 4920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.100281815218089e-06,
+      "loss": 1.1113,
+      "step": 4921
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0993295902868723e-06,
+      "loss": 1.0884,
+      "step": 4922
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.098377661438232e-06,
+      "loss": 1.1289,
+      "step": 4923
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0974260288733916e-06,
+      "loss": 1.0908,
+      "step": 4924
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0964746927935143e-06,
+      "loss": 1.1025,
+      "step": 4925
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0955236533996993e-06,
+      "loss": 1.1152,
+      "step": 4926
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.094572910892982e-06,
+      "loss": 1.0693,
+      "step": 4927
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0936224654743385e-06,
+      "loss": 1.1187,
+      "step": 4928
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0926723173446789e-06,
+      "loss": 1.1523,
+      "step": 4929
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0917224667048507e-06,
+      "loss": 1.1001,
+      "step": 4930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0907729137556407e-06,
+      "loss": 1.1562,
+      "step": 4931
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0898236586977703e-06,
+      "loss": 1.1543,
+      "step": 4932
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0888747017318997e-06,
+      "loss": 1.0762,
+      "step": 4933
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.087926043058626e-06,
+      "loss": 1.1357,
+      "step": 4934
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0869776828784814e-06,
+      "loss": 1.1221,
+      "step": 4935
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0860296213919376e-06,
+      "loss": 1.0781,
+      "step": 4936
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0850818587994014e-06,
+      "loss": 1.0347,
+      "step": 4937
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0841343953012156e-06,
+      "loss": 1.1064,
+      "step": 4938
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.083187231097663e-06,
+      "loss": 1.0767,
+      "step": 4939
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.08224036638896e-06,
+      "loss": 1.0625,
+      "step": 4940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0812938013752601e-06,
+      "loss": 1.1074,
+      "step": 4941
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0803475362566562e-06,
+      "loss": 1.0806,
+      "step": 4942
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.079401571233174e-06,
+      "loss": 1.0947,
+      "step": 4943
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0784559065047781e-06,
+      "loss": 1.1328,
+      "step": 4944
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0775105422713705e-06,
+      "loss": 1.0781,
+      "step": 4945
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.076565478732786e-06,
+      "loss": 1.084,
+      "step": 4946
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0756207160888003e-06,
+      "loss": 1.125,
+      "step": 4947
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0746762545391218e-06,
+      "loss": 1.1143,
+      "step": 4948
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0737320942833968e-06,
+      "loss": 1.1064,
+      "step": 4949
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0727882355212087e-06,
+      "loss": 1.1152,
+      "step": 4950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0718446784520764e-06,
+      "loss": 1.1191,
+      "step": 4951
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.070901423275453e-06,
+      "loss": 1.1362,
+      "step": 4952
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0699584701907326e-06,
+      "loss": 1.0601,
+      "step": 4953
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.06901581939724e-06,
+      "loss": 1.1143,
+      "step": 4954
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0680734710942409e-06,
+      "loss": 1.1338,
+      "step": 4955
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.067131425480934e-06,
+      "loss": 1.1143,
+      "step": 4956
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0661896827564535e-06,
+      "loss": 1.0918,
+      "step": 4957
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0652482431198725e-06,
+      "loss": 1.0693,
+      "step": 4958
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0643071067701989e-06,
+      "loss": 1.0762,
+      "step": 4959
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0633662739063742e-06,
+      "loss": 1.0688,
+      "step": 4960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0624257447272796e-06,
+      "loss": 1.1035,
+      "step": 4961
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.061485519431729e-06,
+      "loss": 1.0996,
+      "step": 4962
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0605455982184725e-06,
+      "loss": 1.0947,
+      "step": 4963
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0596059812861986e-06,
+      "loss": 1.0928,
+      "step": 4964
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0586666688335276e-06,
+      "loss": 1.1221,
+      "step": 4965
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0577276610590186e-06,
+      "loss": 1.0771,
+      "step": 4966
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0567889581611645e-06,
+      "loss": 1.1123,
+      "step": 4967
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0558505603383934e-06,
+      "loss": 1.1152,
+      "step": 4968
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0549124677890703e-06,
+      "loss": 1.1128,
+      "step": 4969
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0539746807114967e-06,
+      "loss": 1.1167,
+      "step": 4970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0530371993039053e-06,
+      "loss": 1.085,
+      "step": 4971
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0521000237644697e-06,
+      "loss": 1.1084,
+      "step": 4972
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0511631542912942e-06,
+      "loss": 1.1279,
+      "step": 4973
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0502265910824196e-06,
+      "loss": 1.0972,
+      "step": 4974
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0492903343358244e-06,
+      "loss": 1.1333,
+      "step": 4975
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0483543842494187e-06,
+      "loss": 1.1514,
+      "step": 4976
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0474187410210516e-06,
+      "loss": 1.1074,
+      "step": 4977
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0464834048485036e-06,
+      "loss": 1.1162,
+      "step": 4978
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0455483759294918e-06,
+      "loss": 1.0771,
+      "step": 4979
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0446136544616703e-06,
+      "loss": 1.0527,
+      "step": 4980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0436792406426241e-06,
+      "loss": 1.1528,
+      "step": 4981
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0427451346698769e-06,
+      "loss": 1.126,
+      "step": 4982
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.041811336740887e-06,
+      "loss": 1.0908,
+      "step": 4983
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0408778470530449e-06,
+      "loss": 1.0977,
+      "step": 4984
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.039944665803677e-06,
+      "loss": 1.1309,
+      "step": 4985
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0390117931900473e-06,
+      "loss": 1.1172,
+      "step": 4986
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.03807922940935e-06,
+      "loss": 1.1089,
+      "step": 4987
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0371469746587182e-06,
+      "loss": 1.1147,
+      "step": 4988
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.036215029135217e-06,
+      "loss": 1.084,
+      "step": 4989
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.035283393035846e-06,
+      "loss": 1.1016,
+      "step": 4990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0343520665575425e-06,
+      "loss": 1.0513,
+      "step": 4991
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0334210498971745e-06,
+      "loss": 1.1182,
+      "step": 4992
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.032490343251545e-06,
+      "loss": 1.1206,
+      "step": 4993
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0315599468173959e-06,
+      "loss": 1.127,
+      "step": 4994
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.030629860791399e-06,
+      "loss": 1.1074,
+      "step": 4995
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0297000853701604e-06,
+      "loss": 1.1504,
+      "step": 4996
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.028770620750224e-06,
+      "loss": 1.0947,
+      "step": 4997
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.027841467128064e-06,
+      "loss": 1.1279,
+      "step": 4998
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0269126247000923e-06,
+      "loss": 1.1318,
+      "step": 4999
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0259840936626531e-06,
+      "loss": 1.1309,
+      "step": 5000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.025055874212024e-06,
+      "loss": 1.1006,
+      "step": 5001
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0241279665444195e-06,
+      "loss": 1.1475,
+      "step": 5002
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.023200370855986e-06,
+      "loss": 1.1035,
+      "step": 5003
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0222730873428036e-06,
+      "loss": 1.0835,
+      "step": 5004
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0213461162008883e-06,
+      "loss": 1.1006,
+      "step": 5005
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0204194576261896e-06,
+      "loss": 1.0483,
+      "step": 5006
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0194931118145895e-06,
+      "loss": 1.1182,
+      "step": 5007
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.018567078961906e-06,
+      "loss": 1.0742,
+      "step": 5008
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0176413592638878e-06,
+      "loss": 1.0874,
+      "step": 5009
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0167159529162218e-06,
+      "loss": 1.1064,
+      "step": 5010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.015790860114525e-06,
+      "loss": 1.1035,
+      "step": 5011
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0148660810543487e-06,
+      "loss": 1.0874,
+      "step": 5012
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.01394161593118e-06,
+      "loss": 1.1201,
+      "step": 5013
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0130174649404375e-06,
+      "loss": 1.1338,
+      "step": 5014
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0120936282774732e-06,
+      "loss": 1.0972,
+      "step": 5015
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0111701061375752e-06,
+      "loss": 1.1104,
+      "step": 5016
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.010246898715962e-06,
+      "loss": 1.2012,
+      "step": 5017
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0093240062077872e-06,
+      "loss": 1.1338,
+      "step": 5018
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0084014288081396e-06,
+      "loss": 1.0981,
+      "step": 5019
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0074791667120368e-06,
+      "loss": 1.1025,
+      "step": 5020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0065572201144344e-06,
+      "loss": 1.1318,
+      "step": 5021
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0056355892102185e-06,
+      "loss": 1.0962,
+      "step": 5022
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0047142741942088e-06,
+      "loss": 1.1216,
+      "step": 5023
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0037932752611596e-06,
+      "loss": 1.0889,
+      "step": 5024
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.002872592605757e-06,
+      "loss": 1.1382,
+      "step": 5025
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0019522264226201e-06,
+      "loss": 1.1504,
+      "step": 5026
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0010321769063033e-06,
+      "loss": 1.0942,
+      "step": 5027
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0001124442512907e-06,
+      "loss": 1.1191,
+      "step": 5028
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.99193028652003e-07,
+      "loss": 1.1406,
+      "step": 5029
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.982739303027903e-07,
+      "loss": 1.0957,
+      "step": 5030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.973551493979384e-07,
+      "loss": 1.1084,
+      "step": 5031
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.964366861316655e-07,
+      "loss": 1.1011,
+      "step": 5032
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.95518540698122e-07,
+      "loss": 1.0601,
+      "step": 5033
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.946007132913898e-07,
+      "loss": 1.1494,
+      "step": 5034
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.93683204105487e-07,
+      "loss": 1.0889,
+      "step": 5035
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.92766013334362e-07,
+      "loss": 1.1289,
+      "step": 5036
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.91849141171895e-07,
+      "loss": 1.123,
+      "step": 5037
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.909325878119019e-07,
+      "loss": 1.0767,
+      "step": 5038
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.900163534481285e-07,
+      "loss": 1.1045,
+      "step": 5039
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.891004382742554e-07,
+      "loss": 1.1045,
+      "step": 5040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.881848424838938e-07,
+      "loss": 1.1533,
+      "step": 5041
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.872695662705864e-07,
+      "loss": 1.0889,
+      "step": 5042
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.863546098278137e-07,
+      "loss": 1.1084,
+      "step": 5043
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.85439973348983e-07,
+      "loss": 1.1567,
+      "step": 5044
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.845256570274352e-07,
+      "loss": 1.1006,
+      "step": 5045
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.836116610564461e-07,
+      "loss": 1.0898,
+      "step": 5046
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.826979856292209e-07,
+      "loss": 1.1123,
+      "step": 5047
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.817846309388973e-07,
+      "loss": 1.1338,
+      "step": 5048
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.808715971785475e-07,
+      "loss": 1.1152,
+      "step": 5049
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.799588845411734e-07,
+      "loss": 1.1221,
+      "step": 5050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.790464932197109e-07,
+      "loss": 1.0884,
+      "step": 5051
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.781344234070265e-07,
+      "loss": 1.1045,
+      "step": 5052
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.772226752959183e-07,
+      "loss": 1.1367,
+      "step": 5053
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.763112490791183e-07,
+      "loss": 1.1016,
+      "step": 5054
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.754001449492907e-07,
+      "loss": 1.1035,
+      "step": 5055
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.74489363099028e-07,
+      "loss": 1.0708,
+      "step": 5056
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.735789037208595e-07,
+      "loss": 1.0996,
+      "step": 5057
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.726687670072427e-07,
+      "loss": 1.0918,
+      "step": 5058
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.717589531505673e-07,
+      "loss": 1.127,
+      "step": 5059
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.70849462343157e-07,
+      "loss": 1.1162,
+      "step": 5060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.699402947772644e-07,
+      "loss": 1.0801,
+      "step": 5061
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.690314506450763e-07,
+      "loss": 1.1318,
+      "step": 5062
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.681229301387095e-07,
+      "loss": 1.0898,
+      "step": 5063
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.672147334502118e-07,
+      "loss": 1.1079,
+      "step": 5064
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.663068607715653e-07,
+      "loss": 1.0806,
+      "step": 5065
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.6539931229468e-07,
+      "loss": 1.0859,
+      "step": 5066
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.644920882114003e-07,
+      "loss": 1.0781,
+      "step": 5067
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.635851887135016e-07,
+      "loss": 1.0645,
+      "step": 5068
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.626786139926897e-07,
+      "loss": 1.0869,
+      "step": 5069
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.617723642406007e-07,
+      "loss": 1.1416,
+      "step": 5070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.608664396488051e-07,
+      "loss": 1.0996,
+      "step": 5071
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.599608404088018e-07,
+      "loss": 1.0996,
+      "step": 5072
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.590555667120235e-07,
+      "loss": 1.0747,
+      "step": 5073
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.581506187498315e-07,
+      "loss": 1.1396,
+      "step": 5074
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.572459967135193e-07,
+      "loss": 1.0898,
+      "step": 5075
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.563417007943127e-07,
+      "loss": 1.1094,
+      "step": 5076
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.554377311833667e-07,
+      "loss": 1.0728,
+      "step": 5077
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.54534088071767e-07,
+      "loss": 1.1221,
+      "step": 5078
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.536307716505341e-07,
+      "loss": 1.1201,
+      "step": 5079
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.527277821106154e-07,
+      "loss": 1.123,
+      "step": 5080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.518251196428896e-07,
+      "loss": 1.1299,
+      "step": 5081
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.509227844381691e-07,
+      "loss": 1.1504,
+      "step": 5082
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.500207766871933e-07,
+      "loss": 1.1377,
+      "step": 5083
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.49119096580636e-07,
+      "loss": 1.1416,
+      "step": 5084
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.482177443090992e-07,
+      "loss": 1.1152,
+      "step": 5085
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.473167200631156e-07,
+      "loss": 1.0684,
+      "step": 5086
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.464160240331513e-07,
+      "loss": 1.084,
+      "step": 5087
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.455156564096002e-07,
+      "loss": 1.1045,
+      "step": 5088
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.446156173827867e-07,
+      "loss": 1.1016,
+      "step": 5089
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.437159071429685e-07,
+      "loss": 1.1289,
+      "step": 5090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.428165258803307e-07,
+      "loss": 1.0762,
+      "step": 5091
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.419174737849904e-07,
+      "loss": 1.123,
+      "step": 5092
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.410187510469962e-07,
+      "loss": 1.0903,
+      "step": 5093
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.401203578563239e-07,
+      "loss": 1.1211,
+      "step": 5094
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.392222944028836e-07,
+      "loss": 1.0967,
+      "step": 5095
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.383245608765124e-07,
+      "loss": 1.166,
+      "step": 5096
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.374271574669782e-07,
+      "loss": 1.0825,
+      "step": 5097
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.365300843639816e-07,
+      "loss": 1.0815,
+      "step": 5098
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.356333417571506e-07,
+      "loss": 1.0854,
+      "step": 5099
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.347369298360434e-07,
+      "loss": 1.126,
+      "step": 5100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.33840848790151e-07,
+      "loss": 1.1064,
+      "step": 5101
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.329450988088912e-07,
+      "loss": 1.1221,
+      "step": 5102
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.32049680081614e-07,
+      "loss": 1.1094,
+      "step": 5103
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.311545927975993e-07,
+      "loss": 1.1035,
+      "step": 5104
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.30259837146055e-07,
+      "loss": 1.0615,
+      "step": 5105
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.293654133161214e-07,
+      "loss": 1.0576,
+      "step": 5106
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.284713214968674e-07,
+      "loss": 1.0957,
+      "step": 5107
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.275775618772903e-07,
+      "loss": 1.1299,
+      "step": 5108
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.266841346463204e-07,
+      "loss": 1.0977,
+      "step": 5109
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.257910399928155e-07,
+      "loss": 1.0928,
+      "step": 5110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.248982781055624e-07,
+      "loss": 1.124,
+      "step": 5111
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.240058491732806e-07,
+      "loss": 1.1006,
+      "step": 5112
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.231137533846158e-07,
+      "loss": 1.1108,
+      "step": 5113
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.222219909281466e-07,
+      "loss": 1.0801,
+      "step": 5114
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.213305619923771e-07,
+      "loss": 1.1289,
+      "step": 5115
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.204394667657448e-07,
+      "loss": 1.1084,
+      "step": 5116
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.195487054366153e-07,
+      "loss": 1.1191,
+      "step": 5117
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.186582781932832e-07,
+      "loss": 1.123,
+      "step": 5118
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.177681852239711e-07,
+      "loss": 1.127,
+      "step": 5119
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.168784267168346e-07,
+      "loss": 1.124,
+      "step": 5120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.159890028599552e-07,
+      "loss": 1.0771,
+      "step": 5121
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.150999138413446e-07,
+      "loss": 1.1113,
+      "step": 5122
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.142111598489455e-07,
+      "loss": 1.1211,
+      "step": 5123
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.133227410706269e-07,
+      "loss": 1.0825,
+      "step": 5124
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.124346576941897e-07,
+      "loss": 1.1201,
+      "step": 5125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.11546909907362e-07,
+      "loss": 1.1338,
+      "step": 5126
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.106594978977998e-07,
+      "loss": 1.0825,
+      "step": 5127
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.097724218530931e-07,
+      "loss": 1.1064,
+      "step": 5128
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.088856819607562e-07,
+      "loss": 1.1162,
+      "step": 5129
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.079992784082328e-07,
+      "loss": 1.1191,
+      "step": 5130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.071132113828982e-07,
+      "loss": 1.1465,
+      "step": 5131
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.06227481072054e-07,
+      "loss": 1.1035,
+      "step": 5132
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.053420876629307e-07,
+      "loss": 1.0864,
+      "step": 5133
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.044570313426898e-07,
+      "loss": 1.1079,
+      "step": 5134
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.035723122984189e-07,
+      "loss": 1.1611,
+      "step": 5135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.026879307171368e-07,
+      "loss": 1.1104,
+      "step": 5136
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.01803886785789e-07,
+      "loss": 1.1396,
+      "step": 5137
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.009201806912494e-07,
+      "loss": 1.1094,
+      "step": 5138
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.000368126203221e-07,
+      "loss": 1.1167,
+      "step": 5139
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.991537827597402e-07,
+      "loss": 1.0859,
+      "step": 5140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.982710912961623e-07,
+      "loss": 1.0942,
+      "step": 5141
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.973887384161789e-07,
+      "loss": 1.0801,
+      "step": 5142
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.965067243063063e-07,
+      "loss": 1.1123,
+      "step": 5143
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.9562504915299e-07,
+      "loss": 1.0903,
+      "step": 5144
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.947437131426054e-07,
+      "loss": 1.1465,
+      "step": 5145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.938627164614531e-07,
+      "loss": 1.1021,
+      "step": 5146
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.929820592957656e-07,
+      "loss": 1.0947,
+      "step": 5147
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.921017418317008e-07,
+      "loss": 1.0605,
+      "step": 5148
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.91221764255345e-07,
+      "loss": 1.0947,
+      "step": 5149
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.903421267527152e-07,
+      "loss": 1.0996,
+      "step": 5150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.89462829509753e-07,
+      "loss": 1.104,
+      "step": 5151
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.885838727123306e-07,
+      "loss": 1.1079,
+      "step": 5152
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.877052565462483e-07,
+      "loss": 1.0898,
+      "step": 5153
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.868269811972327e-07,
+      "loss": 1.1611,
+      "step": 5154
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.859490468509385e-07,
+      "loss": 1.1377,
+      "step": 5155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.850714536929506e-07,
+      "loss": 1.1641,
+      "step": 5156
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.841942019087784e-07,
+      "loss": 1.0947,
+      "step": 5157
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.833172916838629e-07,
+      "loss": 1.1172,
+      "step": 5158
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.824407232035698e-07,
+      "loss": 1.0996,
+      "step": 5159
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.815644966531931e-07,
+      "loss": 1.1045,
+      "step": 5160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.806886122179567e-07,
+      "loss": 1.0913,
+      "step": 5161
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.798130700830087e-07,
+      "loss": 1.1069,
+      "step": 5162
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.789378704334287e-07,
+      "loss": 1.0928,
+      "step": 5163
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.7806301345422e-07,
+      "loss": 1.1016,
+      "step": 5164
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.771884993303176e-07,
+      "loss": 1.1299,
+      "step": 5165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.763143282465797e-07,
+      "loss": 1.1035,
+      "step": 5166
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.754405003877958e-07,
+      "loss": 1.1123,
+      "step": 5167
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.745670159386796e-07,
+      "loss": 1.127,
+      "step": 5168
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.736938750838752e-07,
+      "loss": 1.123,
+      "step": 5169
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.728210780079524e-07,
+      "loss": 1.125,
+      "step": 5170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.719486248954074e-07,
+      "loss": 1.0986,
+      "step": 5171
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.710765159306661e-07,
+      "loss": 1.0786,
+      "step": 5172
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.702047512980794e-07,
+      "loss": 1.0854,
+      "step": 5173
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.693333311819279e-07,
+      "loss": 1.1143,
+      "step": 5174
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.684622557664168e-07,
+      "loss": 1.1025,
+      "step": 5175
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.67591525235679e-07,
+      "loss": 1.1211,
+      "step": 5176
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.66721139773776e-07,
+      "loss": 1.0903,
+      "step": 5177
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.658510995646957e-07,
+      "loss": 1.0889,
+      "step": 5178
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.649814047923513e-07,
+      "loss": 1.1289,
+      "step": 5179
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.64112055640586e-07,
+      "loss": 1.0923,
+      "step": 5180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.632430522931679e-07,
+      "loss": 1.1113,
+      "step": 5181
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.623743949337909e-07,
+      "loss": 1.1123,
+      "step": 5182
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.615060837460795e-07,
+      "loss": 1.0898,
+      "step": 5183
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.60638118913581e-07,
+      "loss": 1.1104,
+      "step": 5184
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.597705006197729e-07,
+      "loss": 1.0908,
+      "step": 5185
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.589032290480571e-07,
+      "loss": 1.1064,
+      "step": 5186
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.580363043817618e-07,
+      "loss": 1.1299,
+      "step": 5187
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.571697268041446e-07,
+      "loss": 1.0908,
+      "step": 5188
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.56303496498388e-07,
+      "loss": 1.0918,
+      "step": 5189
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.554376136476003e-07,
+      "loss": 1.0981,
+      "step": 5190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.545720784348188e-07,
+      "loss": 1.1152,
+      "step": 5191
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.537068910430049e-07,
+      "loss": 1.1172,
+      "step": 5192
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.528420516550465e-07,
+      "loss": 1.0986,
+      "step": 5193
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.519775604537606e-07,
+      "loss": 1.1104,
+      "step": 5194
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.511134176218871e-07,
+      "loss": 1.1201,
+      "step": 5195
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.502496233420959e-07,
+      "loss": 1.1494,
+      "step": 5196
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.4938617779698e-07,
+      "loss": 1.1113,
+      "step": 5197
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.485230811690595e-07,
+      "loss": 1.1172,
+      "step": 5198
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.476603336407827e-07,
+      "loss": 1.1025,
+      "step": 5199
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.467979353945211e-07,
+      "loss": 1.1396,
+      "step": 5200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.459358866125747e-07,
+      "loss": 1.1025,
+      "step": 5201
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.450741874771695e-07,
+      "loss": 1.0938,
+      "step": 5202
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.442128381704562e-07,
+      "loss": 1.0908,
+      "step": 5203
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.433518388745116e-07,
+      "loss": 1.0498,
+      "step": 5204
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.424911897713409e-07,
+      "loss": 1.1387,
+      "step": 5205
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.416308910428714e-07,
+      "loss": 1.1011,
+      "step": 5206
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.407709428709604e-07,
+      "loss": 1.1221,
+      "step": 5207
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.399113454373886e-07,
+      "loss": 1.1074,
+      "step": 5208
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.39052098923862e-07,
+      "loss": 1.0728,
+      "step": 5209
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.381932035120155e-07,
+      "loss": 1.1504,
+      "step": 5210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.373346593834067e-07,
+      "loss": 1.0635,
+      "step": 5211
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.364764667195188e-07,
+      "loss": 1.1191,
+      "step": 5212
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.356186257017648e-07,
+      "loss": 1.1064,
+      "step": 5213
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.347611365114793e-07,
+      "loss": 1.0884,
+      "step": 5214
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.339039993299233e-07,
+      "loss": 1.1333,
+      "step": 5215
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.330472143382848e-07,
+      "loss": 1.123,
+      "step": 5216
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.32190781717675e-07,
+      "loss": 1.1201,
+      "step": 5217
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.313347016491338e-07,
+      "loss": 1.0952,
+      "step": 5218
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.304789743136241e-07,
+      "loss": 1.0693,
+      "step": 5219
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.29623599892034e-07,
+      "loss": 1.0811,
+      "step": 5220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.287685785651797e-07,
+      "loss": 1.1133,
+      "step": 5221
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.279139105138001e-07,
+      "loss": 1.0869,
+      "step": 5222
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.270595959185595e-07,
+      "loss": 1.0894,
+      "step": 5223
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.2620563496005e-07,
+      "loss": 1.0889,
+      "step": 5224
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.253520278187857e-07,
+      "loss": 1.0879,
+      "step": 5225
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.244987746752084e-07,
+      "loss": 1.0723,
+      "step": 5226
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.236458757096847e-07,
+      "loss": 1.1045,
+      "step": 5227
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.22793331102504e-07,
+      "loss": 1.1201,
+      "step": 5228
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.219411410338849e-07,
+      "loss": 1.0747,
+      "step": 5229
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.210893056839673e-07,
+      "loss": 1.1152,
+      "step": 5230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.20237825232817e-07,
+      "loss": 1.1094,
+      "step": 5231
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.193866998604266e-07,
+      "loss": 1.1631,
+      "step": 5232
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.185359297467121e-07,
+      "loss": 1.0977,
+      "step": 5233
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.176855150715135e-07,
+      "loss": 1.0815,
+      "step": 5234
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.168354560145985e-07,
+      "loss": 1.1016,
+      "step": 5235
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.159857527556567e-07,
+      "loss": 1.0757,
+      "step": 5236
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.151364054743038e-07,
+      "loss": 1.0869,
+      "step": 5237
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.142874143500817e-07,
+      "loss": 1.0601,
+      "step": 5238
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.134387795624546e-07,
+      "loss": 1.0996,
+      "step": 5239
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.125905012908111e-07,
+      "loss": 1.0654,
+      "step": 5240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.117425797144676e-07,
+      "loss": 1.127,
+      "step": 5241
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.108950150126613e-07,
+      "loss": 1.1338,
+      "step": 5242
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.100478073645577e-07,
+      "loss": 1.0947,
+      "step": 5243
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.09200956949244e-07,
+      "loss": 1.1182,
+      "step": 5244
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.083544639457316e-07,
+      "loss": 1.106,
+      "step": 5245
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.075083285329596e-07,
+      "loss": 1.1289,
+      "step": 5246
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.066625508897879e-07,
+      "loss": 1.1226,
+      "step": 5247
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.058171311950036e-07,
+      "loss": 1.1553,
+      "step": 5248
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.049720696273156e-07,
+      "loss": 1.1016,
+      "step": 5249
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.041273663653598e-07,
+      "loss": 1.0674,
+      "step": 5250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.03283021587693e-07,
+      "loss": 1.1099,
+      "step": 5251
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.024390354728004e-07,
+      "loss": 1.0918,
+      "step": 5252
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.01595408199087e-07,
+      "loss": 1.1025,
+      "step": 5253
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.007521399448858e-07,
+      "loss": 1.0996,
+      "step": 5254
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.999092308884518e-07,
+      "loss": 1.0977,
+      "step": 5255
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.990666812079628e-07,
+      "loss": 1.1313,
+      "step": 5256
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.982244910815249e-07,
+      "loss": 1.1162,
+      "step": 5257
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.973826606871629e-07,
+      "loss": 1.1011,
+      "step": 5258
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.965411902028309e-07,
+      "loss": 1.0684,
+      "step": 5259
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.957000798064027e-07,
+      "loss": 1.1074,
+      "step": 5260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.948593296756768e-07,
+      "loss": 1.1475,
+      "step": 5261
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.940189399883775e-07,
+      "loss": 1.166,
+      "step": 5262
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.931789109221522e-07,
+      "loss": 1.1211,
+      "step": 5263
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.923392426545701e-07,
+      "loss": 1.1045,
+      "step": 5264
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.91499935363127e-07,
+      "loss": 1.0996,
+      "step": 5265
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.906609892252404e-07,
+      "loss": 1.0957,
+      "step": 5266
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.898224044182509e-07,
+      "loss": 1.1123,
+      "step": 5267
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.88984181119426e-07,
+      "loss": 1.0674,
+      "step": 5268
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.881463195059525e-07,
+      "loss": 1.1406,
+      "step": 5269
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.87308819754945e-07,
+      "loss": 1.0884,
+      "step": 5270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.864716820434382e-07,
+      "loss": 1.123,
+      "step": 5271
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.856349065483912e-07,
+      "loss": 1.1289,
+      "step": 5272
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.847984934466873e-07,
+      "loss": 1.1035,
+      "step": 5273
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.83962442915134e-07,
+      "loss": 1.1074,
+      "step": 5274
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.831267551304592e-07,
+      "loss": 1.1138,
+      "step": 5275
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.822914302693171e-07,
+      "loss": 1.1094,
+      "step": 5276
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.814564685082837e-07,
+      "loss": 1.0527,
+      "step": 5277
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.806218700238577e-07,
+      "loss": 1.0972,
+      "step": 5278
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.79787634992463e-07,
+      "loss": 1.1309,
+      "step": 5279
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.78953763590444e-07,
+      "loss": 1.1104,
+      "step": 5280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.781202559940715e-07,
+      "loss": 1.0928,
+      "step": 5281
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.772871123795364e-07,
+      "loss": 1.1138,
+      "step": 5282
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.764543329229537e-07,
+      "loss": 1.0679,
+      "step": 5283
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.756219178003626e-07,
+      "loss": 1.0938,
+      "step": 5284
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.747898671877232e-07,
+      "loss": 1.126,
+      "step": 5285
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.739581812609201e-07,
+      "loss": 1.0859,
+      "step": 5286
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.731268601957612e-07,
+      "loss": 1.1328,
+      "step": 5287
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.722959041679756e-07,
+      "loss": 1.1074,
+      "step": 5288
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.714653133532152e-07,
+      "loss": 1.1309,
+      "step": 5289
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.706350879270571e-07,
+      "loss": 1.0986,
+      "step": 5290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.698052280649984e-07,
+      "loss": 1.1113,
+      "step": 5291
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.689757339424611e-07,
+      "loss": 1.0962,
+      "step": 5292
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.681466057347886e-07,
+      "loss": 1.1445,
+      "step": 5293
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.673178436172462e-07,
+      "loss": 1.1494,
+      "step": 5294
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.664894477650247e-07,
+      "loss": 1.083,
+      "step": 5295
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.656614183532349e-07,
+      "loss": 1.1108,
+      "step": 5296
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.648337555569099e-07,
+      "loss": 1.1289,
+      "step": 5297
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.640064595510068e-07,
+      "loss": 1.1572,
+      "step": 5298
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.631795305104062e-07,
+      "loss": 1.0576,
+      "step": 5299
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.623529686099074e-07,
+      "loss": 1.1201,
+      "step": 5300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.615267740242363e-07,
+      "loss": 1.0771,
+      "step": 5301
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.607009469280372e-07,
+      "loss": 1.1299,
+      "step": 5302
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.598754874958802e-07,
+      "loss": 1.0845,
+      "step": 5303
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.590503959022558e-07,
+      "loss": 1.0952,
+      "step": 5304
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.582256723215761e-07,
+      "loss": 1.0532,
+      "step": 5305
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.574013169281777e-07,
+      "loss": 1.1055,
+      "step": 5306
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.565773298963175e-07,
+      "loss": 1.1113,
+      "step": 5307
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.557537114001742e-07,
+      "loss": 1.1201,
+      "step": 5308
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.549304616138512e-07,
+      "loss": 1.0991,
+      "step": 5309
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.541075807113706e-07,
+      "loss": 1.1055,
+      "step": 5310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.53285068866679e-07,
+      "loss": 1.0947,
+      "step": 5311
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.52462926253645e-07,
+      "loss": 1.0786,
+      "step": 5312
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.516411530460566e-07,
+      "loss": 1.0874,
+      "step": 5313
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.50819749417627e-07,
+      "loss": 1.0967,
+      "step": 5314
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.499987155419891e-07,
+      "loss": 1.0977,
+      "step": 5315
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.491780515926972e-07,
+      "loss": 1.0596,
+      "step": 5316
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.4835775774323e-07,
+      "loss": 1.0527,
+      "step": 5317
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.475378341669858e-07,
+      "loss": 1.1201,
+      "step": 5318
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.467182810372844e-07,
+      "loss": 1.125,
+      "step": 5319
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.458990985273698e-07,
+      "loss": 1.0962,
+      "step": 5320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.450802868104043e-07,
+      "loss": 1.0889,
+      "step": 5321
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.442618460594742e-07,
+      "loss": 1.1191,
+      "step": 5322
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.434437764475874e-07,
+      "loss": 1.0859,
+      "step": 5323
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.426260781476716e-07,
+      "loss": 1.0688,
+      "step": 5324
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.418087513325781e-07,
+      "loss": 1.1328,
+      "step": 5325
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.409917961750779e-07,
+      "loss": 1.0908,
+      "step": 5326
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.401752128478631e-07,
+      "loss": 1.1182,
+      "step": 5327
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.393590015235507e-07,
+      "loss": 1.1118,
+      "step": 5328
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.385431623746749e-07,
+      "loss": 1.1201,
+      "step": 5329
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.377276955736928e-07,
+      "loss": 1.1387,
+      "step": 5330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.369126012929842e-07,
+      "loss": 1.1284,
+      "step": 5331
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.360978797048473e-07,
+      "loss": 1.1016,
+      "step": 5332
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.352835309815046e-07,
+      "loss": 1.1392,
+      "step": 5333
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.34469555295097e-07,
+      "loss": 1.1387,
+      "step": 5334
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.336559528176884e-07,
+      "loss": 1.1436,
+      "step": 5335
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.328427237212643e-07,
+      "loss": 1.0557,
+      "step": 5336
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.320298681777291e-07,
+      "loss": 1.1328,
+      "step": 5337
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.312173863589087e-07,
+      "loss": 1.0977,
+      "step": 5338
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.30405278436552e-07,
+      "loss": 1.1211,
+      "step": 5339
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.295935445823272e-07,
+      "loss": 1.1074,
+      "step": 5340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.287821849678226e-07,
+      "loss": 1.1279,
+      "step": 5341
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.279711997645502e-07,
+      "loss": 1.1318,
+      "step": 5342
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.271605891439396e-07,
+      "loss": 1.0918,
+      "step": 5343
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.263503532773445e-07,
+      "loss": 1.0752,
+      "step": 5344
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.255404923360365e-07,
+      "loss": 1.0552,
+      "step": 5345
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.24731006491208e-07,
+      "loss": 1.0889,
+      "step": 5346
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.239218959139763e-07,
+      "loss": 1.1406,
+      "step": 5347
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.231131607753744e-07,
+      "loss": 1.1172,
+      "step": 5348
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.223048012463574e-07,
+      "loss": 1.1055,
+      "step": 5349
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.214968174978032e-07,
+      "loss": 1.0913,
+      "step": 5350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.206892097005072e-07,
+      "loss": 1.1377,
+      "step": 5351
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.198819780251865e-07,
+      "loss": 1.0884,
+      "step": 5352
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.190751226424797e-07,
+      "loss": 1.0537,
+      "step": 5353
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.182686437229438e-07,
+      "loss": 1.0659,
+      "step": 5354
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.174625414370592e-07,
+      "loss": 1.1035,
+      "step": 5355
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.166568159552239e-07,
+      "loss": 1.1367,
+      "step": 5356
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.158514674477562e-07,
+      "loss": 1.1074,
+      "step": 5357
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.150464960848979e-07,
+      "loss": 1.0576,
+      "step": 5358
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.142419020368069e-07,
+      "loss": 1.1123,
+      "step": 5359
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.134376854735642e-07,
+      "loss": 1.0967,
+      "step": 5360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.126338465651713e-07,
+      "loss": 1.1514,
+      "step": 5361
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.118303854815478e-07,
+      "loss": 1.085,
+      "step": 5362
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.110273023925335e-07,
+      "loss": 1.1465,
+      "step": 5363
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.102245974678909e-07,
+      "loss": 1.0967,
+      "step": 5364
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.094222708772991e-07,
+      "loss": 1.1094,
+      "step": 5365
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.086203227903607e-07,
+      "loss": 1.0977,
+      "step": 5366
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.078187533765957e-07,
+      "loss": 1.0488,
+      "step": 5367
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.070175628054443e-07,
+      "loss": 1.1055,
+      "step": 5368
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.062167512462684e-07,
+      "loss": 1.0781,
+      "step": 5369
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.054163188683483e-07,
+      "loss": 1.1406,
+      "step": 5370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.046162658408826e-07,
+      "loss": 1.061,
+      "step": 5371
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.038165923329948e-07,
+      "loss": 1.0464,
+      "step": 5372
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.030172985137234e-07,
+      "loss": 1.0869,
+      "step": 5373
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.022183845520275e-07,
+      "loss": 1.0781,
+      "step": 5374
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.01419850616788e-07,
+      "loss": 1.1196,
+      "step": 5375
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.006216968768023e-07,
+      "loss": 1.0957,
+      "step": 5376
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.998239235007911e-07,
+      "loss": 1.1187,
+      "step": 5377
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.990265306573921e-07,
+      "loss": 1.0986,
+      "step": 5378
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.982295185151619e-07,
+      "loss": 1.1436,
+      "step": 5379
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.9743288724258e-07,
+      "loss": 1.0679,
+      "step": 5380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.966366370080426e-07,
+      "loss": 1.0835,
+      "step": 5381
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.958407679798651e-07,
+      "loss": 1.124,
+      "step": 5382
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.950452803262841e-07,
+      "loss": 1.0645,
+      "step": 5383
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.942501742154559e-07,
+      "loss": 1.1289,
+      "step": 5384
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.934554498154533e-07,
+      "loss": 1.1377,
+      "step": 5385
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.926611072942715e-07,
+      "loss": 1.1279,
+      "step": 5386
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.918671468198227e-07,
+      "loss": 1.1182,
+      "step": 5387
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.910735685599401e-07,
+      "loss": 1.1289,
+      "step": 5388
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.902803726823751e-07,
+      "loss": 1.165,
+      "step": 5389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.894875593547975e-07,
+      "loss": 1.1494,
+      "step": 5390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.886951287447985e-07,
+      "loss": 1.1475,
+      "step": 5391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.879030810198867e-07,
+      "loss": 1.1406,
+      "step": 5392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.871114163474893e-07,
+      "loss": 1.1084,
+      "step": 5393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.863201348949547e-07,
+      "loss": 1.084,
+      "step": 5394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.855292368295477e-07,
+      "loss": 1.1167,
+      "step": 5395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.847387223184543e-07,
+      "loss": 1.1416,
+      "step": 5396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.839485915287785e-07,
+      "loss": 1.1396,
+      "step": 5397
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.831588446275422e-07,
+      "loss": 1.1504,
+      "step": 5398
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.823694817816887e-07,
+      "loss": 1.0996,
+      "step": 5399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.815805031580772e-07,
+      "loss": 1.1191,
+      "step": 5400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.807919089234869e-07,
+      "loss": 1.0771,
+      "step": 5401
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.800036992446169e-07,
+      "loss": 1.0444,
+      "step": 5402
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.792158742880833e-07,
+      "loss": 1.125,
+      "step": 5403
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.784284342204206e-07,
+      "loss": 1.0981,
+      "step": 5404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.776413792080849e-07,
+      "loss": 1.1265,
+      "step": 5405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.768547094174468e-07,
+      "loss": 1.0894,
+      "step": 5406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.760684250147986e-07,
+      "loss": 1.0859,
+      "step": 5407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.75282526166351e-07,
+      "loss": 1.1426,
+      "step": 5408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.7449701303823e-07,
+      "loss": 1.0967,
+      "step": 5409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.737118857964847e-07,
+      "loss": 1.1152,
+      "step": 5410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.729271446070793e-07,
+      "loss": 1.0981,
+      "step": 5411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.721427896358962e-07,
+      "loss": 1.041,
+      "step": 5412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.713588210487393e-07,
+      "loss": 1.1201,
+      "step": 5413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.705752390113276e-07,
+      "loss": 1.0962,
+      "step": 5414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.697920436892994e-07,
+      "loss": 1.0898,
+      "step": 5415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.690092352482127e-07,
+      "loss": 1.1309,
+      "step": 5416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.682268138535411e-07,
+      "loss": 1.1094,
+      "step": 5417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.674447796706793e-07,
+      "loss": 1.0732,
+      "step": 5418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.666631328649367e-07,
+      "loss": 1.105,
+      "step": 5419
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.658818736015443e-07,
+      "loss": 1.0898,
+      "step": 5420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.651010020456497e-07,
+      "loss": 1.1162,
+      "step": 5421
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.643205183623183e-07,
+      "loss": 1.0781,
+      "step": 5422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.635404227165323e-07,
+      "loss": 1.1323,
+      "step": 5423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.627607152731952e-07,
+      "loss": 1.1118,
+      "step": 5424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.619813961971256e-07,
+      "loss": 1.1396,
+      "step": 5425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.612024656530599e-07,
+      "loss": 1.0879,
+      "step": 5426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.604239238056553e-07,
+      "loss": 1.0952,
+      "step": 5427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.596457708194834e-07,
+      "loss": 1.1143,
+      "step": 5428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.588680068590361e-07,
+      "loss": 1.1514,
+      "step": 5429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.580906320887217e-07,
+      "loss": 1.0884,
+      "step": 5430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.57313646672866e-07,
+      "loss": 1.1045,
+      "step": 5431
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.565370507757137e-07,
+      "loss": 1.0986,
+      "step": 5432
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.557608445614269e-07,
+      "loss": 1.0698,
+      "step": 5433
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.549850281940842e-07,
+      "loss": 1.0835,
+      "step": 5434
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.542096018376834e-07,
+      "loss": 1.1387,
+      "step": 5435
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.534345656561389e-07,
+      "loss": 1.1016,
+      "step": 5436
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.526599198132813e-07,
+      "loss": 1.1064,
+      "step": 5437
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.518856644728621e-07,
+      "loss": 1.0962,
+      "step": 5438
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.511117997985469e-07,
+      "loss": 1.1055,
+      "step": 5439
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.503383259539215e-07,
+      "loss": 1.082,
+      "step": 5440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.495652431024866e-07,
+      "loss": 1.167,
+      "step": 5441
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.48792551407661e-07,
+      "loss": 1.0747,
+      "step": 5442
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.480202510327827e-07,
+      "loss": 1.1055,
+      "step": 5443
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.472483421411036e-07,
+      "loss": 1.124,
+      "step": 5444
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.464768248957953e-07,
+      "loss": 1.1045,
+      "step": 5445
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.457056994599475e-07,
+      "loss": 1.0996,
+      "step": 5446
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.449349659965643e-07,
+      "loss": 1.1211,
+      "step": 5447
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.441646246685676e-07,
+      "loss": 1.1069,
+      "step": 5448
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.433946756387982e-07,
+      "loss": 1.0986,
+      "step": 5449
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.426251190700117e-07,
+      "loss": 1.0991,
+      "step": 5450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.418559551248835e-07,
+      "loss": 1.0859,
+      "step": 5451
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.410871839660029e-07,
+      "loss": 1.1035,
+      "step": 5452
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.403188057558777e-07,
+      "loss": 1.0562,
+      "step": 5453
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.395508206569331e-07,
+      "loss": 1.1333,
+      "step": 5454
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.387832288315099e-07,
+      "loss": 1.0938,
+      "step": 5455
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.380160304418672e-07,
+      "loss": 1.1055,
+      "step": 5456
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.372492256501805e-07,
+      "loss": 1.0757,
+      "step": 5457
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.364828146185418e-07,
+      "loss": 1.0967,
+      "step": 5458
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.357167975089588e-07,
+      "loss": 1.1152,
+      "step": 5459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.349511744833584e-07,
+      "loss": 1.1172,
+      "step": 5460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.341859457035812e-07,
+      "loss": 1.1367,
+      "step": 5461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.334211113313882e-07,
+      "loss": 1.0918,
+      "step": 5462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.326566715284537e-07,
+      "loss": 1.125,
+      "step": 5463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.318926264563696e-07,
+      "loss": 1.0986,
+      "step": 5464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.311289762766454e-07,
+      "loss": 1.0405,
+      "step": 5465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.303657211507053e-07,
+      "loss": 1.1113,
+      "step": 5466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.296028612398922e-07,
+      "loss": 1.0698,
+      "step": 5467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.28840396705463e-07,
+      "loss": 1.0986,
+      "step": 5468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.280783277085936e-07,
+      "loss": 1.0869,
+      "step": 5469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.273166544103737e-07,
+      "loss": 1.0996,
+      "step": 5470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.265553769718117e-07,
+      "loss": 1.1069,
+      "step": 5471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.2579449555383e-07,
+      "loss": 1.0781,
+      "step": 5472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.250340103172702e-07,
+      "loss": 1.0811,
+      "step": 5473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.242739214228874e-07,
+      "loss": 1.105,
+      "step": 5474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.235142290313537e-07,
+      "loss": 1.1152,
+      "step": 5475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.227549333032587e-07,
+      "loss": 1.1021,
+      "step": 5476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.219960343991058e-07,
+      "loss": 1.1416,
+      "step": 5477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.212375324793174e-07,
+      "loss": 1.1001,
+      "step": 5478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.204794277042298e-07,
+      "loss": 1.1113,
+      "step": 5479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.197217202340949e-07,
+      "loss": 1.124,
+      "step": 5480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.189644102290829e-07,
+      "loss": 1.0786,
+      "step": 5481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.182074978492791e-07,
+      "loss": 1.127,
+      "step": 5482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.174509832546832e-07,
+      "loss": 1.1064,
+      "step": 5483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.166948666052134e-07,
+      "loss": 1.1113,
+      "step": 5484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.159391480607019e-07,
+      "loss": 1.0996,
+      "step": 5485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.151838277808961e-07,
+      "loss": 1.1494,
+      "step": 5486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.144289059254624e-07,
+      "loss": 1.126,
+      "step": 5487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.13674382653979e-07,
+      "loss": 1.1348,
+      "step": 5488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.129202581259436e-07,
+      "loss": 1.1553,
+      "step": 5489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.121665325007669e-07,
+      "loss": 1.1074,
+      "step": 5490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.114132059377756e-07,
+      "loss": 1.106,
+      "step": 5491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.10660278596214e-07,
+      "loss": 1.1162,
+      "step": 5492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.099077506352391e-07,
+      "loss": 1.1328,
+      "step": 5493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.091556222139261e-07,
+      "loss": 1.0854,
+      "step": 5494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.08403893491265e-07,
+      "loss": 1.1084,
+      "step": 5495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.076525646261602e-07,
+      "loss": 1.0859,
+      "step": 5496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.069016357774321e-07,
+      "loss": 1.1152,
+      "step": 5497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.061511071038182e-07,
+      "loss": 1.1045,
+      "step": 5498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.054009787639678e-07,
+      "loss": 1.0591,
+      "step": 5499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.046512509164501e-07,
+      "loss": 1.0186,
+      "step": 5500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.039019237197461e-07,
+      "loss": 1.0664,
+      "step": 5501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.031529973322528e-07,
+      "loss": 1.1133,
+      "step": 5502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.024044719122843e-07,
+      "loss": 1.0913,
+      "step": 5503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.016563476180681e-07,
+      "loss": 1.1064,
+      "step": 5504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.009086246077458e-07,
+      "loss": 1.1162,
+      "step": 5505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.001613030393791e-07,
+      "loss": 1.0537,
+      "step": 5506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.994143830709395e-07,
+      "loss": 1.0854,
+      "step": 5507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.986678648603156e-07,
+      "loss": 1.0542,
+      "step": 5508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.979217485653118e-07,
+      "loss": 1.0747,
+      "step": 5509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.971760343436464e-07,
+      "loss": 1.1064,
+      "step": 5510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.964307223529539e-07,
+      "loss": 1.0801,
+      "step": 5511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.956858127507828e-07,
+      "loss": 1.0991,
+      "step": 5512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.949413056945957e-07,
+      "loss": 1.1006,
+      "step": 5513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.941972013417729e-07,
+      "loss": 1.1113,
+      "step": 5514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.934534998496072e-07,
+      "loss": 1.0723,
+      "step": 5515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.927102013753058e-07,
+      "loss": 1.1338,
+      "step": 5516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.919673060759931e-07,
+      "loss": 1.1182,
+      "step": 5517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.912248141087076e-07,
+      "loss": 1.0713,
+      "step": 5518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.904827256304003e-07,
+      "loss": 1.0898,
+      "step": 5519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.897410407979401e-07,
+      "loss": 1.1099,
+      "step": 5520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.889997597681074e-07,
+      "loss": 1.061,
+      "step": 5521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.882588826976007e-07,
+      "loss": 1.1006,
+      "step": 5522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.875184097430303e-07,
+      "loss": 1.1699,
+      "step": 5523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.867783410609215e-07,
+      "loss": 1.0986,
+      "step": 5524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.860386768077161e-07,
+      "loss": 1.1436,
+      "step": 5525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.852994171397683e-07,
+      "loss": 1.1138,
+      "step": 5526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.845605622133466e-07,
+      "loss": 1.0503,
+      "step": 5527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.838221121846363e-07,
+      "loss": 1.1201,
+      "step": 5528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.830840672097346e-07,
+      "loss": 1.1201,
+      "step": 5529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.823464274446542e-07,
+      "loss": 1.0977,
+      "step": 5530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.816091930453232e-07,
+      "loss": 1.1328,
+      "step": 5531
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.808723641675823e-07,
+      "loss": 1.123,
+      "step": 5532
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.801359409671858e-07,
+      "loss": 1.1084,
+      "step": 5533
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.793999235998055e-07,
+      "loss": 1.1357,
+      "step": 5534
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.786643122210234e-07,
+      "loss": 1.1191,
+      "step": 5535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.779291069863396e-07,
+      "loss": 1.0879,
+      "step": 5536
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.771943080511655e-07,
+      "loss": 1.1455,
+      "step": 5537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.764599155708267e-07,
+      "loss": 1.0293,
+      "step": 5538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.757259297005657e-07,
+      "loss": 1.0996,
+      "step": 5539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.749923505955352e-07,
+      "loss": 1.0913,
+      "step": 5540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.742591784108043e-07,
+      "loss": 1.0654,
+      "step": 5541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.735264133013568e-07,
+      "loss": 1.127,
+      "step": 5542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.727940554220885e-07,
+      "loss": 1.0542,
+      "step": 5543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.720621049278085e-07,
+      "loss": 1.1436,
+      "step": 5544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.713305619732431e-07,
+      "loss": 1.1201,
+      "step": 5545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.705994267130291e-07,
+      "loss": 1.1313,
+      "step": 5546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.698686993017197e-07,
+      "loss": 1.1035,
+      "step": 5547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.691383798937802e-07,
+      "loss": 1.1055,
+      "step": 5548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.684084686435889e-07,
+      "loss": 1.0879,
+      "step": 5549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.676789657054412e-07,
+      "loss": 1.0825,
+      "step": 5550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.669498712335419e-07,
+      "loss": 1.1113,
+      "step": 5551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.662211853820137e-07,
+      "loss": 1.0801,
+      "step": 5552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.654929083048891e-07,
+      "loss": 1.1279,
+      "step": 5553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.647650401561172e-07,
+      "loss": 1.1299,
+      "step": 5554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.640375810895581e-07,
+      "loss": 1.1152,
+      "step": 5555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.633105312589881e-07,
+      "loss": 1.1211,
+      "step": 5556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.625838908180944e-07,
+      "loss": 1.1094,
+      "step": 5557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.618576599204797e-07,
+      "loss": 1.1011,
+      "step": 5558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.611318387196593e-07,
+      "loss": 1.1338,
+      "step": 5559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.604064273690607e-07,
+      "loss": 1.1133,
+      "step": 5560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.596814260220274e-07,
+      "loss": 1.1074,
+      "step": 5561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.589568348318133e-07,
+      "loss": 1.0796,
+      "step": 5562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.58232653951589e-07,
+      "loss": 1.1357,
+      "step": 5563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.57508883534435e-07,
+      "loss": 1.1201,
+      "step": 5564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.567855237333461e-07,
+      "loss": 1.0889,
+      "step": 5565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.560625747012316e-07,
+      "loss": 1.126,
+      "step": 5566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.553400365909134e-07,
+      "loss": 1.0986,
+      "step": 5567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.546179095551249e-07,
+      "loss": 1.0723,
+      "step": 5568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.538961937465154e-07,
+      "loss": 1.105,
+      "step": 5569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.531748893176447e-07,
+      "loss": 1.125,
+      "step": 5570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.524539964209863e-07,
+      "loss": 1.1162,
+      "step": 5571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.517335152089287e-07,
+      "loss": 1.1328,
+      "step": 5572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.510134458337699e-07,
+      "loss": 1.0747,
+      "step": 5573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.502937884477244e-07,
+      "loss": 1.1074,
+      "step": 5574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.495745432029173e-07,
+      "loss": 1.0854,
+      "step": 5575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.488557102513862e-07,
+      "loss": 1.0752,
+      "step": 5576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.481372897450843e-07,
+      "loss": 1.0908,
+      "step": 5577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.474192818358742e-07,
+      "loss": 1.0977,
+      "step": 5578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.467016866755337e-07,
+      "loss": 1.1113,
+      "step": 5579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.459845044157536e-07,
+      "loss": 1.0742,
+      "step": 5580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.452677352081353e-07,
+      "loss": 1.0952,
+      "step": 5581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.445513792041937e-07,
+      "loss": 1.0986,
+      "step": 5582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.438354365553578e-07,
+      "loss": 1.0591,
+      "step": 5583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.431199074129667e-07,
+      "loss": 1.1201,
+      "step": 5584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.424047919282748e-07,
+      "loss": 1.1045,
+      "step": 5585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.416900902524475e-07,
+      "loss": 1.0918,
+      "step": 5586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.409758025365619e-07,
+      "loss": 1.1123,
+      "step": 5587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.402619289316102e-07,
+      "loss": 1.105,
+      "step": 5588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.395484695884945e-07,
+      "loss": 1.1016,
+      "step": 5589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.388354246580294e-07,
+      "loss": 1.1245,
+      "step": 5590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.381227942909451e-07,
+      "loss": 1.0986,
+      "step": 5591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.374105786378811e-07,
+      "loss": 1.125,
+      "step": 5592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.36698777849389e-07,
+      "loss": 1.1348,
+      "step": 5593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.359873920759354e-07,
+      "loss": 1.1377,
+      "step": 5594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.35276421467896e-07,
+      "loss": 1.1064,
+      "step": 5595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.345658661755615e-07,
+      "loss": 1.0767,
+      "step": 5596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.338557263491329e-07,
+      "loss": 1.1357,
+      "step": 5597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.331460021387236e-07,
+      "loss": 1.1104,
+      "step": 5598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.324366936943609e-07,
+      "loss": 1.0967,
+      "step": 5599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.317278011659824e-07,
+      "loss": 1.0796,
+      "step": 5600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.31019324703437e-07,
+      "loss": 1.1201,
+      "step": 5601
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.303112644564881e-07,
+      "loss": 1.1084,
+      "step": 5602
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.296036205748106e-07,
+      "loss": 1.0615,
+      "step": 5603
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.288963932079893e-07,
+      "loss": 1.0435,
+      "step": 5604
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.281895825055239e-07,
+      "loss": 1.1104,
+      "step": 5605
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.27483188616823e-07,
+      "loss": 1.1309,
+      "step": 5606
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.267772116912101e-07,
+      "loss": 1.0908,
+      "step": 5607
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.26071651877918e-07,
+      "loss": 1.0811,
+      "step": 5608
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.253665093260924e-07,
+      "loss": 1.1055,
+      "step": 5609
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.246617841847915e-07,
+      "loss": 1.1045,
+      "step": 5610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.239574766029842e-07,
+      "loss": 1.1084,
+      "step": 5611
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.232535867295508e-07,
+      "loss": 1.0781,
+      "step": 5612
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.225501147132852e-07,
+      "loss": 1.1211,
+      "step": 5613
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.218470607028906e-07,
+      "loss": 1.1387,
+      "step": 5614
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.211444248469838e-07,
+      "loss": 1.085,
+      "step": 5615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.204422072940926e-07,
+      "loss": 1.1279,
+      "step": 5616
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.197404081926554e-07,
+      "loss": 1.126,
+      "step": 5617
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.19039027691024e-07,
+      "loss": 1.0825,
+      "step": 5618
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.183380659374602e-07,
+      "loss": 1.105,
+      "step": 5619
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.176375230801364e-07,
+      "loss": 1.04,
+      "step": 5620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.169373992671398e-07,
+      "loss": 1.0825,
+      "step": 5621
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.162376946464662e-07,
+      "loss": 1.0903,
+      "step": 5622
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.155384093660229e-07,
+      "loss": 1.1143,
+      "step": 5623
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.148395435736303e-07,
+      "loss": 1.0928,
+      "step": 5624
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.141410974170183e-07,
+      "loss": 1.127,
+      "step": 5625
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.134430710438296e-07,
+      "loss": 1.123,
+      "step": 5626
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.127454646016164e-07,
+      "loss": 1.1162,
+      "step": 5627
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.120482782378438e-07,
+      "loss": 1.0791,
+      "step": 5628
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.113515120998882e-07,
+      "loss": 1.0977,
+      "step": 5629
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.106551663350357e-07,
+      "loss": 1.123,
+      "step": 5630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.099592410904836e-07,
+      "loss": 1.0684,
+      "step": 5631
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.092637365133424e-07,
+      "loss": 1.1406,
+      "step": 5632
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.085686527506315e-07,
+      "loss": 1.1045,
+      "step": 5633
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.078739899492816e-07,
+      "loss": 1.0723,
+      "step": 5634
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.071797482561358e-07,
+      "loss": 1.1387,
+      "step": 5635
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.064859278179465e-07,
+      "loss": 1.1455,
+      "step": 5636
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.057925287813789e-07,
+      "loss": 1.1289,
+      "step": 5637
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.050995512930073e-07,
+      "loss": 1.1045,
+      "step": 5638
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.04406995499317e-07,
+      "loss": 1.085,
+      "step": 5639
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.037148615467067e-07,
+      "loss": 1.0928,
+      "step": 5640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.03023149581483e-07,
+      "loss": 1.0967,
+      "step": 5641
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.023318597498635e-07,
+      "loss": 1.1338,
+      "step": 5642
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.016409921979795e-07,
+      "loss": 1.1172,
+      "step": 5643
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.009505470718695e-07,
+      "loss": 1.1133,
+      "step": 5644
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.002605245174838e-07,
+      "loss": 1.1069,
+      "step": 5645
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.995709246806848e-07,
+      "loss": 1.1523,
+      "step": 5646
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.988817477072436e-07,
+      "loss": 1.1543,
+      "step": 5647
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.981929937428439e-07,
+      "loss": 1.0898,
+      "step": 5648
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.975046629330782e-07,
+      "loss": 1.0791,
+      "step": 5649
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.968167554234493e-07,
+      "loss": 1.1182,
+      "step": 5650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.961292713593727e-07,
+      "loss": 1.1143,
+      "step": 5651
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.954422108861731e-07,
+      "loss": 1.1187,
+      "step": 5652
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.94755574149085e-07,
+      "loss": 1.0947,
+      "step": 5653
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.940693612932551e-07,
+      "loss": 1.1104,
+      "step": 5654
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.933835724637384e-07,
+      "loss": 1.125,
+      "step": 5655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.926982078055012e-07,
+      "loss": 1.1084,
+      "step": 5656
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.920132674634212e-07,
+      "loss": 1.0957,
+      "step": 5657
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.913287515822843e-07,
+      "loss": 1.1201,
+      "step": 5658
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.906446603067888e-07,
+      "loss": 1.1025,
+      "step": 5659
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.89960993781542e-07,
+      "loss": 1.1465,
+      "step": 5660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.892777521510608e-07,
+      "loss": 1.1367,
+      "step": 5661
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.885949355597743e-07,
+      "loss": 1.0703,
+      "step": 5662
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.879125441520194e-07,
+      "loss": 1.0967,
+      "step": 5663
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.872305780720452e-07,
+      "loss": 1.1152,
+      "step": 5664
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.865490374640103e-07,
+      "loss": 1.1143,
+      "step": 5665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.858679224719825e-07,
+      "loss": 1.1582,
+      "step": 5666
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.851872332399396e-07,
+      "loss": 1.0864,
+      "step": 5667
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.845069699117713e-07,
+      "loss": 1.0928,
+      "step": 5668
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.838271326312749e-07,
+      "loss": 1.1113,
+      "step": 5669
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.831477215421595e-07,
+      "loss": 1.124,
+      "step": 5670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.824687367880432e-07,
+      "loss": 1.1543,
+      "step": 5671
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.817901785124532e-07,
+      "loss": 1.0835,
+      "step": 5672
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.811120468588284e-07,
+      "loss": 1.1201,
+      "step": 5673
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.804343419705165e-07,
+      "loss": 1.0957,
+      "step": 5674
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.797570639907736e-07,
+      "loss": 1.063,
+      "step": 5675
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.790802130627695e-07,
+      "loss": 1.1255,
+      "step": 5676
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.784037893295798e-07,
+      "loss": 1.1309,
+      "step": 5677
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.777277929341905e-07,
+      "loss": 1.1318,
+      "step": 5678
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.770522240194999e-07,
+      "loss": 1.1377,
+      "step": 5679
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.7637708272831237e-07,
+      "loss": 1.1514,
+      "step": 5680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.757023692033444e-07,
+      "loss": 1.0938,
+      "step": 5681
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.750280835872212e-07,
+      "loss": 1.1133,
+      "step": 5682
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.7435422602247675e-07,
+      "loss": 1.0996,
+      "step": 5683
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.7368079665155666e-07,
+      "loss": 1.1182,
+      "step": 5684
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.730077956168136e-07,
+      "loss": 1.0693,
+      "step": 5685
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.723352230605105e-07,
+      "loss": 1.0762,
+      "step": 5686
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.7166307912482087e-07,
+      "loss": 1.1006,
+      "step": 5687
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.7099136395182684e-07,
+      "loss": 1.0923,
+      "step": 5688
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.70320077683519e-07,
+      "loss": 1.0488,
+      "step": 5689
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.696492204617992e-07,
+      "loss": 1.0581,
+      "step": 5690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.689787924284761e-07,
+      "loss": 1.0898,
+      "step": 5691
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.683087937252703e-07,
+      "loss": 1.127,
+      "step": 5692
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6763922449380995e-07,
+      "loss": 1.1279,
+      "step": 5693
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6697008487563227e-07,
+      "loss": 1.0898,
+      "step": 5694
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6630137501218504e-07,
+      "loss": 1.1084,
+      "step": 5695
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.65633095044824e-07,
+      "loss": 1.0928,
+      "step": 5696
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.64965245114814e-07,
+      "loss": 1.0967,
+      "step": 5697
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.642978253633301e-07,
+      "loss": 1.083,
+      "step": 5698
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6363083593145507e-07,
+      "loss": 1.1094,
+      "step": 5699
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6296427696018137e-07,
+      "loss": 1.1416,
+      "step": 5700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6229814859041183e-07,
+      "loss": 1.0996,
+      "step": 5701
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6163245096295494e-07,
+      "loss": 1.0693,
+      "step": 5702
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6096718421853146e-07,
+      "loss": 1.0938,
+      "step": 5703
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.603023484977695e-07,
+      "loss": 1.0947,
+      "step": 5704
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.596379439412049e-07,
+      "loss": 1.1436,
+      "step": 5705
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5897397068928565e-07,
+      "loss": 1.1265,
+      "step": 5706
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5831042888236586e-07,
+      "loss": 1.1113,
+      "step": 5707
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.576473186607081e-07,
+      "loss": 1.0864,
+      "step": 5708
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5698464016448644e-07,
+      "loss": 1.0732,
+      "step": 5709
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5632239353378073e-07,
+      "loss": 1.0483,
+      "step": 5710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.556605789085822e-07,
+      "loss": 1.1318,
+      "step": 5711
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.549991964287878e-07,
+      "loss": 1.0757,
+      "step": 5712
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.543382462342058e-07,
+      "loss": 1.0791,
+      "step": 5713
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.536777284645527e-07,
+      "loss": 1.1172,
+      "step": 5714
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5301764325945185e-07,
+      "loss": 1.0947,
+      "step": 5715
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5235799075843576e-07,
+      "loss": 1.1211,
+      "step": 5716
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.516987711009477e-07,
+      "loss": 1.1143,
+      "step": 5717
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5103998442633634e-07,
+      "loss": 1.0537,
+      "step": 5718
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5038163087385983e-07,
+      "loss": 1.0957,
+      "step": 5719
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.497237105826868e-07,
+      "loss": 1.1094,
+      "step": 5720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.490662236918905e-07,
+      "loss": 1.1309,
+      "step": 5721
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4840917034045667e-07,
+      "loss": 1.1021,
+      "step": 5722
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.477525506672764e-07,
+      "loss": 1.1172,
+      "step": 5723
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.47096364811149e-07,
+      "loss": 1.0928,
+      "step": 5724
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4644061291078564e-07,
+      "loss": 1.0957,
+      "step": 5725
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.457852951048022e-07,
+      "loss": 1.0811,
+      "step": 5726
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.451304115317229e-07,
+      "loss": 1.1191,
+      "step": 5727
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.444759623299827e-07,
+      "loss": 1.1157,
+      "step": 5728
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.438219476379227e-07,
+      "loss": 1.1436,
+      "step": 5729
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.43168367593792e-07,
+      "loss": 1.1289,
+      "step": 5730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.425152223357493e-07,
+      "loss": 1.1211,
+      "step": 5731
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4186251200185995e-07,
+      "loss": 1.1079,
+      "step": 5732
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.412102367300988e-07,
+      "loss": 1.1279,
+      "step": 5733
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.405583966583474e-07,
+      "loss": 1.084,
+      "step": 5734
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.3990699192439534e-07,
+      "loss": 1.1294,
+      "step": 5735
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.392560226659412e-07,
+      "loss": 1.1289,
+      "step": 5736
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.386054890205918e-07,
+      "loss": 1.1006,
+      "step": 5737
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.379553911258594e-07,
+      "loss": 1.063,
+      "step": 5738
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.373057291191674e-07,
+      "loss": 1.0825,
+      "step": 5739
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.3665650313784473e-07,
+      "loss": 1.0947,
+      "step": 5740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.3600771331912815e-07,
+      "loss": 1.1133,
+      "step": 5741
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.353593598001646e-07,
+      "loss": 1.0571,
+      "step": 5742
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3471144271800547e-07,
+      "loss": 1.1367,
+      "step": 5743
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3406396220961295e-07,
+      "loss": 1.0469,
+      "step": 5744
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3341691841185475e-07,
+      "loss": 1.1201,
+      "step": 5745
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.32770311461507e-07,
+      "loss": 1.0801,
+      "step": 5746
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3212414149525426e-07,
+      "loss": 1.1475,
+      "step": 5747
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3147840864968734e-07,
+      "loss": 1.103,
+      "step": 5748
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3083311306130524e-07,
+      "loss": 1.0859,
+      "step": 5749
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3018825486651596e-07,
+      "loss": 1.1221,
+      "step": 5750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.295438342016328e-07,
+      "loss": 1.0981,
+      "step": 5751
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2889985120287664e-07,
+      "loss": 1.1104,
+      "step": 5752
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2825630600637794e-07,
+      "loss": 1.1143,
+      "step": 5753
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.276131987481727e-07,
+      "loss": 1.1479,
+      "step": 5754
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2697052956420545e-07,
+      "loss": 1.0747,
+      "step": 5755
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2632829859032775e-07,
+      "loss": 1.1045,
+      "step": 5756
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.256865059622972e-07,
+      "loss": 1.0918,
+      "step": 5757
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2504515181578166e-07,
+      "loss": 1.0718,
+      "step": 5758
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.244042362863532e-07,
+      "loss": 1.1348,
+      "step": 5759
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.23763759509494e-07,
+      "loss": 1.1338,
+      "step": 5760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.231237216205908e-07,
+      "loss": 1.1055,
+      "step": 5761
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2248412275494006e-07,
+      "loss": 1.1289,
+      "step": 5762
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2184496304774283e-07,
+      "loss": 1.0859,
+      "step": 5763
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.212062426341101e-07,
+      "loss": 1.1191,
+      "step": 5764
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2056796164905766e-07,
+      "loss": 1.0688,
+      "step": 5765
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1993012022751023e-07,
+      "loss": 1.084,
+      "step": 5766
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1929271850429833e-07,
+      "loss": 1.0801,
+      "step": 5767
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.186557566141594e-07,
+      "loss": 1.1343,
+      "step": 5768
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.180192346917394e-07,
+      "loss": 1.1221,
+      "step": 5769
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1738315287158964e-07,
+      "loss": 1.0771,
+      "step": 5770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1674751128816997e-07,
+      "loss": 1.1279,
+      "step": 5771
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1611231007584603e-07,
+      "loss": 1.0825,
+      "step": 5772
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1547754936888953e-07,
+      "loss": 1.0889,
+      "step": 5773
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1484322930148146e-07,
+      "loss": 1.1138,
+      "step": 5774
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.142093500077085e-07,
+      "loss": 1.0894,
+      "step": 5775
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.135759116215629e-07,
+      "loss": 1.1553,
+      "step": 5776
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.129429142769467e-07,
+      "loss": 1.0703,
+      "step": 5777
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.123103581076654e-07,
+      "loss": 1.063,
+      "step": 5778
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1167824324743285e-07,
+      "loss": 1.1221,
+      "step": 5779
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1104656982987017e-07,
+      "loss": 1.1504,
+      "step": 5780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.104153379885034e-07,
+      "loss": 1.126,
+      "step": 5781
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0978454785676804e-07,
+      "loss": 1.0327,
+      "step": 5782
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0915419956800316e-07,
+      "loss": 1.0962,
+      "step": 5783
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.085242932554556e-07,
+      "loss": 1.0757,
+      "step": 5784
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.078948290522794e-07,
+      "loss": 1.1226,
+      "step": 5785
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.072658070915353e-07,
+      "loss": 1.0854,
+      "step": 5786
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.066372275061889e-07,
+      "loss": 1.1167,
+      "step": 5787
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.060090904291142e-07,
+      "loss": 1.1284,
+      "step": 5788
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.053813959930905e-07,
+      "loss": 1.124,
+      "step": 5789
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.047541443308031e-07,
+      "loss": 1.1348,
+      "step": 5790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0412733557484566e-07,
+      "loss": 1.1265,
+      "step": 5791
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0350096985771583e-07,
+      "loss": 1.083,
+      "step": 5792
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.028750473118198e-07,
+      "loss": 1.1851,
+      "step": 5793
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.022495680694685e-07,
+      "loss": 1.0918,
+      "step": 5794
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.016245322628792e-07,
+      "loss": 1.1191,
+      "step": 5795
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0099994002417707e-07,
+      "loss": 1.1309,
+      "step": 5796
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0037579148539093e-07,
+      "loss": 1.0947,
+      "step": 5797
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9975208677845823e-07,
+      "loss": 1.0962,
+      "step": 5798
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.991288260352222e-07,
+      "loss": 1.0713,
+      "step": 5799
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.985060093874307e-07,
+      "loss": 1.1035,
+      "step": 5800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9788363696673826e-07,
+      "loss": 1.1064,
+      "step": 5801
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9726170890470705e-07,
+      "loss": 1.0781,
+      "step": 5802
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9664022533280306e-07,
+      "loss": 1.0693,
+      "step": 5803
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9601918638240063e-07,
+      "loss": 1.1196,
+      "step": 5804
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9539859218477805e-07,
+      "loss": 1.1104,
+      "step": 5805
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9477844287112013e-07,
+      "loss": 1.1035,
+      "step": 5806
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.94158738572519e-07,
+      "loss": 1.1201,
+      "step": 5807
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9353947941997125e-07,
+      "loss": 1.1318,
+      "step": 5808
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9292066554437847e-07,
+      "loss": 1.1323,
+      "step": 5809
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9230229707655167e-07,
+      "loss": 1.0645,
+      "step": 5810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9168437414720454e-07,
+      "loss": 1.0864,
+      "step": 5811
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9106689688695714e-07,
+      "loss": 1.0986,
+      "step": 5812
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.904498654263364e-07,
+      "loss": 1.1719,
+      "step": 5813
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8983327989577406e-07,
+      "loss": 1.0718,
+      "step": 5814
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.892171404256073e-07,
+      "loss": 1.1016,
+      "step": 5815
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8860144714608055e-07,
+      "loss": 1.1191,
+      "step": 5816
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.879862001873422e-07,
+      "loss": 1.0879,
+      "step": 5817
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.873713996794476e-07,
+      "loss": 1.0986,
+      "step": 5818
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.86757045752357e-07,
+      "loss": 1.124,
+      "step": 5819
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.86143138535936e-07,
+      "loss": 1.0801,
+      "step": 5820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.855296781599563e-07,
+      "loss": 1.1357,
+      "step": 5821
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8491666475409586e-07,
+      "loss": 1.0522,
+      "step": 5822
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8430409844793597e-07,
+      "loss": 1.1445,
+      "step": 5823
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.836919793709665e-07,
+      "loss": 1.1055,
+      "step": 5824
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.830803076525799e-07,
+      "loss": 1.1172,
+      "step": 5825
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8246908342207484e-07,
+      "loss": 1.1221,
+      "step": 5826
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8185830680865675e-07,
+      "loss": 1.1328,
+      "step": 5827
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8124797794143463e-07,
+      "loss": 1.1338,
+      "step": 5828
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.806380969494247e-07,
+      "loss": 1.0903,
+      "step": 5829
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8002866396154663e-07,
+      "loss": 1.0967,
+      "step": 5830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.79419679106626e-07,
+      "loss": 1.0898,
+      "step": 5831
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7881114251339504e-07,
+      "loss": 1.1416,
+      "step": 5832
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.782030543104884e-07,
+      "loss": 1.1646,
+      "step": 5833
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7759541462644877e-07,
+      "loss": 1.1445,
+      "step": 5834
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.769882235897232e-07,
+      "loss": 1.1108,
+      "step": 5835
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.763814813286629e-07,
+      "loss": 1.064,
+      "step": 5836
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.757751879715246e-07,
+      "loss": 1.1143,
+      "step": 5837
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.751693436464712e-07,
+      "loss": 1.0698,
+      "step": 5838
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.745639484815691e-07,
+      "loss": 1.1011,
+      "step": 5839
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7395900260479177e-07,
+      "loss": 1.1279,
+      "step": 5840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7335450614401554e-07,
+      "loss": 1.1104,
+      "step": 5841
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.727504592270223e-07,
+      "loss": 1.0981,
+      "step": 5842
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.721468619815002e-07,
+      "loss": 1.0859,
+      "step": 5843
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.715437145350409e-07,
+      "loss": 1.1201,
+      "step": 5844
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.709410170151423e-07,
+      "loss": 1.103,
+      "step": 5845
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.703387695492056e-07,
+      "loss": 1.0879,
+      "step": 5846
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.69736972264538e-07,
+      "loss": 1.1084,
+      "step": 5847
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.69135625288351e-07,
+      "loss": 1.0947,
+      "step": 5848
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6853472874776167e-07,
+      "loss": 1.0903,
+      "step": 5849
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6793428276979043e-07,
+      "loss": 1.0928,
+      "step": 5850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6733428748136457e-07,
+      "loss": 1.1045,
+      "step": 5851
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6673474300931403e-07,
+      "loss": 1.0796,
+      "step": 5852
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6613564948037416e-07,
+      "loss": 1.0586,
+      "step": 5853
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6553700702118585e-07,
+      "loss": 1.1084,
+      "step": 5854
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.649388157582931e-07,
+      "loss": 1.0977,
+      "step": 5855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.643410758181465e-07,
+      "loss": 1.1016,
+      "step": 5856
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.637437873270994e-07,
+      "loss": 1.1211,
+      "step": 5857
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6314695041140964e-07,
+      "loss": 1.0908,
+      "step": 5858
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.625505651972414e-07,
+      "loss": 1.1055,
+      "step": 5859
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6195463181066257e-07,
+      "loss": 1.0898,
+      "step": 5860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.613591503776445e-07,
+      "loss": 1.1016,
+      "step": 5861
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.607641210240645e-07,
+      "loss": 1.1104,
+      "step": 5862
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.601695438757036e-07,
+      "loss": 1.1133,
+      "step": 5863
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.595754190582465e-07,
+      "loss": 1.1455,
+      "step": 5864
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.589817466972842e-07,
+      "loss": 1.0664,
+      "step": 5865
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5838852691830963e-07,
+      "loss": 1.0635,
+      "step": 5866
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5779575984672273e-07,
+      "loss": 1.1035,
+      "step": 5867
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5720344560782576e-07,
+      "loss": 1.0977,
+      "step": 5868
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.56611584326825e-07,
+      "loss": 1.0977,
+      "step": 5869
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.560201761288326e-07,
+      "loss": 1.1104,
+      "step": 5870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.554292211388649e-07,
+      "loss": 1.0664,
+      "step": 5871
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5483871948184034e-07,
+      "loss": 1.082,
+      "step": 5872
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.542486712825838e-07,
+      "loss": 1.104,
+      "step": 5873
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5365907666582347e-07,
+      "loss": 1.1494,
+      "step": 5874
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.530699357561906e-07,
+      "loss": 1.0674,
+      "step": 5875
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.524812486782228e-07,
+      "loss": 1.1006,
+      "step": 5876
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.518930155563591e-07,
+      "loss": 1.0811,
+      "step": 5877
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5130523651494554e-07,
+      "loss": 1.0645,
+      "step": 5878
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5071791167822944e-07,
+      "loss": 1.1406,
+      "step": 5879
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.501310411703632e-07,
+      "loss": 1.1206,
+      "step": 5880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.495446251154039e-07,
+      "loss": 1.1191,
+      "step": 5881
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.48958663637311e-07,
+      "loss": 1.1279,
+      "step": 5882
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.483731568599491e-07,
+      "loss": 1.0957,
+      "step": 5883
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.477881049070875e-07,
+      "loss": 1.1074,
+      "step": 5884
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4720350790239666e-07,
+      "loss": 1.1416,
+      "step": 5885
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4661936596945276e-07,
+      "loss": 1.1016,
+      "step": 5886
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.46035679231736e-07,
+      "loss": 1.0352,
+      "step": 5887
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4545244781262863e-07,
+      "loss": 1.0859,
+      "step": 5888
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.448696718354194e-07,
+      "loss": 1.1006,
+      "step": 5889
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4428735142329804e-07,
+      "loss": 1.1152,
+      "step": 5890
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.437054866993589e-07,
+      "loss": 1.1064,
+      "step": 5891
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4312407778660126e-07,
+      "loss": 1.1016,
+      "step": 5892
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.425431248079264e-07,
+      "loss": 1.1377,
+      "step": 5893
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4196262788613906e-07,
+      "loss": 1.1094,
+      "step": 5894
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4138258714394993e-07,
+      "loss": 1.1289,
+      "step": 5895
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.408030027039713e-07,
+      "loss": 1.0806,
+      "step": 5896
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.402238746887182e-07,
+      "loss": 1.0776,
+      "step": 5897
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3964520322061213e-07,
+      "loss": 1.1123,
+      "step": 5898
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3906698842197467e-07,
+      "loss": 1.1035,
+      "step": 5899
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.384892304150339e-07,
+      "loss": 1.1123,
+      "step": 5900
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3791192932191905e-07,
+      "loss": 1.1045,
+      "step": 5901
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3733508526466395e-07,
+      "loss": 1.1387,
+      "step": 5902
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3675869836520577e-07,
+      "loss": 1.0928,
+      "step": 5903
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.361827687453847e-07,
+      "loss": 1.1162,
+      "step": 5904
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3560729652694387e-07,
+      "loss": 1.123,
+      "step": 5905
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3503228183153105e-07,
+      "loss": 1.0996,
+      "step": 5906
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.344577247806957e-07,
+      "loss": 1.1001,
+      "step": 5907
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.338836254958916e-07,
+      "loss": 1.1104,
+      "step": 5908
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.333099840984763e-07,
+      "loss": 1.1084,
+      "step": 5909
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.327368007097084e-07,
+      "loss": 1.085,
+      "step": 5910
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.321640754507524e-07,
+      "loss": 1.0908,
+      "step": 5911
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3159180844267394e-07,
+      "loss": 1.0967,
+      "step": 5912
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.31019999806442e-07,
+      "loss": 1.1162,
+      "step": 5913
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.304486496629297e-07,
+      "loss": 1.1221,
+      "step": 5914
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2987775813291266e-07,
+      "loss": 1.0957,
+      "step": 5915
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.293073253370688e-07,
+      "loss": 1.1143,
+      "step": 5916
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.28737351395981e-07,
+      "loss": 1.083,
+      "step": 5917
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2816783643013243e-07,
+      "loss": 1.0762,
+      "step": 5918
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.275987805599121e-07,
+      "loss": 1.1309,
+      "step": 5919
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.270301839056106e-07,
+      "loss": 1.1191,
+      "step": 5920
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2646204658742046e-07,
+      "loss": 1.1064,
+      "step": 5921
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.258943687254393e-07,
+      "loss": 1.1123,
+      "step": 5922
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.25327150439666e-07,
+      "loss": 1.0742,
+      "step": 5923
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2476039185000195e-07,
+      "loss": 1.1533,
+      "step": 5924
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2419409307625384e-07,
+      "loss": 1.1055,
+      "step": 5925
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2362825423812834e-07,
+      "loss": 1.1377,
+      "step": 5926
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.230628754552356e-07,
+      "loss": 1.0884,
+      "step": 5927
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.224979568470901e-07,
+      "loss": 1.1006,
+      "step": 5928
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2193349853310714e-07,
+      "loss": 1.1338,
+      "step": 5929
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.213695006326062e-07,
+      "loss": 1.0869,
+      "step": 5930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.208059632648078e-07,
+      "loss": 1.1284,
+      "step": 5931
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2024288654883664e-07,
+      "loss": 1.0864,
+      "step": 5932
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.196802706037197e-07,
+      "loss": 1.103,
+      "step": 5933
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.191181155483861e-07,
+      "loss": 1.1055,
+      "step": 5934
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1855642150166707e-07,
+      "loss": 1.1523,
+      "step": 5935
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1799518858229796e-07,
+      "loss": 1.1338,
+      "step": 5936
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1743441690891525e-07,
+      "loss": 1.1602,
+      "step": 5937
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.16874106600058e-07,
+      "loss": 1.1523,
+      "step": 5938
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.163142577741693e-07,
+      "loss": 1.0986,
+      "step": 5939
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.157548705495922e-07,
+      "loss": 1.0957,
+      "step": 5940
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.151959450445749e-07,
+      "loss": 1.0669,
+      "step": 5941
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.146374813772657e-07,
+      "loss": 1.0894,
+      "step": 5942
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.140794796657154e-07,
+      "loss": 1.0601,
+      "step": 5943
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.135219400278802e-07,
+      "loss": 1.0947,
+      "step": 5944
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.129648625816148e-07,
+      "loss": 1.0703,
+      "step": 5945
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.124082474446774e-07,
+      "loss": 1.127,
+      "step": 5946
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.118520947347303e-07,
+      "loss": 1.0913,
+      "step": 5947
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1129640456933565e-07,
+      "loss": 1.0889,
+      "step": 5948
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1074117706595827e-07,
+      "loss": 1.1055,
+      "step": 5949
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.101864123419668e-07,
+      "loss": 1.0811,
+      "step": 5950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.0963211051462967e-07,
+      "loss": 1.0693,
+      "step": 5951
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.090782717011201e-07,
+      "loss": 1.1006,
+      "step": 5952
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.085248960185111e-07,
+      "loss": 1.1001,
+      "step": 5953
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0797198358377857e-07,
+      "loss": 1.1045,
+      "step": 5954
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0741953451380117e-07,
+      "loss": 1.1211,
+      "step": 5955
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0686754892535945e-07,
+      "loss": 1.1357,
+      "step": 5956
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0631602693513453e-07,
+      "loss": 1.1348,
+      "step": 5957
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.057649686597117e-07,
+      "loss": 1.0703,
+      "step": 5958
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0521437421557635e-07,
+      "loss": 1.0596,
+      "step": 5959
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.046642437191169e-07,
+      "loss": 1.1299,
+      "step": 5960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.041145772866236e-07,
+      "loss": 1.085,
+      "step": 5961
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.035653750342879e-07,
+      "loss": 1.1079,
+      "step": 5962
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.030166370782045e-07,
+      "loss": 1.0933,
+      "step": 5963
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.024683635343684e-07,
+      "loss": 1.1357,
+      "step": 5964
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0192055451867707e-07,
+      "loss": 1.0635,
+      "step": 5965
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.013732101469305e-07,
+      "loss": 1.1094,
+      "step": 5966
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.008263305348294e-07,
+      "loss": 1.0938,
+      "step": 5967
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0027991579797576e-07,
+      "loss": 1.1211,
+      "step": 5968
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9973396605187583e-07,
+      "loss": 1.1055,
+      "step": 5969
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9918848141193527e-07,
+      "loss": 1.0625,
+      "step": 5970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.986434619934617e-07,
+      "loss": 1.125,
+      "step": 5971
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.980989079116653e-07,
+      "loss": 1.0771,
+      "step": 5972
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9755481928165685e-07,
+      "loss": 1.0947,
+      "step": 5973
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9701119621844994e-07,
+      "loss": 1.0845,
+      "step": 5974
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9646803883695843e-07,
+      "loss": 1.0977,
+      "step": 5975
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.959253472519982e-07,
+      "loss": 1.0767,
+      "step": 5976
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9538312157828764e-07,
+      "loss": 1.0845,
+      "step": 5977
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.948413619304455e-07,
+      "loss": 1.082,
+      "step": 5978
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.943000684229916e-07,
+      "loss": 1.1299,
+      "step": 5979
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.937592411703488e-07,
+      "loss": 1.0581,
+      "step": 5980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9321888028684076e-07,
+      "loss": 1.0737,
+      "step": 5981
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.926789858866916e-07,
+      "loss": 1.1187,
+      "step": 5982
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.921395580840286e-07,
+      "loss": 1.082,
+      "step": 5983
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.916005969928784e-07,
+      "loss": 1.0884,
+      "step": 5984
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9106210272717124e-07,
+      "loss": 1.0864,
+      "step": 5985
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.905240754007363e-07,
+      "loss": 1.1367,
+      "step": 5986
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8998651512730543e-07,
+      "loss": 1.1074,
+      "step": 5987
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8944942202051186e-07,
+      "loss": 1.0947,
+      "step": 5988
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.889127961938898e-07,
+      "loss": 1.0869,
+      "step": 5989
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8837663776087405e-07,
+      "loss": 1.1387,
+      "step": 5990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8784094683480174e-07,
+      "loss": 1.1436,
+      "step": 5991
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.873057235289098e-07,
+      "loss": 1.0718,
+      "step": 5992
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.86770967956338e-07,
+      "loss": 1.0898,
+      "step": 5993
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.862366802301261e-07,
+      "loss": 1.145,
+      "step": 5994
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.85702860463215e-07,
+      "loss": 1.1455,
+      "step": 5995
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8516950876844753e-07,
+      "loss": 1.0601,
+      "step": 5996
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.846366252585667e-07,
+      "loss": 1.1055,
+      "step": 5997
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8410421004621574e-07,
+      "loss": 1.1084,
+      "step": 5998
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.835722632439414e-07,
+      "loss": 1.1455,
+      "step": 5999
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8304078496418947e-07,
+      "loss": 1.0576,
+      "step": 6000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.825097753193065e-07,
+      "loss": 1.1206,
+      "step": 6001
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8197923442154163e-07,
+      "loss": 1.1021,
+      "step": 6002
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.814491623830432e-07,
+      "loss": 1.0264,
+      "step": 6003
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8091955931586144e-07,
+      "loss": 1.1191,
+      "step": 6004
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8039042533194813e-07,
+      "loss": 1.124,
+      "step": 6005
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7986176054315336e-07,
+      "loss": 1.1011,
+      "step": 6006
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7933356506123113e-07,
+      "loss": 1.1074,
+      "step": 6007
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7880583899783375e-07,
+      "loss": 1.124,
+      "step": 6008
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.782785824645154e-07,
+      "loss": 1.0557,
+      "step": 6009
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7775179557273155e-07,
+      "loss": 1.1035,
+      "step": 6010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.772254784338371e-07,
+      "loss": 1.1035,
+      "step": 6011
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7669963115908805e-07,
+      "loss": 1.0889,
+      "step": 6012
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7617425385964234e-07,
+      "loss": 1.0977,
+      "step": 6013
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.756493466465565e-07,
+      "loss": 1.0757,
+      "step": 6014
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.751249096307895e-07,
+      "loss": 1.0791,
+      "step": 6015
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7460094292319965e-07,
+      "loss": 1.1377,
+      "step": 6016
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.740774466345464e-07,
+      "loss": 1.0898,
+      "step": 6017
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7355442087549035e-07,
+      "loss": 1.126,
+      "step": 6018
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.730318657565917e-07,
+      "loss": 1.0967,
+      "step": 6019
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.725097813883104e-07,
+      "loss": 1.1133,
+      "step": 6020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.719881678810096e-07,
+      "loss": 1.0874,
+      "step": 6021
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7146702534495056e-07,
+      "loss": 1.0576,
+      "step": 6022
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7094635389029513e-07,
+      "loss": 1.1377,
+      "step": 6023
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7042615362710687e-07,
+      "loss": 1.1426,
+      "step": 6024
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.699064246653485e-07,
+      "loss": 1.0723,
+      "step": 6025
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.693871671148843e-07,
+      "loss": 1.0732,
+      "step": 6026
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6886838108547754e-07,
+      "loss": 1.1152,
+      "step": 6027
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.683500666867919e-07,
+      "loss": 1.0796,
+      "step": 6028
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.67832224028394e-07,
+      "loss": 1.127,
+      "step": 6029
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6731485321974724e-07,
+      "loss": 1.0815,
+      "step": 6030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.667979543702168e-07,
+      "loss": 1.0801,
+      "step": 6031
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6628152758906843e-07,
+      "loss": 1.1221,
+      "step": 6032
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6576557298546773e-07,
+      "loss": 1.0708,
+      "step": 6033
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.652500906684796e-07,
+      "loss": 1.0957,
+      "step": 6034
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.647350807470711e-07,
+      "loss": 1.0557,
+      "step": 6035
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6422054333010703e-07,
+      "loss": 1.084,
+      "step": 6036
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.637064785263549e-07,
+      "loss": 1.0996,
+      "step": 6037
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.631928864444805e-07,
+      "loss": 1.1172,
+      "step": 6038
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6267976719304947e-07,
+      "loss": 1.166,
+      "step": 6039
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.621671208805293e-07,
+      "loss": 1.1084,
+      "step": 6040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.616549476152852e-07,
+      "loss": 1.1191,
+      "step": 6041
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6114324750558443e-07,
+      "loss": 1.126,
+      "step": 6042
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6063202065959387e-07,
+      "loss": 1.0688,
+      "step": 6043
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.601212671853792e-07,
+      "loss": 1.126,
+      "step": 6044
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.596109871909061e-07,
+      "loss": 1.1187,
+      "step": 6045
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.591011807840424e-07,
+      "loss": 1.0317,
+      "step": 6046
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.585918480725524e-07,
+      "loss": 1.1562,
+      "step": 6047
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5808298916410346e-07,
+      "loss": 1.1318,
+      "step": 6048
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5757460416626105e-07,
+      "loss": 1.0898,
+      "step": 6049
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.570666931864901e-07,
+      "loss": 1.1001,
+      "step": 6050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.56559256332157e-07,
+      "loss": 1.125,
+      "step": 6051
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5605229371052615e-07,
+      "loss": 1.1338,
+      "step": 6052
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5554580542876295e-07,
+      "loss": 1.1074,
+      "step": 6053
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.550397915939326e-07,
+      "loss": 1.1074,
+      "step": 6054
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.545342523129987e-07,
+      "loss": 1.1191,
+      "step": 6055
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5402918769282554e-07,
+      "loss": 1.0894,
+      "step": 6056
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5352459784017707e-07,
+      "loss": 1.0986,
+      "step": 6057
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.530204828617164e-07,
+      "loss": 1.085,
+      "step": 6058
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.525168428640071e-07,
+      "loss": 1.0933,
+      "step": 6059
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.520136779535112e-07,
+      "loss": 1.0806,
+      "step": 6060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5151098823659096e-07,
+      "loss": 1.0845,
+      "step": 6061
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.510087738195086e-07,
+      "loss": 1.1338,
+      "step": 6062
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.505070348084243e-07,
+      "loss": 1.0918,
+      "step": 6063
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.500057713094001e-07,
+      "loss": 1.083,
+      "step": 6064
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.495049834283955e-07,
+      "loss": 1.0957,
+      "step": 6065
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.490046712712707e-07,
+      "loss": 1.1045,
+      "step": 6066
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4850483494378406e-07,
+      "loss": 1.1191,
+      "step": 6067
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.48005474551595e-07,
+      "loss": 1.1133,
+      "step": 6068
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4750659020026076e-07,
+      "loss": 1.082,
+      "step": 6069
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.470081819952397e-07,
+      "loss": 1.1108,
+      "step": 6070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4651025004188753e-07,
+      "loss": 1.1465,
+      "step": 6071
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4601279444546044e-07,
+      "loss": 1.083,
+      "step": 6072
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4551581531111393e-07,
+      "loss": 1.1084,
+      "step": 6073
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.450193127439024e-07,
+      "loss": 1.0869,
+      "step": 6074
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4452328684878014e-07,
+      "loss": 1.0913,
+      "step": 6075
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4402773773059995e-07,
+      "loss": 1.1328,
+      "step": 6076
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4353266549411364e-07,
+      "loss": 1.1123,
+      "step": 6077
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.430380702439733e-07,
+      "loss": 1.1787,
+      "step": 6078
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4254395208473e-07,
+      "loss": 1.1377,
+      "step": 6079
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4205031112083247e-07,
+      "loss": 1.0996,
+      "step": 6080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4155714745663083e-07,
+      "loss": 1.1514,
+      "step": 6081
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.410644611963725e-07,
+      "loss": 1.0591,
+      "step": 6082
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.405722524442042e-07,
+      "loss": 1.1289,
+      "step": 6083
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.400805213041732e-07,
+      "loss": 1.083,
+      "step": 6084
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3958926788022365e-07,
+      "loss": 1.084,
+      "step": 6085
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3909849227620077e-07,
+      "loss": 1.1113,
+      "step": 6086
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.386081945958474e-07,
+      "loss": 1.0991,
+      "step": 6087
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.381183749428051e-07,
+      "loss": 1.1523,
+      "step": 6088
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3762903342061579e-07,
+      "loss": 1.083,
+      "step": 6089
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3714017013271974e-07,
+      "loss": 1.1445,
+      "step": 6090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.366517851824554e-07,
+      "loss": 1.1328,
+      "step": 6091
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3616387867306162e-07,
+      "loss": 1.0771,
+      "step": 6092
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.356764507076742e-07,
+      "loss": 1.0601,
+      "step": 6093
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.351895013893288e-07,
+      "loss": 1.1201,
+      "step": 6094
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3470303082096025e-07,
+      "loss": 1.0928,
+      "step": 6095
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3421703910540133e-07,
+      "loss": 1.0781,
+      "step": 6096
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3373152634538488e-07,
+      "loss": 1.0791,
+      "step": 6097
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3324649264354126e-07,
+      "loss": 1.082,
+      "step": 6098
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3276193810239905e-07,
+      "loss": 1.0684,
+      "step": 6099
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3227786282438796e-07,
+      "loss": 1.1299,
+      "step": 6100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3179426691183394e-07,
+      "loss": 1.1484,
+      "step": 6101
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3131115046696213e-07,
+      "loss": 1.1172,
+      "step": 6102
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3082851359189806e-07,
+      "loss": 1.0659,
+      "step": 6103
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3034635638866404e-07,
+      "loss": 1.0815,
+      "step": 6104
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2986467895918074e-07,
+      "loss": 1.0669,
+      "step": 6105
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2938348140526922e-07,
+      "loss": 1.1201,
+      "step": 6106
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2890276382864784e-07,
+      "loss": 1.1045,
+      "step": 6107
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2842252633093308e-07,
+      "loss": 1.0981,
+      "step": 6108
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2794276901364125e-07,
+      "loss": 1.1406,
+      "step": 6109
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2746349197818624e-07,
+      "loss": 1.126,
+      "step": 6110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2698469532588086e-07,
+      "loss": 1.0547,
+      "step": 6111
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.265063791579361e-07,
+      "loss": 1.1074,
+      "step": 6112
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2602854357546085e-07,
+      "loss": 1.1172,
+      "step": 6113
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.255511886794634e-07,
+      "loss": 1.0742,
+      "step": 6114
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.250743145708509e-07,
+      "loss": 1.0439,
+      "step": 6115
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2459792135042678e-07,
+      "loss": 1.0723,
+      "step": 6116
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2412200911889503e-07,
+      "loss": 1.0801,
+      "step": 6117
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2364657797685651e-07,
+      "loss": 1.0776,
+      "step": 6118
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2317162802481058e-07,
+      "loss": 1.0864,
+      "step": 6119
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.226971593631558e-07,
+      "loss": 1.0986,
+      "step": 6120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2222317209218774e-07,
+      "loss": 1.1035,
+      "step": 6121
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2174966631210132e-07,
+      "loss": 1.0791,
+      "step": 6122
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.212766421229892e-07,
+      "loss": 1.1211,
+      "step": 6123
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2080409962484145e-07,
+      "loss": 1.1265,
+      "step": 6124
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2033203891754812e-07,
+      "loss": 1.0908,
+      "step": 6125
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.198604601008958e-07,
+      "loss": 1.1309,
+      "step": 6126
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1938936327456973e-07,
+      "loss": 1.1406,
+      "step": 6127
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1891874853815388e-07,
+      "loss": 1.0986,
+      "step": 6128
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1844861599112948e-07,
+      "loss": 1.0605,
+      "step": 6129
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.179789657328757e-07,
+      "loss": 1.1113,
+      "step": 6130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1750979786267113e-07,
+      "loss": 1.0679,
+      "step": 6131
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1704111247969006e-07,
+      "loss": 1.1172,
+      "step": 6132
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.165729096830077e-07,
+      "loss": 1.1426,
+      "step": 6133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1610518957159516e-07,
+      "loss": 1.0869,
+      "step": 6134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1563795224432139e-07,
+      "loss": 1.1035,
+      "step": 6135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1517119779995487e-07,
+      "loss": 1.0718,
+      "step": 6136
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1470492633716034e-07,
+      "loss": 1.127,
+      "step": 6137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.142391379545017e-07,
+      "loss": 1.0786,
+      "step": 6138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1377383275044045e-07,
+      "loss": 1.0454,
+      "step": 6139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1330901082333572e-07,
+      "loss": 1.061,
+      "step": 6140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1284467227144362e-07,
+      "loss": 1.1055,
+      "step": 6141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.123808171929201e-07,
+      "loss": 1.1172,
+      "step": 6142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1191744568581696e-07,
+      "loss": 1.1289,
+      "step": 6143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1145455784808532e-07,
+      "loss": 1.1201,
+      "step": 6144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1099215377757303e-07,
+      "loss": 1.1025,
+      "step": 6145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1053023357202524e-07,
+      "loss": 1.1758,
+      "step": 6146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1006879732908686e-07,
+      "loss": 1.1362,
+      "step": 6147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0960784514629823e-07,
+      "loss": 1.1465,
+      "step": 6148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0914737712109917e-07,
+      "loss": 1.0708,
+      "step": 6149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0868739335082516e-07,
+      "loss": 1.1357,
+      "step": 6150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0822789393271175e-07,
+      "loss": 1.1211,
+      "step": 6151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0776887896388987e-07,
+      "loss": 1.0908,
+      "step": 6152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0731034854138964e-07,
+      "loss": 1.1309,
+      "step": 6153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0685230276213774e-07,
+      "loss": 1.1104,
+      "step": 6154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.063947417229592e-07,
+      "loss": 1.1172,
+      "step": 6155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.059376655205758e-07,
+      "loss": 1.1201,
+      "step": 6156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0548107425160697e-07,
+      "loss": 1.0742,
+      "step": 6157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0502496801257072e-07,
+      "loss": 1.1353,
+      "step": 6158
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0456934689988078e-07,
+      "loss": 1.1025,
+      "step": 6159
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0411421100984985e-07,
+      "loss": 1.1309,
+      "step": 6160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0365956043868708e-07,
+      "loss": 1.1348,
+      "step": 6161
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0320539528249918e-07,
+      "loss": 1.0532,
+      "step": 6162
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0275171563729074e-07,
+      "loss": 1.1211,
+      "step": 6163
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.022985215989637e-07,
+      "loss": 1.0415,
+      "step": 6164
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0184581326331666e-07,
+      "loss": 1.126,
+      "step": 6165
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0139359072604642e-07,
+      "loss": 1.0996,
+      "step": 6166
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0094185408274652e-07,
+      "loss": 1.083,
+      "step": 6167
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0049060342890697e-07,
+      "loss": 1.1064,
+      "step": 6168
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0003983885991757e-07,
+      "loss": 1.1265,
+      "step": 6169
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9958956047106236e-07,
+      "loss": 1.0908,
+      "step": 6170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9913976835752525e-07,
+      "loss": 1.1025,
+      "step": 6171
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.986904626143854e-07,
+      "loss": 1.1133,
+      "step": 6172
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9824164333661993e-07,
+      "loss": 1.0742,
+      "step": 6173
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9779331061910323e-07,
+      "loss": 1.1123,
+      "step": 6174
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.973454645566067e-07,
+      "loss": 1.1104,
+      "step": 6175
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9689810524379848e-07,
+      "loss": 1.0791,
+      "step": 6176
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9645123277524546e-07,
+      "loss": 1.0757,
+      "step": 6177
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.960048472454093e-07,
+      "loss": 1.1211,
+      "step": 6178
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.955589487486495e-07,
+      "loss": 1.0996,
+      "step": 6179
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.95113537379224e-07,
+      "loss": 1.0566,
+      "step": 6180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9466861323128582e-07,
+      "loss": 1.1143,
+      "step": 6181
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.942241763988864e-07,
+      "loss": 1.1108,
+      "step": 6182
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9378022697597336e-07,
+      "loss": 1.1113,
+      "step": 6183
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9333676505639109e-07,
+      "loss": 1.0957,
+      "step": 6184
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.928937907338821e-07,
+      "loss": 1.1162,
+      "step": 6185
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9245130410208478e-07,
+      "loss": 1.0718,
+      "step": 6186
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.920093052545341e-07,
+      "loss": 1.0674,
+      "step": 6187
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9156779428466414e-07,
+      "loss": 1.1279,
+      "step": 6188
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9112677128580337e-07,
+      "loss": 1.0752,
+      "step": 6189
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9068623635117772e-07,
+      "loss": 1.0918,
+      "step": 6190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9024618957391077e-07,
+      "loss": 1.1377,
+      "step": 6191
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8980663104702201e-07,
+      "loss": 1.1738,
+      "step": 6192
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.893675608634285e-07,
+      "loss": 1.0938,
+      "step": 6193
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8892897911594376e-07,
+      "loss": 1.0513,
+      "step": 6194
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8849088589727698e-07,
+      "loss": 1.1216,
+      "step": 6195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.880532813000363e-07,
+      "loss": 1.0908,
+      "step": 6196
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8761616541672496e-07,
+      "loss": 1.1318,
+      "step": 6197
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8717953833974262e-07,
+      "loss": 1.1006,
+      "step": 6198
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8674340016138687e-07,
+      "loss": 1.0557,
+      "step": 6199
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8630775097385172e-07,
+      "loss": 1.1396,
+      "step": 6200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.858725908692266e-07,
+      "loss": 1.0928,
+      "step": 6201
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8543791993949927e-07,
+      "loss": 1.1338,
+      "step": 6202
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8500373827655267e-07,
+      "loss": 1.0811,
+      "step": 6203
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8457004597216695e-07,
+      "loss": 1.1089,
+      "step": 6204
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8413684311801905e-07,
+      "loss": 1.0947,
+      "step": 6205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.837041298056813e-07,
+      "loss": 1.1172,
+      "step": 6206
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8327190612662444e-07,
+      "loss": 1.0996,
+      "step": 6207
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8284017217221395e-07,
+      "loss": 1.1123,
+      "step": 6208
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8240892803371268e-07,
+      "loss": 1.0825,
+      "step": 6209
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8197817380227968e-07,
+      "loss": 1.1196,
+      "step": 6210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.815479095689704e-07,
+      "loss": 1.1396,
+      "step": 6211
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.811181354247371e-07,
+      "loss": 1.0923,
+      "step": 6212
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8068885146042824e-07,
+      "loss": 1.124,
+      "step": 6213
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8026005776678817e-07,
+      "loss": 1.0757,
+      "step": 6214
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7983175443445855e-07,
+      "loss": 1.063,
+      "step": 6215
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.794039415539764e-07,
+      "loss": 1.1025,
+      "step": 6216
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7897661921577547e-07,
+      "loss": 1.1602,
+      "step": 6217
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.785497875101863e-07,
+      "loss": 1.0791,
+      "step": 6218
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7812344652743507e-07,
+      "loss": 1.1182,
+      "step": 6219
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7769759635764383e-07,
+      "loss": 1.1416,
+      "step": 6220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7727223709083252e-07,
+      "loss": 1.1641,
+      "step": 6221
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7684736881691532e-07,
+      "loss": 1.0913,
+      "step": 6222
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.764229916257043e-07,
+      "loss": 1.1113,
+      "step": 6223
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7599910560690686e-07,
+      "loss": 1.125,
+      "step": 6224
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.755757108501263e-07,
+      "loss": 1.1025,
+      "step": 6225
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.751528074448633e-07,
+      "loss": 1.0747,
+      "step": 6226
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7473039548051296e-07,
+      "loss": 1.1318,
+      "step": 6227
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7430847504636778e-07,
+      "loss": 1.0801,
+      "step": 6228
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.738870462316161e-07,
+      "loss": 1.084,
+      "step": 6229
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.73466109125342e-07,
+      "loss": 1.0874,
+      "step": 6230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7304566381652565e-07,
+      "loss": 1.1304,
+      "step": 6231
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7262571039404402e-07,
+      "loss": 1.1445,
+      "step": 6232
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7220624894666888e-07,
+      "loss": 1.1011,
+      "step": 6233
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.717872795630693e-07,
+      "loss": 1.1826,
+      "step": 6234
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.713688023318094e-07,
+      "loss": 1.062,
+      "step": 6235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.7095081734134849e-07,
+      "loss": 1.0894,
+      "step": 6236
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.7053332468004474e-07,
+      "loss": 1.1084,
+      "step": 6237
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.7011632443614952e-07,
+      "loss": 1.0996,
+      "step": 6238
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.696998166978106e-07,
+      "loss": 1.0688,
+      "step": 6239
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6928380155307288e-07,
+      "loss": 1.0767,
+      "step": 6240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6886827908987573e-07,
+      "loss": 1.1572,
+      "step": 6241
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6845324939605472e-07,
+      "loss": 1.0742,
+      "step": 6242
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.680387125593419e-07,
+      "loss": 1.1191,
+      "step": 6243
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.676246686673641e-07,
+      "loss": 1.0898,
+      "step": 6244
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.672111178076455e-07,
+      "loss": 1.0786,
+      "step": 6245
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6679806006760447e-07,
+      "loss": 1.0972,
+      "step": 6246
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6638549553455562e-07,
+      "loss": 1.0728,
+      "step": 6247
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.659734242957095e-07,
+      "loss": 1.1074,
+      "step": 6248
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.655618464381728e-07,
+      "loss": 1.125,
+      "step": 6249
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6515076204894702e-07,
+      "loss": 1.0376,
+      "step": 6250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6474017121493018e-07,
+      "loss": 1.1118,
+      "step": 6251
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6433007402291533e-07,
+      "loss": 1.083,
+      "step": 6252
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6392047055959093e-07,
+      "loss": 1.1143,
+      "step": 6253
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6351136091154242e-07,
+      "loss": 1.0747,
+      "step": 6254
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6310274516524922e-07,
+      "loss": 1.0942,
+      "step": 6255
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6269462340708808e-07,
+      "loss": 1.0635,
+      "step": 6256
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6228699572332946e-07,
+      "loss": 1.1377,
+      "step": 6257
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6187986220014025e-07,
+      "loss": 1.1318,
+      "step": 6258
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6147322292358354e-07,
+      "loss": 1.0474,
+      "step": 6259
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6106707797961696e-07,
+      "loss": 1.1055,
+      "step": 6260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.606614274540938e-07,
+      "loss": 1.1055,
+      "step": 6261
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.602562714327635e-07,
+      "loss": 1.0977,
+      "step": 6262
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5985161000127064e-07,
+      "loss": 1.0698,
+      "step": 6263
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5944744324515426e-07,
+      "loss": 1.0747,
+      "step": 6264
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.590437712498505e-07,
+      "loss": 1.1426,
+      "step": 6265
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5864059410068938e-07,
+      "loss": 1.1006,
+      "step": 6266
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5823791188289776e-07,
+      "loss": 1.1152,
+      "step": 6267
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5783572468159703e-07,
+      "loss": 1.103,
+      "step": 6268
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5743403258180357e-07,
+      "loss": 1.1504,
+      "step": 6269
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5703283566843003e-07,
+      "loss": 1.1279,
+      "step": 6270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5663213402628413e-07,
+      "loss": 1.1162,
+      "step": 6271
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.562319277400673e-07,
+      "loss": 1.1045,
+      "step": 6272
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.558322168943799e-07,
+      "loss": 1.1621,
+      "step": 6273
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5543300157371404e-07,
+      "loss": 1.083,
+      "step": 6274
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.550342818624581e-07,
+      "loss": 1.1025,
+      "step": 6275
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.546360578448969e-07,
+      "loss": 1.0947,
+      "step": 6276
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5423832960520884e-07,
+      "loss": 1.0986,
+      "step": 6277
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.53841097227469e-07,
+      "loss": 1.105,
+      "step": 6278
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5344436079564596e-07,
+      "loss": 1.0967,
+      "step": 6279
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.530481203936046e-07,
+      "loss": 1.1396,
+      "step": 6280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5265237610510548e-07,
+      "loss": 1.1475,
+      "step": 6281
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5225712801380272e-07,
+      "loss": 1.0908,
+      "step": 6282
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5186237620324646e-07,
+      "loss": 1.1201,
+      "step": 6283
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5146812075688162e-07,
+      "loss": 1.0967,
+      "step": 6284
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5107436175804957e-07,
+      "loss": 1.1191,
+      "step": 6285
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5068109928998405e-07,
+      "loss": 1.1064,
+      "step": 6286
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5028833343581684e-07,
+      "loss": 1.1006,
+      "step": 6287
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4989606427857185e-07,
+      "loss": 1.0977,
+      "step": 6288
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4950429190117049e-07,
+      "loss": 1.1235,
+      "step": 6289
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4911301638642794e-07,
+      "loss": 1.126,
+      "step": 6290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4872223781705358e-07,
+      "loss": 1.1104,
+      "step": 6291
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4833195627565383e-07,
+      "loss": 1.124,
+      "step": 6292
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4794217184472854e-07,
+      "loss": 1.125,
+      "step": 6293
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4755288460667205e-07,
+      "loss": 1.0835,
+      "step": 6294
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.471640946437755e-07,
+      "loss": 1.1299,
+      "step": 6295
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.467758020382226e-07,
+      "loss": 1.1084,
+      "step": 6296
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4638800687209408e-07,
+      "loss": 1.1514,
+      "step": 6297
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4600070922736436e-07,
+      "loss": 1.1699,
+      "step": 6298
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4561390918590212e-07,
+      "loss": 1.0845,
+      "step": 6299
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.452276068294728e-07,
+      "loss": 1.0815,
+      "step": 6300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4484180223973464e-07,
+      "loss": 1.1128,
+      "step": 6301
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4445649549824138e-07,
+      "loss": 1.125,
+      "step": 6302
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4407168668644222e-07,
+      "loss": 1.085,
+      "step": 6303
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4368737588567993e-07,
+      "loss": 1.1475,
+      "step": 6304
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4330356317719257e-07,
+      "loss": 1.0659,
+      "step": 6305
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.429202486421133e-07,
+      "loss": 1.1372,
+      "step": 6306
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4253743236146893e-07,
+      "loss": 1.1465,
+      "step": 6307
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.421551144161823e-07,
+      "loss": 1.0601,
+      "step": 6308
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.417732948870698e-07,
+      "loss": 1.1572,
+      "step": 6309
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4139197385484276e-07,
+      "loss": 1.1045,
+      "step": 6310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.410111514001078e-07,
+      "loss": 1.1182,
+      "step": 6311
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4063082760336522e-07,
+      "loss": 1.0664,
+      "step": 6312
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402510025450099e-07,
+      "loss": 1.0596,
+      "step": 6313
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3987167630533233e-07,
+      "loss": 1.0723,
+      "step": 6314
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.394928489645164e-07,
+      "loss": 1.0806,
+      "step": 6315
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.391145206026412e-07,
+      "loss": 1.1113,
+      "step": 6316
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3873669129968015e-07,
+      "loss": 1.0747,
+      "step": 6317
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.383593611355011e-07,
+      "loss": 1.1016,
+      "step": 6318
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3798253018986713e-07,
+      "loss": 1.083,
+      "step": 6319
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3760619854243429e-07,
+      "loss": 1.1289,
+      "step": 6320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3723036627275366e-07,
+      "loss": 1.0898,
+      "step": 6321
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3685503346027252e-07,
+      "loss": 1.0688,
+      "step": 6322
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3648020018433017e-07,
+      "loss": 1.123,
+      "step": 6323
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.36105866524161e-07,
+      "loss": 1.082,
+      "step": 6324
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3573203255889484e-07,
+      "loss": 1.0938,
+      "step": 6325
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.353586983675545e-07,
+      "loss": 1.125,
+      "step": 6326
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3498586402905744e-07,
+      "loss": 1.1025,
+      "step": 6327
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3461352962221646e-07,
+      "loss": 1.0845,
+      "step": 6328
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3424169522573722e-07,
+      "loss": 1.1157,
+      "step": 6329
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.33870360918221e-07,
+      "loss": 1.0859,
+      "step": 6330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3349952677816303e-07,
+      "loss": 1.1016,
+      "step": 6331
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.331291928839515e-07,
+      "loss": 1.1172,
+      "step": 6332
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.327593593138707e-07,
+      "loss": 1.1318,
+      "step": 6333
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3239002614609865e-07,
+      "loss": 1.1104,
+      "step": 6334
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3202119345870678e-07,
+      "loss": 1.1021,
+      "step": 6335
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3165286132966164e-07,
+      "loss": 1.0786,
+      "step": 6336
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3128502983682345e-07,
+      "loss": 1.1118,
+      "step": 6337
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3091769905794637e-07,
+      "loss": 1.0752,
+      "step": 6338
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3055086907068e-07,
+      "loss": 1.1123,
+      "step": 6339
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3018453995256612e-07,
+      "loss": 1.1514,
+      "step": 6340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.298187117810426e-07,
+      "loss": 1.124,
+      "step": 6341
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.294533846334403e-07,
+      "loss": 1.126,
+      "step": 6342
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.290885585869839e-07,
+      "loss": 1.1396,
+      "step": 6343
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2872423371879328e-07,
+      "loss": 1.0952,
+      "step": 6344
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2836041010588107e-07,
+      "loss": 1.1069,
+      "step": 6345
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.279970878251549e-07,
+      "loss": 1.106,
+      "step": 6346
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.276342669534167e-07,
+      "loss": 1.1074,
+      "step": 6347
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2727194756736117e-07,
+      "loss": 1.0957,
+      "step": 6348
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2691012974357765e-07,
+      "loss": 1.0576,
+      "step": 6349
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2654881355854992e-07,
+      "loss": 1.1074,
+      "step": 6350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2618799908865498e-07,
+      "loss": 1.0557,
+      "step": 6351
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.258276864101643e-07,
+      "loss": 1.0771,
+      "step": 6352
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.254678755992428e-07,
+      "loss": 1.0972,
+      "step": 6353
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2510856673194933e-07,
+      "loss": 1.1221,
+      "step": 6354
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2474975988423786e-07,
+      "loss": 1.1089,
+      "step": 6355
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.243914551319539e-07,
+      "loss": 1.1348,
+      "step": 6356
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.24033652550839e-07,
+      "loss": 1.0713,
+      "step": 6357
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2367635221652797e-07,
+      "loss": 1.0938,
+      "step": 6358
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.23319554204549e-07,
+      "loss": 1.0664,
+      "step": 6359
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.229632585903237e-07,
+      "loss": 1.0781,
+      "step": 6360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.226074654491688e-07,
+      "loss": 1.0488,
+      "step": 6361
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2225217485629353e-07,
+      "loss": 1.1011,
+      "step": 6362
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.218973868868023e-07,
+      "loss": 1.0942,
+      "step": 6363
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2154310161569177e-07,
+      "loss": 1.1035,
+      "step": 6364
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2118931911785315e-07,
+      "loss": 1.0996,
+      "step": 6365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2083603946807133e-07,
+      "loss": 1.1035,
+      "step": 6366
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2048326274102434e-07,
+      "loss": 1.1123,
+      "step": 6367
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2013098901128506e-07,
+      "loss": 1.0874,
+      "step": 6368
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1977921835331917e-07,
+      "loss": 1.1147,
+      "step": 6369
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1942795084148555e-07,
+      "loss": 1.0615,
+      "step": 6370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1907718655003787e-07,
+      "loss": 1.1523,
+      "step": 6371
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1872692555312315e-07,
+      "loss": 1.0601,
+      "step": 6372
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.183771679247811e-07,
+      "loss": 1.083,
+      "step": 6373
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1802791373894645e-07,
+      "loss": 1.0781,
+      "step": 6374
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1767916306944655e-07,
+      "loss": 1.1084,
+      "step": 6375
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1733091599000184e-07,
+      "loss": 1.0459,
+      "step": 6376
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1698317257422792e-07,
+      "loss": 1.1162,
+      "step": 6377
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1663593289563235e-07,
+      "loss": 1.1113,
+      "step": 6378
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1628919702761726e-07,
+      "loss": 1.1035,
+      "step": 6379
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1594296504347791e-07,
+      "loss": 1.0903,
+      "step": 6380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1559723701640213e-07,
+      "loss": 1.061,
+      "step": 6381
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1525201301947314e-07,
+      "loss": 1.0703,
+      "step": 6382
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1490729312566618e-07,
+      "loss": 1.0811,
+      "step": 6383
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1456307740785017e-07,
+      "loss": 1.1147,
+      "step": 6384
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1421936593878829e-07,
+      "loss": 1.0791,
+      "step": 6385
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1387615879113578e-07,
+      "loss": 1.1094,
+      "step": 6386
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1353345603744181e-07,
+      "loss": 1.1074,
+      "step": 6387
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1319125775014955e-07,
+      "loss": 1.1006,
+      "step": 6388
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1284956400159503e-07,
+      "loss": 1.1182,
+      "step": 6389
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1250837486400684e-07,
+      "loss": 1.1064,
+      "step": 6390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.121676904095087e-07,
+      "loss": 1.1289,
+      "step": 6391
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1182751071011605e-07,
+      "loss": 1.1221,
+      "step": 6392
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1148783583773886e-07,
+      "loss": 1.1064,
+      "step": 6393
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1114866586417888e-07,
+      "loss": 1.0552,
+      "step": 6394
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1081000086113236e-07,
+      "loss": 1.1455,
+      "step": 6395
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.104718409001887e-07,
+      "loss": 1.124,
+      "step": 6396
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1013418605283044e-07,
+      "loss": 1.0723,
+      "step": 6397
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0979703639043244e-07,
+      "loss": 1.1094,
+      "step": 6398
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0946039198426433e-07,
+      "loss": 1.0781,
+      "step": 6399
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0912425290548783e-07,
+      "loss": 1.0874,
+      "step": 6400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.087886192251575e-07,
+      "loss": 1.0884,
+      "step": 6401
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0845349101422292e-07,
+      "loss": 1.1104,
+      "step": 6402
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.081188683435247e-07,
+      "loss": 1.0386,
+      "step": 6403
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0778475128379789e-07,
+      "loss": 1.124,
+      "step": 6404
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0745113990567046e-07,
+      "loss": 1.1494,
+      "step": 6405
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0711803427966211e-07,
+      "loss": 1.0869,
+      "step": 6406
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0678543447618845e-07,
+      "loss": 1.0762,
+      "step": 6407
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0645334056555573e-07,
+      "loss": 1.082,
+      "step": 6408
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0612175261796393e-07,
+      "loss": 1.0996,
+      "step": 6409
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0579067070350669e-07,
+      "loss": 1.0898,
+      "step": 6410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0546009489216996e-07,
+      "loss": 1.0781,
+      "step": 6411
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0513002525383259e-07,
+      "loss": 1.0835,
+      "step": 6412
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0480046185826736e-07,
+      "loss": 1.1094,
+      "step": 6413
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0447140477513883e-07,
+      "loss": 1.0972,
+      "step": 6414
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.041428540740061e-07,
+      "loss": 1.1104,
+      "step": 6415
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.038148098243194e-07,
+      "loss": 1.0815,
+      "step": 6416
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0348727209542303e-07,
+      "loss": 1.1025,
+      "step": 6417
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0316024095655408e-07,
+      "loss": 1.1201,
+      "step": 6418
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0283371647684282e-07,
+      "loss": 1.0718,
+      "step": 6419
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0250769872531152e-07,
+      "loss": 1.1362,
+      "step": 6420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0218218777087646e-07,
+      "loss": 1.1562,
+      "step": 6421
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0185718368234565e-07,
+      "loss": 1.0703,
+      "step": 6422
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0153268652842025e-07,
+      "loss": 1.1123,
+      "step": 6423
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0120869637769537e-07,
+      "loss": 1.1289,
+      "step": 6424
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0088521329865763e-07,
+      "loss": 1.0688,
+      "step": 6425
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0056223735968706e-07,
+      "loss": 1.1133,
+      "step": 6426
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0023976862905627e-07,
+      "loss": 1.0747,
+      "step": 6427
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.991780717493043e-08,
+      "loss": 1.1367,
+      "step": 6428
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.959635306536875e-08,
+      "loss": 1.1177,
+      "step": 6429
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.9275406368321e-08,
+      "loss": 1.0728,
+      "step": 6430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.895496715163155e-08,
+      "loss": 1.0938,
+      "step": 6431
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.863503548303705e-08,
+      "loss": 1.125,
+      "step": 6432
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.831561143016671e-08,
+      "loss": 1.0732,
+      "step": 6433
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.799669506054182e-08,
+      "loss": 1.1318,
+      "step": 6434
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.767828644157761e-08,
+      "loss": 1.1328,
+      "step": 6435
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.736038564058054e-08,
+      "loss": 1.0586,
+      "step": 6436
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.704299272475154e-08,
+      "loss": 1.1177,
+      "step": 6437
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.672610776118225e-08,
+      "loss": 1.0923,
+      "step": 6438
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.640973081685796e-08,
+      "loss": 1.0918,
+      "step": 6439
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.609386195865683e-08,
+      "loss": 1.1309,
+      "step": 6440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.577850125334882e-08,
+      "loss": 1.0742,
+      "step": 6441
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.546364876759723e-08,
+      "loss": 1.0718,
+      "step": 6442
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.51493045679569e-08,
+      "loss": 1.1045,
+      "step": 6443
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.48354687208769e-08,
+      "loss": 1.0967,
+      "step": 6444
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.452214129269693e-08,
+      "loss": 1.0947,
+      "step": 6445
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.420932234965068e-08,
+      "loss": 1.1729,
+      "step": 6446
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.389701195786333e-08,
+      "loss": 1.0991,
+      "step": 6447
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.358521018335371e-08,
+      "loss": 1.1021,
+      "step": 6448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.327391709203187e-08,
+      "loss": 1.1138,
+      "step": 6449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.296313274970104e-08,
+      "loss": 1.1104,
+      "step": 6450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.265285722205697e-08,
+      "loss": 1.1455,
+      "step": 6451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.234309057468722e-08,
+      "loss": 1.1357,
+      "step": 6452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.203383287307299e-08,
+      "loss": 1.082,
+      "step": 6453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.172508418258675e-08,
+      "loss": 1.1211,
+      "step": 6454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.14168445684932e-08,
+      "loss": 1.1289,
+      "step": 6455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.110911409595024e-08,
+      "loss": 1.0791,
+      "step": 6456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.08018928300089e-08,
+      "loss": 1.1211,
+      "step": 6457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.049518083560999e-08,
+      "loss": 1.1025,
+      "step": 6458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.018897817758943e-08,
+      "loss": 1.0874,
+      "step": 6459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.988328492067405e-08,
+      "loss": 1.0723,
+      "step": 6460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.957810112948245e-08,
+      "loss": 1.0757,
+      "step": 6461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.927342686852747e-08,
+      "loss": 1.0581,
+      "step": 6462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.896926220221202e-08,
+      "loss": 1.1025,
+      "step": 6463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.86656071948333e-08,
+      "loss": 1.0757,
+      "step": 6464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.836246191057912e-08,
+      "loss": 1.1143,
+      "step": 6465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.805982641353017e-08,
+      "loss": 1.0947,
+      "step": 6466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.775770076766e-08,
+      "loss": 1.0938,
+      "step": 6467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.745608503683361e-08,
+      "loss": 1.124,
+      "step": 6468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.715497928480838e-08,
+      "loss": 1.1182,
+      "step": 6469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.685438357523445e-08,
+      "loss": 1.1299,
+      "step": 6470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.655429797165299e-08,
+      "loss": 1.0659,
+      "step": 6471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.625472253749795e-08,
+      "loss": 1.1221,
+      "step": 6472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.59556573360959e-08,
+      "loss": 1.127,
+      "step": 6473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.565710243066489e-08,
+      "loss": 1.1221,
+      "step": 6474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.535905788431554e-08,
+      "loss": 1.0762,
+      "step": 6475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.50615237600505e-08,
+      "loss": 1.1045,
+      "step": 6476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.476450012076365e-08,
+      "loss": 1.0986,
+      "step": 6477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.44679870292428e-08,
+      "loss": 1.126,
+      "step": 6478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.417198454816588e-08,
+      "loss": 1.0649,
+      "step": 6479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.387649274010396e-08,
+      "loss": 1.0767,
+      "step": 6480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.358151166752066e-08,
+      "loss": 1.0845,
+      "step": 6481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.328704139277028e-08,
+      "loss": 1.1021,
+      "step": 6482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.299308197809996e-08,
+      "loss": 1.061,
+      "step": 6483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.269963348564863e-08,
+      "loss": 1.1064,
+      "step": 6484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.240669597744722e-08,
+      "loss": 1.1074,
+      "step": 6485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.211426951541923e-08,
+      "loss": 1.0562,
+      "step": 6486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.182235416137913e-08,
+      "loss": 1.1021,
+      "step": 6487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.153094997703365e-08,
+      "loss": 1.1055,
+      "step": 6488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.124005702398214e-08,
+      "loss": 1.0425,
+      "step": 6489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.094967536371512e-08,
+      "loss": 1.0811,
+      "step": 6490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.065980505761517e-08,
+      "loss": 1.0801,
+      "step": 6491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.037044616695743e-08,
+      "loss": 1.0771,
+      "step": 6492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.0081598752908e-08,
+      "loss": 1.0977,
+      "step": 6493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.97932628765255e-08,
+      "loss": 1.1592,
+      "step": 6494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.950543859876008e-08,
+      "loss": 1.0728,
+      "step": 6495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.921812598045336e-08,
+      "loss": 1.0757,
+      "step": 6496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.893132508234036e-08,
+      "loss": 1.1494,
+      "step": 6497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.864503596504619e-08,
+      "loss": 1.0713,
+      "step": 6498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.835925868908828e-08,
+      "loss": 1.1514,
+      "step": 6499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.807399331487664e-08,
+      "loss": 1.1338,
+      "step": 6500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.77892399027122e-08,
+      "loss": 1.1255,
+      "step": 6501
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.750499851278764e-08,
+      "loss": 1.0981,
+      "step": 6502
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.722126920518769e-08,
+      "loss": 1.1484,
+      "step": 6503
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.693805203988963e-08,
+      "loss": 1.0415,
+      "step": 6504
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.665534707676086e-08,
+      "loss": 1.0933,
+      "step": 6505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.637315437556192e-08,
+      "loss": 1.1299,
+      "step": 6506
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.609147399594397e-08,
+      "loss": 1.1592,
+      "step": 6507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.581030599745076e-08,
+      "loss": 1.0928,
+      "step": 6508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.552965043951754e-08,
+      "loss": 1.1045,
+      "step": 6509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.524950738147018e-08,
+      "loss": 1.1299,
+      "step": 6510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.496987688252828e-08,
+      "loss": 1.0923,
+      "step": 6511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.469075900180117e-08,
+      "loss": 1.0723,
+      "step": 6512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.441215379829031e-08,
+      "loss": 1.0879,
+      "step": 6513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.413406133088968e-08,
+      "loss": 1.0493,
+      "step": 6514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.385648165838393e-08,
+      "loss": 1.0928,
+      "step": 6515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.357941483944947e-08,
+      "loss": 1.0923,
+      "step": 6516
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.33028609326547e-08,
+      "loss": 1.1289,
+      "step": 6517
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.302681999645927e-08,
+      "loss": 1.1318,
+      "step": 6518
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.275129208921455e-08,
+      "loss": 1.1396,
+      "step": 6519
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.24762772691634e-08,
+      "loss": 1.0864,
+      "step": 6520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.220177559443959e-08,
+      "loss": 1.0977,
+      "step": 6521
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.192778712306975e-08,
+      "loss": 1.123,
+      "step": 6522
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.165431191297117e-08,
+      "loss": 1.0928,
+      "step": 6523
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.138135002195207e-08,
+      "loss": 1.1016,
+      "step": 6524
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.110890150771404e-08,
+      "loss": 1.0776,
+      "step": 6525
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.083696642784771e-08,
+      "loss": 1.1172,
+      "step": 6526
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.056554483983763e-08,
+      "loss": 1.1338,
+      "step": 6527
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.029463680105764e-08,
+      "loss": 1.1201,
+      "step": 6528
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.002424236877442e-08,
+      "loss": 1.0889,
+      "step": 6529
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.975436160014615e-08,
+      "loss": 1.0977,
+      "step": 6530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.948499455222107e-08,
+      "loss": 1.0811,
+      "step": 6531
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.921614128194004e-08,
+      "loss": 1.0898,
+      "step": 6532
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.894780184613537e-08,
+      "loss": 1.0874,
+      "step": 6533
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.867997630152972e-08,
+      "loss": 1.0942,
+      "step": 6534
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.841266470473783e-08,
+      "loss": 1.1055,
+      "step": 6535
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.814586711226589e-08,
+      "loss": 1.1035,
+      "step": 6536
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.787958358051128e-08,
+      "loss": 1.0776,
+      "step": 6537
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.761381416576285e-08,
+      "loss": 1.1631,
+      "step": 6538
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.734855892420039e-08,
+      "loss": 1.1104,
+      "step": 6539
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.708381791189488e-08,
+      "loss": 1.1318,
+      "step": 6540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.681959118480963e-08,
+      "loss": 1.0635,
+      "step": 6541
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.655587879879855e-08,
+      "loss": 1.1074,
+      "step": 6542
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.629268080960621e-08,
+      "loss": 1.082,
+      "step": 6543
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.602999727286951e-08,
+      "loss": 1.0801,
+      "step": 6544
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.576782824411627e-08,
+      "loss": 1.125,
+      "step": 6545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.55061737787649e-08,
+      "loss": 1.0708,
+      "step": 6546
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.524503393212617e-08,
+      "loss": 1.061,
+      "step": 6547
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.498440875940094e-08,
+      "loss": 1.1113,
+      "step": 6548
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.472429831568234e-08,
+      "loss": 1.124,
+      "step": 6549
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.446470265595389e-08,
+      "loss": 1.1016,
+      "step": 6550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.420562183509033e-08,
+      "loss": 1.1133,
+      "step": 6551
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.39470559078581e-08,
+      "loss": 1.1172,
+      "step": 6552
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.368900492891489e-08,
+      "loss": 1.1094,
+      "step": 6553
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.343146895280816e-08,
+      "loss": 1.1035,
+      "step": 6554
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.317444803397855e-08,
+      "loss": 1.0518,
+      "step": 6555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.29179422267559e-08,
+      "loss": 1.1016,
+      "step": 6556
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.266195158536242e-08,
+      "loss": 1.0586,
+      "step": 6557
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.240647616391121e-08,
+      "loss": 1.0591,
+      "step": 6558
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.215151601640601e-08,
+      "loss": 1.1084,
+      "step": 6559
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.189707119674204e-08,
+      "loss": 1.1143,
+      "step": 6560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.164314175870545e-08,
+      "loss": 1.0659,
+      "step": 6561
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.138972775597301e-08,
+      "loss": 1.0684,
+      "step": 6562
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.113682924211383e-08,
+      "loss": 1.1143,
+      "step": 6563
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.088444627058626e-08,
+      "loss": 1.022,
+      "step": 6564
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.063257889474122e-08,
+      "loss": 1.1318,
+      "step": 6565
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.038122716782002e-08,
+      "loss": 1.1104,
+      "step": 6566
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.013039114295515e-08,
+      "loss": 1.085,
+      "step": 6567
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.988007087316894e-08,
+      "loss": 1.0503,
+      "step": 6568
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.963026641137681e-08,
+      "loss": 1.0938,
+      "step": 6569
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.938097781038349e-08,
+      "loss": 1.0806,
+      "step": 6570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.9132205122885143e-08,
+      "loss": 1.0996,
+      "step": 6571
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.888394840146944e-08,
+      "loss": 1.0928,
+      "step": 6572
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.863620769861328e-08,
+      "loss": 1.123,
+      "step": 6573
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.8388983066687e-08,
+      "loss": 1.0757,
+      "step": 6574
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.8142274557949907e-08,
+      "loss": 1.0557,
+      "step": 6575
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.789608222455251e-08,
+      "loss": 1.082,
+      "step": 6576
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.765040611853678e-08,
+      "loss": 1.0781,
+      "step": 6577
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.740524629183536e-08,
+      "loss": 1.0933,
+      "step": 6578
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.716060279627178e-08,
+      "loss": 1.0908,
+      "step": 6579
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.6916475683560235e-08,
+      "loss": 1.1582,
+      "step": 6580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.667286500530556e-08,
+      "loss": 1.0771,
+      "step": 6581
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.642977081300433e-08,
+      "loss": 1.1211,
+      "step": 6582
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.618719315804294e-08,
+      "loss": 1.1035,
+      "step": 6583
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.594513209169872e-08,
+      "loss": 1.1094,
+      "step": 6584
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.5703587665140726e-08,
+      "loss": 1.124,
+      "step": 6585
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.546255992942784e-08,
+      "loss": 1.1226,
+      "step": 6586
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.522204893550959e-08,
+      "loss": 1.1289,
+      "step": 6587
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.498205473422752e-08,
+      "loss": 1.1699,
+      "step": 6588
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.4742577376312165e-08,
+      "loss": 1.0723,
+      "step": 6589
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.450361691238665e-08,
+      "loss": 1.0972,
+      "step": 6590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.4265173392963885e-08,
+      "loss": 1.0923,
+      "step": 6591
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.4027246868446903e-08,
+      "loss": 1.0928,
+      "step": 6592
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.378983738913074e-08,
+      "loss": 1.1055,
+      "step": 6593
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.3552945005200255e-08,
+      "loss": 1.1191,
+      "step": 6594
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.331656976673122e-08,
+      "loss": 1.0894,
+      "step": 6595
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.308071172369061e-08,
+      "loss": 1.1348,
+      "step": 6596
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.2845370925934946e-08,
+      "loss": 1.1299,
+      "step": 6597
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.2610547423211924e-08,
+      "loss": 1.0864,
+      "step": 6598
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.237624126516072e-08,
+      "loss": 1.1001,
+      "step": 6599
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.214245250131006e-08,
+      "loss": 1.1147,
+      "step": 6600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.190918118107985e-08,
+      "loss": 1.0728,
+      "step": 6601
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.167642735378037e-08,
+      "loss": 1.1387,
+      "step": 6602
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.144419106861226e-08,
+      "loss": 1.0874,
+      "step": 6603
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.121247237466792e-08,
+      "loss": 1.1016,
+      "step": 6604
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.098127132092901e-08,
+      "loss": 1.0723,
+      "step": 6605
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.0750587956267805e-08,
+      "loss": 1.1172,
+      "step": 6606
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.052042232944865e-08,
+      "loss": 1.1143,
+      "step": 6607
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.029077448912456e-08,
+      "loss": 1.0776,
+      "step": 6608
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.006164448384032e-08,
+      "loss": 1.0859,
+      "step": 6609
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.98330323620308e-08,
+      "loss": 1.1045,
+      "step": 6610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.96049381720215e-08,
+      "loss": 1.0986,
+      "step": 6611
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.93773619620283e-08,
+      "loss": 1.1201,
+      "step": 6612
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.9150303780157725e-08,
+      "loss": 1.1138,
+      "step": 6613
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.8923763674406934e-08,
+      "loss": 1.1172,
+      "step": 6614
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.8697741692663446e-08,
+      "loss": 1.1016,
+      "step": 6615
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.847223788270516e-08,
+      "loss": 1.1152,
+      "step": 6616
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.824725229220034e-08,
+      "loss": 1.0947,
+      "step": 6617
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.8022784968707883e-08,
+      "loss": 1.0884,
+      "step": 6618
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.7798835959677325e-08,
+      "loss": 1.1045,
+      "step": 6619
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.757540531244803e-08,
+      "loss": 1.1226,
+      "step": 6620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.735249307425055e-08,
+      "loss": 1.1094,
+      "step": 6621
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.713009929220552e-08,
+      "loss": 1.1094,
+      "step": 6622
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.690822401332368e-08,
+      "loss": 1.0811,
+      "step": 6623
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.668686728450667e-08,
+      "loss": 1.127,
+      "step": 6624
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.6466029152545664e-08,
+      "loss": 1.1406,
+      "step": 6625
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.624570966412362e-08,
+      "loss": 1.0996,
+      "step": 6626
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.6025908865812986e-08,
+      "loss": 1.0967,
+      "step": 6627
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.580662680407605e-08,
+      "loss": 1.125,
+      "step": 6628
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.558786352526684e-08,
+      "loss": 1.1367,
+      "step": 6629
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.5369619075628365e-08,
+      "loss": 1.0957,
+      "step": 6630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.515189350129428e-08,
+      "loss": 1.1025,
+      "step": 6631
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.4934686848289445e-08,
+      "loss": 1.1069,
+      "step": 6632
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.471799916252795e-08,
+      "loss": 1.1396,
+      "step": 6633
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.45018304898151e-08,
+      "loss": 1.0942,
+      "step": 6634
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.428618087584546e-08,
+      "loss": 1.0898,
+      "step": 6635
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.407105036620452e-08,
+      "loss": 1.1016,
+      "step": 6636
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.3856439006367825e-08,
+      "loss": 1.1152,
+      "step": 6637
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.36423468417016e-08,
+      "loss": 1.0557,
+      "step": 6638
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.3428773917461866e-08,
+      "loss": 1.1201,
+      "step": 6639
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.321572027879528e-08,
+      "loss": 1.063,
+      "step": 6640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.300318597073805e-08,
+      "loss": 1.1094,
+      "step": 6641
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.2791171038217e-08,
+      "loss": 1.0532,
+      "step": 6642
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.2579675526049634e-08,
+      "loss": 1.1055,
+      "step": 6643
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.236869947894268e-08,
+      "loss": 1.0762,
+      "step": 6644
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.215824294149434e-08,
+      "loss": 1.1113,
+      "step": 6645
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.194830595819155e-08,
+      "loss": 1.0718,
+      "step": 6646
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.173888857341185e-08,
+      "loss": 1.0698,
+      "step": 6647
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.152999083142428e-08,
+      "loss": 1.1201,
+      "step": 6648
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.1321612776386025e-08,
+      "loss": 1.0903,
+      "step": 6649
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.1113754452345724e-08,
+      "loss": 1.1016,
+      "step": 6650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.0906415903242126e-08,
+      "loss": 1.0811,
+      "step": 6651
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.069959717290323e-08,
+      "loss": 1.0649,
+      "step": 6652
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.0493298305047666e-08,
+      "loss": 1.1074,
+      "step": 6653
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.028751934328473e-08,
+      "loss": 1.1621,
+      "step": 6654
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.0082260331112675e-08,
+      "loss": 1.1011,
+      "step": 6655
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.9877521311920966e-08,
+      "loss": 1.1084,
+      "step": 6656
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.967330232898858e-08,
+      "loss": 1.0986,
+      "step": 6657
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.946960342548406e-08,
+      "loss": 1.1006,
+      "step": 6658
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.9266424644467114e-08,
+      "loss": 1.0586,
+      "step": 6659
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.9063766028886454e-08,
+      "loss": 1.125,
+      "step": 6660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.8861627621581964e-08,
+      "loss": 1.0293,
+      "step": 6661
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.866000946528253e-08,
+      "loss": 1.0771,
+      "step": 6662
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.8458911602607954e-08,
+      "loss": 1.1191,
+      "step": 6663
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.8258334076066713e-08,
+      "loss": 1.0986,
+      "step": 6664
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.805827692805908e-08,
+      "loss": 1.1289,
+      "step": 6665
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.785874020087343e-08,
+      "loss": 1.0742,
+      "step": 6666
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.765972393669021e-08,
+      "loss": 1.061,
+      "step": 6667
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.746122817757797e-08,
+      "loss": 1.1191,
+      "step": 6668
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.726325296549621e-08,
+      "loss": 1.1201,
+      "step": 6669
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.706579834229424e-08,
+      "loss": 1.1035,
+      "step": 6670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.6868864349710874e-08,
+      "loss": 1.1299,
+      "step": 6671
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.667245102937617e-08,
+      "loss": 1.0884,
+      "step": 6672
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.647655842280828e-08,
+      "loss": 1.1357,
+      "step": 6673
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.6281186571416604e-08,
+      "loss": 1.106,
+      "step": 6674
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.608633551650004e-08,
+      "loss": 1.1543,
+      "step": 6675
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.589200529924758e-08,
+      "loss": 1.1016,
+      "step": 6676
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.569819596073748e-08,
+      "loss": 1.0835,
+      "step": 6677
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.550490754193919e-08,
+      "loss": 1.1133,
+      "step": 6678
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.531214008371087e-08,
+      "loss": 1.1279,
+      "step": 6679
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.5119893626800737e-08,
+      "loss": 1.0957,
+      "step": 6680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.4928168211847136e-08,
+      "loss": 1.1191,
+      "step": 6681
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.473696387937819e-08,
+      "loss": 1.1641,
+      "step": 6682
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.4546280669812135e-08,
+      "loss": 1.103,
+      "step": 6683
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.435611862345645e-08,
+      "loss": 1.1182,
+      "step": 6684
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.4166477780508964e-08,
+      "loss": 1.0693,
+      "step": 6685
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.397735818105707e-08,
+      "loss": 1.0913,
+      "step": 6686
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.3788759865078235e-08,
+      "loss": 1.0757,
+      "step": 6687
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.360068287243945e-08,
+      "loss": 1.1113,
+      "step": 6688
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.34131272428978e-08,
+      "loss": 1.1318,
+      "step": 6689
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.322609301609964e-08,
+      "loss": 1.1069,
+      "step": 6690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.3039580231581655e-08,
+      "loss": 1.041,
+      "step": 6691
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.285358892877011e-08,
+      "loss": 1.0986,
+      "step": 6692
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.2668119146981035e-08,
+      "loss": 1.1309,
+      "step": 6693
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.2483170925420025e-08,
+      "loss": 1.1309,
+      "step": 6694
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.229874430318275e-08,
+      "loss": 1.1177,
+      "step": 6695
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.211483931925441e-08,
+      "loss": 1.1006,
+      "step": 6696
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.1931456012510296e-08,
+      "loss": 1.1006,
+      "step": 6697
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.174859442171469e-08,
+      "loss": 1.125,
+      "step": 6698
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.1566254585522206e-08,
+      "loss": 1.1143,
+      "step": 6699
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.138443654247703e-08,
+      "loss": 1.1216,
+      "step": 6700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.120314033101313e-08,
+      "loss": 1.0747,
+      "step": 6701
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.102236598945402e-08,
+      "loss": 1.1426,
+      "step": 6702
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.084211355601274e-08,
+      "loss": 1.1113,
+      "step": 6703
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.066238306879216e-08,
+      "loss": 1.083,
+      "step": 6704
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.0483174565784926e-08,
+      "loss": 1.1084,
+      "step": 6705
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.0304488084873517e-08,
+      "loss": 1.1201,
+      "step": 6706
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.012632366382939e-08,
+      "loss": 1.1045,
+      "step": 6707
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9948681340314366e-08,
+      "loss": 1.0952,
+      "step": 6708
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9771561151879515e-08,
+      "loss": 1.0986,
+      "step": 6709
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9594963135965437e-08,
+      "loss": 1.0781,
+      "step": 6710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9418887329902833e-08,
+      "loss": 1.1006,
+      "step": 6711
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9243333770911654e-08,
+      "loss": 1.1318,
+      "step": 6712
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9068302496101373e-08,
+      "loss": 1.084,
+      "step": 6713
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.889379354247157e-08,
+      "loss": 1.125,
+      "step": 6714
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.871980694691079e-08,
+      "loss": 1.1143,
+      "step": 6715
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.854634274619711e-08,
+      "loss": 1.1357,
+      "step": 6716
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.8373400976998977e-08,
+      "loss": 1.0391,
+      "step": 6717
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.8200981675873816e-08,
+      "loss": 1.0581,
+      "step": 6718
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.8029084879268585e-08,
+      "loss": 1.1318,
+      "step": 6719
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.785771062352005e-08,
+      "loss": 1.0898,
+      "step": 6720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7686858944853956e-08,
+      "loss": 1.0938,
+      "step": 6721
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7516529879386687e-08,
+      "loss": 1.1387,
+      "step": 6722
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.734672346312306e-08,
+      "loss": 1.1187,
+      "step": 6723
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7177439731957966e-08,
+      "loss": 1.1758,
+      "step": 6724
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7008678721675562e-08,
+      "loss": 1.0967,
+      "step": 6725
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6840440467949812e-08,
+      "loss": 1.0908,
+      "step": 6726
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6672725006343934e-08,
+      "loss": 1.1045,
+      "step": 6727
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.6505532372310406e-08,
+      "loss": 1.0625,
+      "step": 6728
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.6338862601191783e-08,
+      "loss": 1.0615,
+      "step": 6729
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.617271572821961e-08,
+      "loss": 1.1089,
+      "step": 6730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.6007091788515238e-08,
+      "loss": 1.085,
+      "step": 6731
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.584199081708899e-08,
+      "loss": 1.1006,
+      "step": 6732
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.567741284884101e-08,
+      "loss": 1.1089,
+      "step": 6733
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.5513357918561243e-08,
+      "loss": 1.127,
+      "step": 6734
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.5349826060928062e-08,
+      "loss": 1.0898,
+      "step": 6735
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.5186817310510758e-08,
+      "loss": 1.0913,
+      "step": 6736
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.5024331701766214e-08,
+      "loss": 1.1201,
+      "step": 6737
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4862369269041953e-08,
+      "loss": 1.0869,
+      "step": 6738
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4700930046574757e-08,
+      "loss": 1.1025,
+      "step": 6739
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.454001406849066e-08,
+      "loss": 1.1094,
+      "step": 6740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4379621368805228e-08,
+      "loss": 1.0952,
+      "step": 6741
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.421975198142301e-08,
+      "loss": 1.1172,
+      "step": 6742
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4060405940138086e-08,
+      "loss": 1.085,
+      "step": 6743
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.3901583278634344e-08,
+      "loss": 1.0986,
+      "step": 6744
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.3743284030484647e-08,
+      "loss": 1.1191,
+      "step": 6745
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.35855082291514e-08,
+      "loss": 1.0791,
+      "step": 6746
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.3428255907985976e-08,
+      "loss": 1.1025,
+      "step": 6747
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.3271527100229564e-08,
+      "loss": 1.0713,
+      "step": 6748
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.3115321839012328e-08,
+      "loss": 1.0928,
+      "step": 6749
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2959640157354246e-08,
+      "loss": 1.1064,
+      "step": 6750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2804482088163992e-08,
+      "loss": 1.082,
+      "step": 6751
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2649847664240053e-08,
+      "loss": 1.1074,
+      "step": 6752
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2495736918269894e-08,
+      "loss": 1.0796,
+      "step": 6753
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2342149882830512e-08,
+      "loss": 1.0898,
+      "step": 6754
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2189086590387876e-08,
+      "loss": 1.0947,
+      "step": 6755
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2036547073297776e-08,
+      "loss": 1.1211,
+      "step": 6756
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1884531363804972e-08,
+      "loss": 1.1069,
+      "step": 6757
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1733039494043206e-08,
+      "loss": 1.0908,
+      "step": 6758
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1582071496036027e-08,
+      "loss": 1.127,
+      "step": 6759
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1431627401695965e-08,
+      "loss": 1.0967,
+      "step": 6760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1281707242824802e-08,
+      "loss": 1.1289,
+      "step": 6761
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1132311051113575e-08,
+      "loss": 1.063,
+      "step": 6762
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0983438858142857e-08,
+      "loss": 1.0898,
+      "step": 6763
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0835090695381922e-08,
+      "loss": 1.0908,
+      "step": 6764
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.068726659418957e-08,
+      "loss": 1.1123,
+      "step": 6765
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.053996658581414e-08,
+      "loss": 1.0674,
+      "step": 6766
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0393190701392117e-08,
+      "loss": 1.0825,
+      "step": 6767
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0246938971950626e-08,
+      "loss": 1.127,
+      "step": 6768
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0101211428404933e-08,
+      "loss": 1.103,
+      "step": 6769
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9956008101559842e-08,
+      "loss": 1.1025,
+      "step": 6770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9811329022109416e-08,
+      "loss": 1.1064,
+      "step": 6771
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9667174220636964e-08,
+      "loss": 1.0518,
+      "step": 6772
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.952354372761478e-08,
+      "loss": 1.1221,
+      "step": 6773
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.938043757340441e-08,
+      "loss": 1.0591,
+      "step": 6774
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9237855788256378e-08,
+      "loss": 1.0649,
+      "step": 6775
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.909579840231046e-08,
+      "loss": 1.1426,
+      "step": 6776
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8954265445595698e-08,
+      "loss": 1.124,
+      "step": 6777
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8813256948030654e-08,
+      "loss": 1.1372,
+      "step": 6778
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8672772939422045e-08,
+      "loss": 1.1133,
+      "step": 6779
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.853281344946667e-08,
+      "loss": 1.0703,
+      "step": 6780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.839337850774947e-08,
+      "loss": 1.1001,
+      "step": 6781
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.825446814374604e-08,
+      "loss": 1.0947,
+      "step": 6782
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.811608238681928e-08,
+      "loss": 1.0581,
+      "step": 6783
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7978221266222172e-08,
+      "loss": 1.0625,
+      "step": 6784
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.784088481109697e-08,
+      "loss": 1.1035,
+      "step": 6785
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.770407305047461e-08,
+      "loss": 1.0781,
+      "step": 6786
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.75677860132753e-08,
+      "loss": 1.1113,
+      "step": 6787
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.743202372830821e-08,
+      "loss": 1.0737,
+      "step": 6788
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.72967862242715e-08,
+      "loss": 1.1328,
+      "step": 6789
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7162073529752575e-08,
+      "loss": 1.0693,
+      "step": 6790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7027885673228105e-08,
+      "loss": 1.127,
+      "step": 6791
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.68942226830629e-08,
+      "loss": 1.1284,
+      "step": 6792
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.676108458751241e-08,
+      "loss": 1.0547,
+      "step": 6793
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6628471414719682e-08,
+      "loss": 1.0527,
+      "step": 6794
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.649638319271729e-08,
+      "loss": 1.1045,
+      "step": 6795
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.636481994942679e-08,
+      "loss": 1.085,
+      "step": 6796
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6233781712659546e-08,
+      "loss": 1.1201,
+      "step": 6797
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6103268510114235e-08,
+      "loss": 1.0742,
+      "step": 6798
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.597328036938045e-08,
+      "loss": 1.0986,
+      "step": 6799
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5843817317935382e-08,
+      "loss": 1.126,
+      "step": 6800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.571487938314603e-08,
+      "loss": 1.083,
+      "step": 6801
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.558646659226781e-08,
+      "loss": 1.1006,
+      "step": 6802
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5458578972445682e-08,
+      "loss": 1.0977,
+      "step": 6803
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.533121655071329e-08,
+      "loss": 1.0942,
+      "step": 6804
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5204379353993548e-08,
+      "loss": 1.0771,
+      "step": 6805
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5078067409097507e-08,
+      "loss": 1.1025,
+      "step": 6806
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.495228074272603e-08,
+      "loss": 1.1094,
+      "step": 6807
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4827019381468966e-08,
+      "loss": 1.082,
+      "step": 6808
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4702283351804569e-08,
+      "loss": 1.0767,
+      "step": 6809
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4578072680100642e-08,
+      "loss": 1.0713,
+      "step": 6810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.44543873926134e-08,
+      "loss": 1.1104,
+      "step": 6811
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4331227515488045e-08,
+      "loss": 1.1172,
+      "step": 6812
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4208593074759303e-08,
+      "loss": 1.1064,
+      "step": 6813
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.408648409635005e-08,
+      "loss": 1.1611,
+      "step": 6814
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3964900606072696e-08,
+      "loss": 1.0732,
+      "step": 6815
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3843842629628346e-08,
+      "loss": 1.1001,
+      "step": 6816
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3723310192606532e-08,
+      "loss": 1.1006,
+      "step": 6817
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3603303320486872e-08,
+      "loss": 1.1113,
+      "step": 6818
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3483822038636574e-08,
+      "loss": 1.1182,
+      "step": 6819
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3364866372312657e-08,
+      "loss": 1.0972,
+      "step": 6820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.324643634666084e-08,
+      "loss": 1.103,
+      "step": 6821
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3128531986715265e-08,
+      "loss": 1.127,
+      "step": 6822
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3011153317399606e-08,
+      "loss": 1.1055,
+      "step": 6823
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2894300363526236e-08,
+      "loss": 1.0649,
+      "step": 6824
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2777973149795676e-08,
+      "loss": 1.1133,
+      "step": 6825
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2662171700798254e-08,
+      "loss": 1.1377,
+      "step": 6826
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2546896041013001e-08,
+      "loss": 1.1064,
+      "step": 6827
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2432146194807093e-08,
+      "loss": 1.1016,
+      "step": 6828
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2317922186437791e-08,
+      "loss": 1.1348,
+      "step": 6829
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.220422404004995e-08,
+      "loss": 1.126,
+      "step": 6830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2091051779677953e-08,
+      "loss": 1.1455,
+      "step": 6831
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1978405429244888e-08,
+      "loss": 1.0791,
+      "step": 6832
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.186628501256254e-08,
+      "loss": 1.1167,
+      "step": 6833
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1754690553331672e-08,
+      "loss": 1.0693,
+      "step": 6834
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1643622075142303e-08,
+      "loss": 1.0596,
+      "step": 6835
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1533079601472041e-08,
+      "loss": 1.1348,
+      "step": 6836
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1423063155688586e-08,
+      "loss": 1.1895,
+      "step": 6837
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1313572761047498e-08,
+      "loss": 1.085,
+      "step": 6838
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1204608440693876e-08,
+      "loss": 1.1426,
+      "step": 6839
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.109617021766124e-08,
+      "loss": 1.0742,
+      "step": 6840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0988258114871808e-08,
+      "loss": 1.0889,
+      "step": 6841
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.088087215513678e-08,
+      "loss": 1.1309,
+      "step": 6842
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0774012361156328e-08,
+      "loss": 1.1172,
+      "step": 6843
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0667678755518495e-08,
+      "loss": 1.1328,
+      "step": 6844
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0561871360701414e-08,
+      "loss": 1.063,
+      "step": 6845
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.045659019907136e-08,
+      "loss": 1.1172,
+      "step": 6846
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.035183529288275e-08,
+      "loss": 1.0898,
+      "step": 6847
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0247606664279819e-08,
+      "loss": 1.0967,
+      "step": 6848
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0143904335294662e-08,
+      "loss": 1.1084,
+      "step": 6849
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0040728327848914e-08,
+      "loss": 1.0913,
+      "step": 6850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.938078663752626e-09,
+      "loss": 1.1123,
+      "step": 6851
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.835955364703997e-09,
+      "loss": 1.1084,
+      "step": 6852
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.734358452291038e-09,
+      "loss": 1.125,
+      "step": 6853
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.6332879479899e-09,
+      "loss": 1.0972,
+      "step": 6854
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.532743873165162e-09,
+      "loss": 1.1504,
+      "step": 6855
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.432726249070656e-09,
+      "loss": 1.0972,
+      "step": 6856
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.333235096848636e-09,
+      "loss": 1.1128,
+      "step": 6857
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.234270437530335e-09,
+      "loss": 1.1094,
+      "step": 6858
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.135832292035684e-09,
+      "loss": 1.0732,
+      "step": 6859
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.037920681172763e-09,
+      "loss": 1.126,
+      "step": 6860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.940535625639179e-09,
+      "loss": 1.1064,
+      "step": 6861
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.843677146020413e-09,
+      "loss": 1.0732,
+      "step": 6862
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.747345262790918e-09,
+      "loss": 1.1045,
+      "step": 6863
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.651539996314685e-09,
+      "loss": 1.1201,
+      "step": 6864
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.556261366842733e-09,
+      "loss": 1.1143,
+      "step": 6865
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.461509394516177e-09,
+      "loss": 1.1035,
+      "step": 6866
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.36728409936427e-09,
+      "loss": 1.1504,
+      "step": 6867
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.27358550130497e-09,
+      "loss": 1.0801,
+      "step": 6868
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.180413620144378e-09,
+      "loss": 1.1104,
+      "step": 6869
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.087768475578684e-09,
+      "loss": 1.127,
+      "step": 6870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.995650087190832e-09,
+      "loss": 1.0928,
+      "step": 6871
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.904058474454134e-09,
+      "loss": 1.1504,
+      "step": 6872
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.812993656729218e-09,
+      "loss": 1.0986,
+      "step": 6873
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.722455653266515e-09,
+      "loss": 1.0693,
+      "step": 6874
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.63244448320405e-09,
+      "loss": 1.1323,
+      "step": 6875
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.54296016556938e-09,
+      "loss": 1.0645,
+      "step": 6876
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.454002719277654e-09,
+      "loss": 1.1113,
+      "step": 6877
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.365572163133827e-09,
+      "loss": 1.1201,
+      "step": 6878
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.277668515830727e-09,
+      "loss": 1.1113,
+      "step": 6879
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.190291795949878e-09,
+      "loss": 1.1094,
+      "step": 6880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.103442021961504e-09,
+      "loss": 1.1523,
+      "step": 6881
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.01711921222481e-09,
+      "loss": 1.0913,
+      "step": 6882
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.931323384986865e-09,
+      "loss": 1.1182,
+      "step": 6883
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.846054558383996e-09,
+      "loss": 1.0708,
+      "step": 6884
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.7613127504406736e-09,
+      "loss": 1.1592,
+      "step": 6885
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.677097979070346e-09,
+      "loss": 1.0898,
+      "step": 6886
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.593410262074884e-09,
+      "loss": 1.1201,
+      "step": 6887
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.510249617144304e-09,
+      "loss": 1.1782,
+      "step": 6888
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.427616061858155e-09,
+      "loss": 1.0957,
+      "step": 6889
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.345509613683576e-09,
+      "loss": 1.1328,
+      "step": 6890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.263930289976961e-09,
+      "loss": 1.1094,
+      "step": 6891
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.182878107983404e-09,
+      "loss": 1.1289,
+      "step": 6892
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.102353084835588e-09,
+      "loss": 1.0791,
+      "step": 6893
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.02235523755601e-09,
+      "loss": 1.0894,
+      "step": 6894
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.942884583054753e-09,
+      "loss": 1.0557,
+      "step": 6895
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.8639411381308796e-09,
+      "loss": 1.0786,
+      "step": 6896
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.785524919471874e-09,
+      "loss": 1.127,
+      "step": 6897
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.7076359436542e-09,
+      "loss": 1.167,
+      "step": 6898
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.630274227141907e-09,
+      "loss": 1.1084,
+      "step": 6899
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.5534397862888615e-09,
+      "loss": 1.0742,
+      "step": 6900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.4771326373367905e-09,
+      "loss": 1.1255,
+      "step": 6901
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.4013527964152935e-09,
+      "loss": 1.0996,
+      "step": 6902
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.326100279543778e-09,
+      "loss": 1.1318,
+      "step": 6903
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.251375102629519e-09,
+      "loss": 1.1055,
+      "step": 6904
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.177177281468215e-09,
+      "loss": 1.1748,
+      "step": 6905
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.103506831744265e-09,
+      "loss": 1.126,
+      "step": 6906
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.030363769031044e-09,
+      "loss": 1.1104,
+      "step": 6907
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.957748108789517e-09,
+      "loss": 1.1128,
+      "step": 6908
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.8856598663696295e-09,
+      "loss": 1.0698,
+      "step": 6909
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.814099057010302e-09,
+      "loss": 1.0928,
+      "step": 6910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.743065695838045e-09,
+      "loss": 1.1484,
+      "step": 6911
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.672559797868348e-09,
+      "loss": 1.1108,
+      "step": 6912
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.6025813780056775e-09,
+      "loss": 1.0771,
+      "step": 6913
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.533130451042089e-09,
+      "loss": 1.1299,
+      "step": 6914
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.464207031658618e-09,
+      "loss": 1.106,
+      "step": 6915
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.395811134424721e-09,
+      "loss": 1.123,
+      "step": 6916
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.327942773798554e-09,
+      "loss": 1.0908,
+      "step": 6917
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.260601964126421e-09,
+      "loss": 1.1191,
+      "step": 6918
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.193788719643044e-09,
+      "loss": 1.0938,
+      "step": 6919
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.1275030544721266e-09,
+      "loss": 1.1221,
+      "step": 6920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.0617449826255154e-09,
+      "loss": 1.0376,
+      "step": 6921
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.996514518003758e-09,
+      "loss": 1.106,
+      "step": 6922
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.931811674394992e-09,
+      "loss": 1.1152,
+      "step": 6923
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.867636465477165e-09,
+      "loss": 1.0908,
+      "step": 6924
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.8039889048160915e-09,
+      "loss": 1.1421,
+      "step": 6925
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.740869005865455e-09,
+      "loss": 1.1084,
+      "step": 6926
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.678276781968193e-09,
+      "loss": 1.0698,
+      "step": 6927
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.616212246355388e-09,
+      "loss": 1.123,
+      "step": 6928
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5546754121471015e-09,
+      "loss": 1.0791,
+      "step": 6929
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.493666292350706e-09,
+      "loss": 1.106,
+      "step": 6930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.433184899863107e-09,
+      "loss": 1.1162,
+      "step": 6931
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.3732312474688e-09,
+      "loss": 1.0723,
+      "step": 6932
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.313805347841814e-09,
+      "loss": 1.0884,
+      "step": 6933
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2549072135437675e-09,
+      "loss": 1.1396,
+      "step": 6934
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.196536857024701e-09,
+      "loss": 1.0664,
+      "step": 6935
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.138694290623356e-09,
+      "loss": 1.1201,
+      "step": 6936
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0813795265668966e-09,
+      "loss": 1.1079,
+      "step": 6937
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0245925769709086e-09,
+      "loss": 1.1299,
+      "step": 6938
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9683334538394003e-09,
+      "loss": 1.0547,
+      "step": 6939
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9126021690645266e-09,
+      "loss": 1.1533,
+      "step": 6940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8573987344276964e-09,
+      "loss": 1.0967,
+      "step": 6941
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8027231615976316e-09,
+      "loss": 1.1201,
+      "step": 6942
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7485754621320304e-09,
+      "loss": 1.0811,
+      "step": 6943
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.694955647477293e-09,
+      "loss": 1.1279,
+      "step": 6944
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6418637289679636e-09,
+      "loss": 1.0718,
+      "step": 6945
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5892997178264545e-09,
+      "loss": 1.1406,
+      "step": 6946
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5372636251644322e-09,
+      "loss": 1.0996,
+      "step": 6947
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4857554619814317e-09,
+      "loss": 1.0884,
+      "step": 6948
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4347752391656875e-09,
+      "loss": 1.1348,
+      "step": 6949
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.384322967493857e-09,
+      "loss": 1.1074,
+      "step": 6950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3343986576307433e-09,
+      "loss": 1.0996,
+      "step": 6951
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2850023201295703e-09,
+      "loss": 1.083,
+      "step": 6952
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.236133965431986e-09,
+      "loss": 1.124,
+      "step": 6953
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1877936038683377e-09,
+      "loss": 1.1191,
+      "step": 6954
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.139981245657119e-09,
+      "loss": 1.0771,
+      "step": 6955
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.092696900904967e-09,
+      "loss": 1.0972,
+      "step": 6956
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0459405796072197e-09,
+      "loss": 1.0854,
+      "step": 6957
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9997122916476375e-09,
+      "loss": 1.1152,
+      "step": 6958
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.954012046798126e-09,
+      "loss": 1.0835,
+      "step": 6959
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.908839854719291e-09,
+      "loss": 1.083,
+      "step": 6960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8641957249596056e-09,
+      "loss": 1.1055,
+      "step": 6961
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8200796669565202e-09,
+      "loss": 1.0708,
+      "step": 6962
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7764916900353535e-09,
+      "loss": 1.0869,
+      "step": 6963
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7334318034101239e-09,
+      "loss": 1.1094,
+      "step": 6964
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6909000161832723e-09,
+      "loss": 1.0889,
+      "step": 6965
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6488963373448296e-09,
+      "loss": 1.1113,
+      "step": 6966
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6074207757746373e-09,
+      "loss": 1.0996,
+      "step": 6967
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.566473340239294e-09,
+      "loss": 1.0918,
+      "step": 6968
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5260540393952083e-09,
+      "loss": 1.0996,
+      "step": 6969
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4861628817858242e-09,
+      "loss": 1.1104,
+      "step": 6970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4467998758441181e-09,
+      "loss": 1.1201,
+      "step": 6971
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4079650298906567e-09,
+      "loss": 1.1348,
+      "step": 6972
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3696583521344286e-09,
+      "loss": 1.1172,
+      "step": 6973
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.331879850673401e-09,
+      "loss": 1.0942,
+      "step": 6974
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2946295334928527e-09,
+      "loss": 1.0938,
+      "step": 6975
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2579074084673182e-09,
+      "loss": 1.0928,
+      "step": 6976
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.221713483359477e-09,
+      "loss": 1.0703,
+      "step": 6977
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.186047765820153e-09,
+      "loss": 1.1006,
+      "step": 6978
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1509102633883163e-09,
+      "loss": 1.1377,
+      "step": 6979
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1163009834919136e-09,
+      "loss": 1.1016,
+      "step": 6980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0822199334464823e-09,
+      "loss": 1.1299,
+      "step": 6981
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0486671204568145e-09,
+      "loss": 1.0854,
+      "step": 6982
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0156425516150147e-09,
+      "loss": 1.1006,
+      "step": 6983
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.831462339021658e-10,
+      "loss": 1.0928,
+      "step": 6984
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.511781741877724e-10,
+      "loss": 1.083,
+      "step": 6985
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.197383792292069e-10,
+      "loss": 1.1055,
+      "step": 6986
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.888268556725422e-10,
+      "loss": 1.0967,
+      "step": 6987
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.584436100519955e-10,
+      "loss": 1.1274,
+      "step": 6988
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.285886487902073e-10,
+      "loss": 1.0962,
+      "step": 6989
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.992619781982403e-10,
+      "loss": 1.0508,
+      "step": 6990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.704636044747471e-10,
+      "loss": 1.1016,
+      "step": 6991
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.421935337081909e-10,
+      "loss": 1.125,
+      "step": 6992
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.14451771873792e-10,
+      "loss": 1.1104,
+      "step": 6993
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.872383248360259e-10,
+      "loss": 1.1328,
+      "step": 6994
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.605531983475133e-10,
+      "loss": 1.0889,
+      "step": 6995
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.343963980490198e-10,
+      "loss": 1.1064,
+      "step": 6996
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.087679294697335e-10,
+      "loss": 1.125,
+      "step": 6997
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.836677980272654e-10,
+      "loss": 1.124,
+      "step": 6998
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.590960090273711e-10,
+      "loss": 1.1143,
+      "step": 6999
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.350525676642293e-10,
+      "loss": 1.0967,
+      "step": 7000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.115374790201632e-10,
+      "loss": 1.1006,
+      "step": 7001
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.885507480661967e-10,
+      "loss": 1.1201,
+      "step": 7002
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.660923796612205e-10,
+      "loss": 1.0625,
+      "step": 7003
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.441623785525484e-10,
+      "loss": 1.1182,
+      "step": 7004
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.227607493759167e-10,
+      "loss": 1.126,
+      "step": 7005
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0188749665576175e-10,
+      "loss": 1.0435,
+      "step": 7006
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8154262480383233e-10,
+      "loss": 1.0864,
+      "step": 7007
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.617261381208548e-10,
+      "loss": 1.0933,
+      "step": 7008
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.424380407959782e-10,
+      "loss": 1.127,
+      "step": 7009
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.236783369064966e-10,
+      "loss": 1.0938,
+      "step": 7010
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.054470304175716e-10,
+      "loss": 1.0776,
+      "step": 7011
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8774412518334236e-10,
+      "loss": 1.0947,
+      "step": 7012
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7056962494609317e-10,
+      "loss": 1.1572,
+      "step": 7013
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5392353333597577e-10,
+      "loss": 1.1377,
+      "step": 7014
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.37805853871842e-10,
+      "loss": 1.1152,
+      "step": 7015
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2221658996068872e-10,
+      "loss": 1.1504,
+      "step": 7016
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.071557448982131e-10,
+      "loss": 1.0854,
+      "step": 7017
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.926233218674245e-10,
+      "loss": 1.0576,
+      "step": 7018
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7861932394114268e-10,
+      "loss": 1.083,
+      "step": 7019
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6514375407866712e-10,
+      "loss": 1.1104,
+      "step": 7020
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.521966151293852e-10,
+      "loss": 1.0776,
+      "step": 7021
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3977790982971916e-10,
+      "loss": 1.1064,
+      "step": 7022
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2788764080479133e-10,
+      "loss": 1.0923,
+      "step": 7023
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.165258105681466e-10,
+      "loss": 1.1367,
+      "step": 7024
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0569242152147497e-10,
+      "loss": 1.0547,
+      "step": 7025
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.538747595516651e-11,
+      "loss": 1.0947,
+      "step": 7026
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.56109760469237e-11,
+      "loss": 1.1475,
+      "step": 7027
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.636292386370425e-11,
+      "loss": 1.0493,
+      "step": 7028
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.764332136033336e-11,
+      "loss": 1.0513,
+      "step": 7029
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.945217038005879e-11,
+      "loss": 1.1543,
+      "step": 7030
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.178947265455092e-11,
+      "loss": 1.0889,
+      "step": 7031
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.465522980307002e-11,
+      "loss": 1.0991,
+      "step": 7032
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.804944333440919e-11,
+      "loss": 1.0898,
+      "step": 7033
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1972114644396315e-11,
+      "loss": 1.0796,
+      "step": 7034
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.642324501811455e-11,
+      "loss": 1.1108,
+      "step": 7035
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1402835627959416e-11,
+      "loss": 1.0898,
+      "step": 7036
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6910887535581676e-11,
+      "loss": 1.1572,
+      "step": 7037
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2947401690777129e-11,
+      "loss": 1.1587,
+      "step": 7038
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.512378930653932e-12,
+      "loss": 1.0854,
+      "step": 7039
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.605819981853057e-12,
+      "loss": 1.0908,
+      "step": 7040
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.227725458605392e-12,
+      "loss": 1.0737,
+      "step": 7041
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.378095863841967e-12,
+      "loss": 1.0996,
+      "step": 7042
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0569315880837316e-12,
+      "loss": 1.1064,
+      "step": 7043
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6423291110688754e-13,
+      "loss": 1.1108,
+      "step": 7044
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 1.0869,
+      "step": 7045
+    },
+    {
+      "epoch": 1.0,
+      "step": 7045,
+      "total_flos": 1.4118353085990437e+19,
+      "train_loss": 1.1225040476401704,
+      "train_runtime": 82446.7613,
+      "train_samples_per_second": 10.937,
+      "train_steps_per_second": 0.085
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 7045,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 2000,
+  "total_flos": 1.4118353085990437e+19,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}