diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,187525 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "global_step": 31250,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1321961620469085e-08,
+      "loss": 4.0469,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.264392324093817e-08,
+      "loss": 4.3672,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.396588486140725e-08,
+      "loss": 4.4102,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.528784648187634e-08,
+      "loss": 4.7266,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0660980810234542e-07,
+      "loss": 4.3555,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.279317697228145e-07,
+      "loss": 5.0312,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4925373134328358e-07,
+      "loss": 4.1875,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7057569296375268e-07,
+      "loss": 4.0664,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.918976545842218e-07,
+      "loss": 4.375,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1321961620469084e-07,
+      "loss": 4.1836,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3454157782515995e-07,
+      "loss": 4.8945,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.55863539445629e-07,
+      "loss": 4.6484,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.771855010660981e-07,
+      "loss": 4.5508,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9850746268656716e-07,
+      "loss": 4.0156,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1982942430703626e-07,
+      "loss": 4.4219,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4115138592750537e-07,
+      "loss": 4.5469,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.624733475479744e-07,
+      "loss": 4.0273,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.837953091684436e-07,
+      "loss": 4.2656,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0511727078891263e-07,
+      "loss": 4.4102,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.264392324093817e-07,
+      "loss": 4.7109,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4776119402985074e-07,
+      "loss": 4.7031,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.690831556503199e-07,
+      "loss": 4.6367,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.904051172707889e-07,
+      "loss": 4.7266,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.11727078891258e-07,
+      "loss": 4.5625,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.33049040511727e-07,
+      "loss": 4.7344,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.543710021321962e-07,
+      "loss": 4.5078,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.756929637526653e-07,
+      "loss": 4.625,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.970149253731343e-07,
+      "loss": 4.2891,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.183368869936035e-07,
+      "loss": 4.9219,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.396588486140725e-07,
+      "loss": 4.8359,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.609808102345417e-07,
+      "loss": 4.1719,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.823027718550107e-07,
+      "loss": 4.0703,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.036247334754798e-07,
+      "loss": 3.8633,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.249466950959488e-07,
+      "loss": 4.6953,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.462686567164179e-07,
+      "loss": 4.2734,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.675906183368872e-07,
+      "loss": 4.4648,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.889125799573562e-07,
+      "loss": 4.3555,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.102345415778253e-07,
+      "loss": 4.1602,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.315565031982943e-07,
+      "loss": 4.2969,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.528784648187634e-07,
+      "loss": 4.3633,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.742004264392324e-07,
+      "loss": 4.6641,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.955223880597015e-07,
+      "loss": 3.9102,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.168443496801707e-07,
+      "loss": 4.2188,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.381663113006398e-07,
+      "loss": 4.0273,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.594882729211088e-07,
+      "loss": 4.5625,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.808102345415779e-07,
+      "loss": 4.7969,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.002132196162047e-06,
+      "loss": 4.0625,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.023454157782516e-06,
+      "loss": 4.0117,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.044776119402985e-06,
+      "loss": 4.2695,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.066098081023454e-06,
+      "loss": 4.8203,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0874200426439234e-06,
+      "loss": 3.8359,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1087420042643924e-06,
+      "loss": 4.043,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1300639658848615e-06,
+      "loss": 3.9844,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1513859275053305e-06,
+      "loss": 3.8594,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1727078891257996e-06,
+      "loss": 3.7031,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1940298507462686e-06,
+      "loss": 3.7305,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2153518123667379e-06,
+      "loss": 3.8164,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.236673773987207e-06,
+      "loss": 3.7188,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.257995735607676e-06,
+      "loss": 3.8125,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.279317697228145e-06,
+      "loss": 3.8945,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.300639658848614e-06,
+      "loss": 3.7695,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3219616204690834e-06,
+      "loss": 3.1328,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3432835820895524e-06,
+      "loss": 3.3047,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3646055437100215e-06,
+      "loss": 3.7578,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3859275053304905e-06,
+      "loss": 3.2539,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4072494669509596e-06,
+      "loss": 3.2305,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 3.4258,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4498933901918977e-06,
+      "loss": 3.0469,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4712153518123667e-06,
+      "loss": 3.1055,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4925373134328358e-06,
+      "loss": 3.2422,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5138592750533053e-06,
+      "loss": 2.7539,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5351812366737743e-06,
+      "loss": 3.4531,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5565031982942434e-06,
+      "loss": 3.125,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5778251599147124e-06,
+      "loss": 2.6992,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5991471215351815e-06,
+      "loss": 2.957,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6204690831556505e-06,
+      "loss": 2.6406,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6417910447761196e-06,
+      "loss": 2.4297,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6631130063965886e-06,
+      "loss": 2.0566,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6844349680170577e-06,
+      "loss": 2.3086,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7057569296375267e-06,
+      "loss": 2.334,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7270788912579958e-06,
+      "loss": 1.9199,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7484008528784648e-06,
+      "loss": 2.127,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7697228144989339e-06,
+      "loss": 2.0957,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.791044776119403e-06,
+      "loss": 1.9961,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.812366737739872e-06,
+      "loss": 2.0957,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8336886993603415e-06,
+      "loss": 2.043,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8550106609808105e-06,
+      "loss": 2.1953,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8763326226012796e-06,
+      "loss": 1.832,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8976545842217486e-06,
+      "loss": 2.0977,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9189765458422177e-06,
+      "loss": 1.5742,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9402985074626867e-06,
+      "loss": 1.9219,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9616204690831558e-06,
+      "loss": 1.4111,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.982942430703625e-06,
+      "loss": 1.4902,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.004264392324094e-06,
+      "loss": 1.6719,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.025586353944563e-06,
+      "loss": 2.168,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.046908315565032e-06,
+      "loss": 1.4092,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.068230277185501e-06,
+      "loss": 1.5801,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.08955223880597e-06,
+      "loss": 1.9219,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.110874200426439e-06,
+      "loss": 2.0215,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.132196162046908e-06,
+      "loss": 1.8613,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1535181236673773e-06,
+      "loss": 1.5469,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1748400852878467e-06,
+      "loss": 1.7344,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1961620469083158e-06,
+      "loss": 1.5234,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.217484008528785e-06,
+      "loss": 1.75,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.238805970149254e-06,
+      "loss": 1.7188,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.260127931769723e-06,
+      "loss": 1.877,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.281449893390192e-06,
+      "loss": 1.5664,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.302771855010661e-06,
+      "loss": 1.8477,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.32409381663113e-06,
+      "loss": 1.752,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.345415778251599e-06,
+      "loss": 1.9668,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.366737739872068e-06,
+      "loss": 1.3516,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3880597014925373e-06,
+      "loss": 1.9688,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4093816631130067e-06,
+      "loss": 1.6406,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4307036247334758e-06,
+      "loss": 1.6562,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.452025586353945e-06,
+      "loss": 1.6523,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.473347547974414e-06,
+      "loss": 1.6699,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.494669509594883e-06,
+      "loss": 1.6543,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.515991471215352e-06,
+      "loss": 1.6543,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.537313432835821e-06,
+      "loss": 1.6406,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.55863539445629e-06,
+      "loss": 1.4971,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.579957356076759e-06,
+      "loss": 1.7793,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.601279317697228e-06,
+      "loss": 1.623,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6226012793176977e-06,
+      "loss": 1.9297,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6439232409381667e-06,
+      "loss": 1.4648,
+      "step": 124
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6652452025586358e-06,
+      "loss": 2.0391,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.686567164179105e-06,
+      "loss": 1.4248,
+      "step": 126
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.707889125799574e-06,
+      "loss": 1.5898,
+      "step": 127
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.729211087420043e-06,
+      "loss": 1.9707,
+      "step": 128
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.750533049040512e-06,
+      "loss": 1.7266,
+      "step": 129
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.771855010660981e-06,
+      "loss": 1.8145,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.79317697228145e-06,
+      "loss": 1.5186,
+      "step": 131
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.814498933901919e-06,
+      "loss": 1.8262,
+      "step": 132
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.835820895522388e-06,
+      "loss": 2.3418,
+      "step": 133
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 1.7617,
+      "step": 134
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8784648187633263e-06,
+      "loss": 2.0176,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8997867803837954e-06,
+      "loss": 1.4805,
+      "step": 136
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9211087420042644e-06,
+      "loss": 1.9883,
+      "step": 137
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9424307036247335e-06,
+      "loss": 1.3594,
+      "step": 138
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9637526652452025e-06,
+      "loss": 1.6787,
+      "step": 139
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9850746268656716e-06,
+      "loss": 1.4375,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.006396588486141e-06,
+      "loss": 1.6621,
+      "step": 141
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0277185501066105e-06,
+      "loss": 2.0312,
+      "step": 142
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0490405117270796e-06,
+      "loss": 1.8574,
+      "step": 143
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0703624733475486e-06,
+      "loss": 2.0449,
+      "step": 144
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0916844349680177e-06,
+      "loss": 1.7441,
+      "step": 145
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1130063965884867e-06,
+      "loss": 1.0693,
+      "step": 146
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1343283582089558e-06,
+      "loss": 1.9531,
+      "step": 147
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.155650319829425e-06,
+      "loss": 1.8281,
+      "step": 148
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.176972281449894e-06,
+      "loss": 1.5859,
+      "step": 149
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.198294243070363e-06,
+      "loss": 1.9395,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.219616204690832e-06,
+      "loss": 1.8652,
+      "step": 151
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.240938166311301e-06,
+      "loss": 1.6543,
+      "step": 152
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.26226012793177e-06,
+      "loss": 1.3809,
+      "step": 153
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.283582089552239e-06,
+      "loss": 1.4121,
+      "step": 154
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.304904051172708e-06,
+      "loss": 1.2109,
+      "step": 155
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3262260127931773e-06,
+      "loss": 1.3994,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3475479744136463e-06,
+      "loss": 1.4287,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3688699360341154e-06,
+      "loss": 1.7344,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3901918976545844e-06,
+      "loss": 1.332,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4115138592750535e-06,
+      "loss": 1.46,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4328358208955225e-06,
+      "loss": 1.5352,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4541577825159916e-06,
+      "loss": 1.5859,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4754797441364606e-06,
+      "loss": 1.7852,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4968017057569297e-06,
+      "loss": 1.8516,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5181236673773987e-06,
+      "loss": 1.3574,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5394456289978678e-06,
+      "loss": 1.7012,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.560767590618337e-06,
+      "loss": 1.6035,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.582089552238806e-06,
+      "loss": 1.418,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.603411513859275e-06,
+      "loss": 1.6318,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.624733475479744e-06,
+      "loss": 1.7773,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6460554371002135e-06,
+      "loss": 1.3809,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.667377398720683e-06,
+      "loss": 1.7148,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.688699360341152e-06,
+      "loss": 1.8594,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.710021321961621e-06,
+      "loss": 1.4316,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.73134328358209e-06,
+      "loss": 1.4395,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.752665245202559e-06,
+      "loss": 1.8457,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.773987206823028e-06,
+      "loss": 1.5664,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7953091684434973e-06,
+      "loss": 2.0371,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.816631130063966e-06,
+      "loss": 1.7188,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.837953091684435e-06,
+      "loss": 1.8379,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.859275053304904e-06,
+      "loss": 1.709,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8805970149253735e-06,
+      "loss": 1.2656,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9019189765458425e-06,
+      "loss": 1.4629,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9232409381663116e-06,
+      "loss": 1.4707,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.944562899786781e-06,
+      "loss": 1.457,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.96588486140725e-06,
+      "loss": 1.4629,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.987206823027719e-06,
+      "loss": 2.0332,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.008528784648188e-06,
+      "loss": 1.3359,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.029850746268657e-06,
+      "loss": 1.2383,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.051172707889126e-06,
+      "loss": 1.75,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.072494669509595e-06,
+      "loss": 1.9551,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.093816631130064e-06,
+      "loss": 1.8848,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.115138592750533e-06,
+      "loss": 1.9648,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.136460554371002e-06,
+      "loss": 1.5508,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.157782515991471e-06,
+      "loss": 1.4902,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.17910447761194e-06,
+      "loss": 1.5391,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.200426439232409e-06,
+      "loss": 1.498,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.221748400852878e-06,
+      "loss": 1.2715,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.243070362473347e-06,
+      "loss": 1.4922,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.264392324093816e-06,
+      "loss": 1.6289,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 1.4121,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3070362473347545e-06,
+      "loss": 1.6484,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3283582089552236e-06,
+      "loss": 1.4336,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3496801705756935e-06,
+      "loss": 1.5645,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3710021321961625e-06,
+      "loss": 1.7168,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3923240938166316e-06,
+      "loss": 1.6152,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.413646055437101e-06,
+      "loss": 1.8184,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.43496801705757e-06,
+      "loss": 1.7344,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.456289978678039e-06,
+      "loss": 1.7559,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.477611940298508e-06,
+      "loss": 1.6582,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.498933901918977e-06,
+      "loss": 1.4023,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.520255863539446e-06,
+      "loss": 1.584,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.541577825159915e-06,
+      "loss": 1.5156,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.562899786780384e-06,
+      "loss": 1.8438,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.584221748400853e-06,
+      "loss": 1.4609,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.605543710021322e-06,
+      "loss": 1.5469,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.626865671641791e-06,
+      "loss": 1.5381,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.64818763326226e-06,
+      "loss": 2.0254,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.669509594882729e-06,
+      "loss": 1.6895,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.690831556503198e-06,
+      "loss": 1.7266,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.712153518123667e-06,
+      "loss": 1.6895,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.733475479744136e-06,
+      "loss": 1.6113,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7547974413646055e-06,
+      "loss": 1.7812,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7761194029850745e-06,
+      "loss": 1.7207,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.797441364605544e-06,
+      "loss": 1.3301,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8187633262260135e-06,
+      "loss": 1.7559,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8400852878464825e-06,
+      "loss": 1.6094,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8614072494669516e-06,
+      "loss": 1.4961,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.882729211087421e-06,
+      "loss": 1.6211,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.90405117270789e-06,
+      "loss": 1.709,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.925373134328359e-06,
+      "loss": 1.2793,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.946695095948828e-06,
+      "loss": 1.5508,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.968017057569297e-06,
+      "loss": 1.6328,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.989339019189766e-06,
+      "loss": 1.4531,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.010660980810235e-06,
+      "loss": 1.6133,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.031982942430704e-06,
+      "loss": 1.6533,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.053304904051173e-06,
+      "loss": 1.7852,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.074626865671642e-06,
+      "loss": 1.6113,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.095948827292111e-06,
+      "loss": 1.7559,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.11727078891258e-06,
+      "loss": 1.4189,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.138592750533049e-06,
+      "loss": 1.8965,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.159914712153518e-06,
+      "loss": 1.3604,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.181236673773987e-06,
+      "loss": 1.3398,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.202558635394456e-06,
+      "loss": 1.6406,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.2238805970149255e-06,
+      "loss": 1.5088,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.245202558635395e-06,
+      "loss": 1.7129,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.2665245202558636e-06,
+      "loss": 1.709,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.2878464818763335e-06,
+      "loss": 1.3086,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.309168443496802e-06,
+      "loss": 1.6299,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.3304904051172716e-06,
+      "loss": 1.3418,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.351812366737741e-06,
+      "loss": 1.4512,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.37313432835821e-06,
+      "loss": 1.4453,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.394456289978679e-06,
+      "loss": 1.2852,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.415778251599148e-06,
+      "loss": 1.9434,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.437100213219617e-06,
+      "loss": 1.4727,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.458422174840086e-06,
+      "loss": 1.4658,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.479744136460555e-06,
+      "loss": 1.4375,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.501066098081024e-06,
+      "loss": 1.6855,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.522388059701493e-06,
+      "loss": 1.4453,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.543710021321962e-06,
+      "loss": 1.8496,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.565031982942431e-06,
+      "loss": 1.4951,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.5863539445629e-06,
+      "loss": 1.8047,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.607675906183369e-06,
+      "loss": 1.459,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.628997867803838e-06,
+      "loss": 1.6895,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.650319829424308e-06,
+      "loss": 1.6094,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.671641791044776e-06,
+      "loss": 1.3574,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.692963752665246e-06,
+      "loss": 1.6426,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 1.5332,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.735607675906184e-06,
+      "loss": 1.8066,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.756929637526653e-06,
+      "loss": 1.8789,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7782515991471225e-06,
+      "loss": 1.9453,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.799573560767591e-06,
+      "loss": 1.6992,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.820895522388061e-06,
+      "loss": 1.4551,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.842217484008529e-06,
+      "loss": 2.0977,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.863539445628999e-06,
+      "loss": 1.7773,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.884861407249467e-06,
+      "loss": 1.3994,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.906183368869937e-06,
+      "loss": 1.3613,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.927505330490405e-06,
+      "loss": 1.793,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.948827292110875e-06,
+      "loss": 1.8457,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.970149253731343e-06,
+      "loss": 1.7695,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.991471215351813e-06,
+      "loss": 2.002,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.012793176972282e-06,
+      "loss": 1.6074,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.034115138592751e-06,
+      "loss": 1.418,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.055437100213221e-06,
+      "loss": 1.4092,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.076759061833689e-06,
+      "loss": 1.4414,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.098081023454159e-06,
+      "loss": 1.752,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.119402985074627e-06,
+      "loss": 1.6348,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.140724946695097e-06,
+      "loss": 1.416,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.1620469083155655e-06,
+      "loss": 1.5898,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.183368869936035e-06,
+      "loss": 1.5273,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2046908315565036e-06,
+      "loss": 1.6299,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2260127931769735e-06,
+      "loss": 1.4971,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.247334754797442e-06,
+      "loss": 1.7715,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2686567164179116e-06,
+      "loss": 1.5908,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.28997867803838e-06,
+      "loss": 1.2617,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.31130063965885e-06,
+      "loss": 1.4082,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.332622601279318e-06,
+      "loss": 1.377,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.353944562899788e-06,
+      "loss": 1.3496,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.375266524520256e-06,
+      "loss": 1.6338,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.396588486140726e-06,
+      "loss": 1.6426,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.417910447761194e-06,
+      "loss": 1.3291,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.439232409381664e-06,
+      "loss": 1.6367,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.460554371002132e-06,
+      "loss": 1.3926,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.481876332622602e-06,
+      "loss": 2.041,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.50319829424307e-06,
+      "loss": 1.0068,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.52452025586354e-06,
+      "loss": 1.4375,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.545842217484008e-06,
+      "loss": 1.4814,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.567164179104478e-06,
+      "loss": 1.8906,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.5884861407249465e-06,
+      "loss": 1.5234,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.609808102345416e-06,
+      "loss": 1.6367,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.631130063965885e-06,
+      "loss": 1.4531,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.6524520255863545e-06,
+      "loss": 1.3945,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.673773987206824e-06,
+      "loss": 1.3438,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.695095948827293e-06,
+      "loss": 1.4414,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.7164179104477625e-06,
+      "loss": 1.6289,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.737739872068231e-06,
+      "loss": 1.6973,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.759061833688701e-06,
+      "loss": 1.5771,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.780383795309169e-06,
+      "loss": 1.4609,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.801705756929639e-06,
+      "loss": 1.7637,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.823027718550107e-06,
+      "loss": 1.8242,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.844349680170577e-06,
+      "loss": 1.6582,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.865671641791045e-06,
+      "loss": 1.6348,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.886993603411515e-06,
+      "loss": 1.5117,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.908315565031983e-06,
+      "loss": 1.9707,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.929637526652453e-06,
+      "loss": 1.5469,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.950959488272921e-06,
+      "loss": 1.5029,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.972281449893391e-06,
+      "loss": 1.582,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.993603411513859e-06,
+      "loss": 1.5957,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.014925373134329e-06,
+      "loss": 1.5547,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.0362473347547975e-06,
+      "loss": 1.6016,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.057569296375267e-06,
+      "loss": 1.8164,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.0788912579957356e-06,
+      "loss": 1.4277,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1002132196162055e-06,
+      "loss": 1.4102,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.121535181236674e-06,
+      "loss": 1.5664,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1428571428571436e-06,
+      "loss": 1.4648,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.164179104477612e-06,
+      "loss": 1.4102,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.185501066098082e-06,
+      "loss": 2.0039,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.20682302771855e-06,
+      "loss": 1.5488,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.22814498933902e-06,
+      "loss": 1.8398,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.249466950959488e-06,
+      "loss": 1.2754,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.270788912579958e-06,
+      "loss": 1.9316,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.292110874200427e-06,
+      "loss": 1.5117,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.313432835820896e-06,
+      "loss": 1.6641,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.334754797441366e-06,
+      "loss": 1.3613,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.356076759061834e-06,
+      "loss": 1.7285,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.377398720682304e-06,
+      "loss": 2.1328,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.398720682302772e-06,
+      "loss": 1.291,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.420042643923242e-06,
+      "loss": 1.4258,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.44136460554371e-06,
+      "loss": 1.9473,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.46268656716418e-06,
+      "loss": 1.4785,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.484008528784648e-06,
+      "loss": 1.3945,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.505330490405118e-06,
+      "loss": 1.2852,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.5266524520255865e-06,
+      "loss": 1.584,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.547974413646056e-06,
+      "loss": 1.4512,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.569296375266525e-06,
+      "loss": 1.6855,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.5906183368869945e-06,
+      "loss": 1.3438,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.611940298507463e-06,
+      "loss": 1.7676,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.633262260127933e-06,
+      "loss": 1.6914,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.654584221748402e-06,
+      "loss": 1.8672,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.67590618336887e-06,
+      "loss": 1.5879,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.69722814498934e-06,
+      "loss": 1.5703,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.718550106609809e-06,
+      "loss": 1.793,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.739872068230278e-06,
+      "loss": 1.748,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.761194029850747e-06,
+      "loss": 1.7676,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.782515991471216e-06,
+      "loss": 1.0449,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.803837953091685e-06,
+      "loss": 1.2588,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.825159914712154e-06,
+      "loss": 1.6699,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.846481876332623e-06,
+      "loss": 1.6562,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.867803837953092e-06,
+      "loss": 1.4551,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.889125799573561e-06,
+      "loss": 1.5879,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.91044776119403e-06,
+      "loss": 1.793,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.9317697228145e-06,
+      "loss": 1.8008,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.953091684434968e-06,
+      "loss": 1.4229,
+      "step": 373
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.974413646055437e-06,
+      "loss": 1.3545,
+      "step": 374
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.995735607675907e-06,
+      "loss": 1.4189,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.017057569296376e-06,
+      "loss": 1.5938,
+      "step": 376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.038379530916846e-06,
+      "loss": 1.7363,
+      "step": 377
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.059701492537314e-06,
+      "loss": 1.1924,
+      "step": 378
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.081023454157784e-06,
+      "loss": 1.748,
+      "step": 379
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.102345415778252e-06,
+      "loss": 1.5566,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.123667377398723e-06,
+      "loss": 1.4922,
+      "step": 381
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.14498933901919e-06,
+      "loss": 1.3418,
+      "step": 382
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.16631130063966e-06,
+      "loss": 1.7285,
+      "step": 383
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.187633262260128e-06,
+      "loss": 1.4512,
+      "step": 384
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.208955223880599e-06,
+      "loss": 1.6465,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.230277185501066e-06,
+      "loss": 1.6641,
+      "step": 386
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.251599147121537e-06,
+      "loss": 1.4648,
+      "step": 387
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.272921108742004e-06,
+      "loss": 1.541,
+      "step": 388
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.294243070362475e-06,
+      "loss": 1.3584,
+      "step": 389
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.315565031982942e-06,
+      "loss": 1.5908,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.336886993603413e-06,
+      "loss": 1.2793,
+      "step": 391
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.35820895522388e-06,
+      "loss": 1.5703,
+      "step": 392
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.379530916844351e-06,
+      "loss": 1.6094,
+      "step": 393
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.400852878464819e-06,
+      "loss": 1.877,
+      "step": 394
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.42217484008529e-06,
+      "loss": 1.7773,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.443496801705757e-06,
+      "loss": 2.1016,
+      "step": 396
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.464818763326227e-06,
+      "loss": 1.5361,
+      "step": 397
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.486140724946695e-06,
+      "loss": 1.458,
+      "step": 398
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.507462686567165e-06,
+      "loss": 1.7227,
+      "step": 399
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.528784648187633e-06,
+      "loss": 1.9473,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.550106609808104e-06,
+      "loss": 1.8477,
+      "step": 401
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 1.7852,
+      "step": 402
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.592750533049042e-06,
+      "loss": 1.6973,
+      "step": 403
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.614072494669509e-06,
+      "loss": 1.8633,
+      "step": 404
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.63539445628998e-06,
+      "loss": 1.2949,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.656716417910447e-06,
+      "loss": 1.2803,
+      "step": 406
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.678038379530918e-06,
+      "loss": 1.5195,
+      "step": 407
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.699360341151387e-06,
+      "loss": 1.6602,
+      "step": 408
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.720682302771856e-06,
+      "loss": 1.1953,
+      "step": 409
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.742004264392325e-06,
+      "loss": 1.791,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.763326226012794e-06,
+      "loss": 1.4922,
+      "step": 411
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.784648187633263e-06,
+      "loss": 1.627,
+      "step": 412
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.805970149253732e-06,
+      "loss": 1.6582,
+      "step": 413
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.827292110874201e-06,
+      "loss": 1.7051,
+      "step": 414
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.84861407249467e-06,
+      "loss": 1.5195,
+      "step": 415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.86993603411514e-06,
+      "loss": 1.584,
+      "step": 416
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.891257995735608e-06,
+      "loss": 1.4609,
+      "step": 417
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.912579957356077e-06,
+      "loss": 1.502,
+      "step": 418
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.933901918976547e-06,
+      "loss": 0.958,
+      "step": 419
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.955223880597016e-06,
+      "loss": 1.4941,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.976545842217485e-06,
+      "loss": 1.5791,
+      "step": 421
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.997867803837954e-06,
+      "loss": 1.418,
+      "step": 422
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.019189765458423e-06,
+      "loss": 1.416,
+      "step": 423
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.040511727078892e-06,
+      "loss": 1.5742,
+      "step": 424
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.06183368869936e-06,
+      "loss": 1.582,
+      "step": 425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.08315565031983e-06,
+      "loss": 1.6855,
+      "step": 426
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.104477611940299e-06,
+      "loss": 1.457,
+      "step": 427
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.125799573560768e-06,
+      "loss": 1.2373,
+      "step": 428
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.147121535181237e-06,
+      "loss": 1.5234,
+      "step": 429
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.168443496801706e-06,
+      "loss": 1.6416,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.189765458422175e-06,
+      "loss": 1.5566,
+      "step": 431
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.211087420042644e-06,
+      "loss": 1.2246,
+      "step": 432
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.232409381663113e-06,
+      "loss": 1.3369,
+      "step": 433
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.253731343283582e-06,
+      "loss": 1.584,
+      "step": 434
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.275053304904051e-06,
+      "loss": 1.585,
+      "step": 435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.29637526652452e-06,
+      "loss": 1.5352,
+      "step": 436
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.31769722814499e-06,
+      "loss": 1.5098,
+      "step": 437
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.339019189765458e-06,
+      "loss": 1.4668,
+      "step": 438
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.36034115138593e-06,
+      "loss": 1.3555,
+      "step": 439
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.381663113006397e-06,
+      "loss": 1.3467,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.402985074626867e-06,
+      "loss": 1.7129,
+      "step": 441
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.424307036247335e-06,
+      "loss": 1.584,
+      "step": 442
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.445628997867805e-06,
+      "loss": 1.4141,
+      "step": 443
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.466950959488273e-06,
+      "loss": 1.6113,
+      "step": 444
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.488272921108744e-06,
+      "loss": 1.584,
+      "step": 445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.509594882729211e-06,
+      "loss": 1.2744,
+      "step": 446
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.530916844349682e-06,
+      "loss": 1.2607,
+      "step": 447
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.552238805970149e-06,
+      "loss": 1.5752,
+      "step": 448
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.57356076759062e-06,
+      "loss": 1.6406,
+      "step": 449
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.594882729211089e-06,
+      "loss": 1.5264,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.616204690831558e-06,
+      "loss": 1.4072,
+      "step": 451
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.637526652452027e-06,
+      "loss": 1.9863,
+      "step": 452
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.658848614072496e-06,
+      "loss": 1.7051,
+      "step": 453
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.680170575692965e-06,
+      "loss": 1.4961,
+      "step": 454
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.701492537313434e-06,
+      "loss": 1.4033,
+      "step": 455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.722814498933903e-06,
+      "loss": 1.627,
+      "step": 456
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.744136460554372e-06,
+      "loss": 1.8398,
+      "step": 457
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.765458422174841e-06,
+      "loss": 1.25,
+      "step": 458
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.78678038379531e-06,
+      "loss": 1.7734,
+      "step": 459
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.80810234541578e-06,
+      "loss": 1.1719,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.829424307036248e-06,
+      "loss": 1.7227,
+      "step": 461
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.850746268656717e-06,
+      "loss": 1.3223,
+      "step": 462
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.872068230277187e-06,
+      "loss": 1.6758,
+      "step": 463
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.893390191897656e-06,
+      "loss": 1.4023,
+      "step": 464
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.914712153518125e-06,
+      "loss": 1.6816,
+      "step": 465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.936034115138594e-06,
+      "loss": 1.3984,
+      "step": 466
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.957356076759063e-06,
+      "loss": 1.668,
+      "step": 467
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.978678038379532e-06,
+      "loss": 1.4277,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1e-05,
+      "loss": 1.5381,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.002132196162047e-05,
+      "loss": 1.459,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.004264392324094e-05,
+      "loss": 1.7461,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0063965884861408e-05,
+      "loss": 1.7676,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0085287846481877e-05,
+      "loss": 1.4883,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0106609808102346e-05,
+      "loss": 1.4238,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0127931769722817e-05,
+      "loss": 1.9004,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0149253731343284e-05,
+      "loss": 1.6299,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0170575692963753e-05,
+      "loss": 1.1919,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0191897654584222e-05,
+      "loss": 1.627,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0213219616204693e-05,
+      "loss": 1.5703,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.023454157782516e-05,
+      "loss": 1.584,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.025586353944563e-05,
+      "loss": 1.4883,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0277185501066098e-05,
+      "loss": 1.3262,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.029850746268657e-05,
+      "loss": 1.542,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0319829424307037e-05,
+      "loss": 1.6465,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0341151385927506e-05,
+      "loss": 1.7324,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0362473347547975e-05,
+      "loss": 1.2461,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0383795309168445e-05,
+      "loss": 1.1982,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0405117270788913e-05,
+      "loss": 1.5371,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0426439232409382e-05,
+      "loss": 1.7637,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0447761194029851e-05,
+      "loss": 1.5723,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0469083155650322e-05,
+      "loss": 1.3281,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.049040511727079e-05,
+      "loss": 1.5566,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0511727078891258e-05,
+      "loss": 1.5605,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0533049040511727e-05,
+      "loss": 1.4688,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0554371002132198e-05,
+      "loss": 1.2881,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0575692963752667e-05,
+      "loss": 1.2656,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0597014925373134e-05,
+      "loss": 1.4785,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0618336886993603e-05,
+      "loss": 1.4199,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0639658848614074e-05,
+      "loss": 1.6553,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0660980810234543e-05,
+      "loss": 1.9473,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.068230277185501e-05,
+      "loss": 1.5742,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0703624733475481e-05,
+      "loss": 1.4492,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.072494669509595e-05,
+      "loss": 1.5352,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.074626865671642e-05,
+      "loss": 1.498,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0767590618336887e-05,
+      "loss": 1.3438,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0788912579957357e-05,
+      "loss": 1.7617,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0810234541577827e-05,
+      "loss": 1.5391,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0831556503198296e-05,
+      "loss": 1.9609,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0852878464818763e-05,
+      "loss": 1.4922,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0874200426439234e-05,
+      "loss": 1.7148,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0895522388059703e-05,
+      "loss": 1.0469,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0916844349680172e-05,
+      "loss": 1.8965,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0938166311300639e-05,
+      "loss": 1.3057,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.095948827292111e-05,
+      "loss": 1.9414,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0980810234541579e-05,
+      "loss": 1.2637,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1002132196162048e-05,
+      "loss": 1.9277,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1023454157782515e-05,
+      "loss": 1.6211,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1044776119402986e-05,
+      "loss": 1.3848,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1066098081023455e-05,
+      "loss": 1.9375,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1087420042643924e-05,
+      "loss": 1.5381,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1108742004264392e-05,
+      "loss": 2.1562,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1130063965884862e-05,
+      "loss": 1.1738,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1151385927505331e-05,
+      "loss": 1.6523,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.11727078891258e-05,
+      "loss": 1.9941,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1194029850746268e-05,
+      "loss": 1.1494,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1215351812366738e-05,
+      "loss": 1.3633,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1236673773987208e-05,
+      "loss": 1.9551,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1257995735607677e-05,
+      "loss": 1.7051,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1279317697228146e-05,
+      "loss": 1.6943,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1300639658848616e-05,
+      "loss": 1.4277,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1321961620469084e-05,
+      "loss": 1.4111,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1343283582089553e-05,
+      "loss": 1.9219,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1364605543710024e-05,
+      "loss": 1.625,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1385927505330493e-05,
+      "loss": 1.8184,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.140724946695096e-05,
+      "loss": 1.7129,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 1.4502,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.14498933901919e-05,
+      "loss": 1.5,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1471215351812369e-05,
+      "loss": 1.7344,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1492537313432836e-05,
+      "loss": 1.5449,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1513859275053305e-05,
+      "loss": 1.4219,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1535181236673776e-05,
+      "loss": 1.3965,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1556503198294245e-05,
+      "loss": 1.4404,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1577825159914712e-05,
+      "loss": 1.8281,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1599147121535181e-05,
+      "loss": 2.0996,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1620469083155652e-05,
+      "loss": 1.2324,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1641791044776121e-05,
+      "loss": 1.3252,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1663113006396589e-05,
+      "loss": 1.0215,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1684434968017058e-05,
+      "loss": 1.417,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1705756929637528e-05,
+      "loss": 1.6152,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1727078891257997e-05,
+      "loss": 1.4668,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1748400852878465e-05,
+      "loss": 1.249,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1769722814498934e-05,
+      "loss": 1.2012,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1791044776119405e-05,
+      "loss": 1.7266,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1812366737739874e-05,
+      "loss": 1.666,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1833688699360341e-05,
+      "loss": 1.5801,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.185501066098081e-05,
+      "loss": 1.4668,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.187633262260128e-05,
+      "loss": 1.4121,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.189765458422175e-05,
+      "loss": 1.3789,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1918976545842217e-05,
+      "loss": 1.2998,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1940298507462686e-05,
+      "loss": 1.5781,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1961620469083157e-05,
+      "loss": 1.5879,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1982942430703626e-05,
+      "loss": 1.4424,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2004264392324093e-05,
+      "loss": 1.6074,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2025586353944564e-05,
+      "loss": 1.873,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2046908315565033e-05,
+      "loss": 1.3848,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2068230277185502e-05,
+      "loss": 1.0791,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2089552238805971e-05,
+      "loss": 1.541,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2110874200426442e-05,
+      "loss": 1.7402,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.213219616204691e-05,
+      "loss": 1.6621,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2153518123667378e-05,
+      "loss": 1.8262,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2174840085287848e-05,
+      "loss": 1.3379,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2196162046908318e-05,
+      "loss": 1.7246,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2217484008528786e-05,
+      "loss": 1.6582,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2238805970149255e-05,
+      "loss": 1.7617,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2260127931769724e-05,
+      "loss": 0.9629,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2281449893390195e-05,
+      "loss": 1.8301,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2302771855010662e-05,
+      "loss": 1.5361,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2324093816631131e-05,
+      "loss": 1.332,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.23454157782516e-05,
+      "loss": 1.3789,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.236673773987207e-05,
+      "loss": 1.2734,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2388059701492538e-05,
+      "loss": 1.4062,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2409381663113007e-05,
+      "loss": 1.5264,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2430703624733476e-05,
+      "loss": 1.2891,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2452025586353947e-05,
+      "loss": 1.8926,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2473347547974414e-05,
+      "loss": 1.5117,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2494669509594883e-05,
+      "loss": 1.1084,
+      "step": 586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2515991471215352e-05,
+      "loss": 1.4824,
+      "step": 587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2537313432835823e-05,
+      "loss": 1.3652,
+      "step": 588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.255863539445629e-05,
+      "loss": 1.5742,
+      "step": 589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.257995735607676e-05,
+      "loss": 1.7012,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2601279317697229e-05,
+      "loss": 1.6572,
+      "step": 591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.26226012793177e-05,
+      "loss": 1.2275,
+      "step": 592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2643923240938167e-05,
+      "loss": 1.6621,
+      "step": 593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2665245202558636e-05,
+      "loss": 1.665,
+      "step": 594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2686567164179107e-05,
+      "loss": 1.459,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2707889125799576e-05,
+      "loss": 1.8535,
+      "step": 596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2729211087420043e-05,
+      "loss": 1.2617,
+      "step": 597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2750533049040512e-05,
+      "loss": 1.623,
+      "step": 598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2771855010660983e-05,
+      "loss": 1.584,
+      "step": 599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2793176972281452e-05,
+      "loss": 1.168,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2814498933901919e-05,
+      "loss": 1.2646,
+      "step": 601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2835820895522388e-05,
+      "loss": 1.5957,
+      "step": 602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2857142857142859e-05,
+      "loss": 1.541,
+      "step": 603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2878464818763328e-05,
+      "loss": 1.0801,
+      "step": 604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2899786780383795e-05,
+      "loss": 1.8008,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2921108742004264e-05,
+      "loss": 1.7363,
+      "step": 606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2942430703624735e-05,
+      "loss": 1.6211,
+      "step": 607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2963752665245204e-05,
+      "loss": 1.5137,
+      "step": 608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2985074626865673e-05,
+      "loss": 1.6738,
+      "step": 609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.300639658848614e-05,
+      "loss": 1.8193,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3027718550106611e-05,
+      "loss": 1.6953,
+      "step": 611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.304904051172708e-05,
+      "loss": 1.4629,
+      "step": 612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.307036247334755e-05,
+      "loss": 1.6621,
+      "step": 613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3091684434968017e-05,
+      "loss": 1.4863,
+      "step": 614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3113006396588488e-05,
+      "loss": 1.7959,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3134328358208957e-05,
+      "loss": 1.4814,
+      "step": 616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3155650319829426e-05,
+      "loss": 1.3115,
+      "step": 617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3176972281449893e-05,
+      "loss": 1.293,
+      "step": 618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3198294243070364e-05,
+      "loss": 1.5605,
+      "step": 619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3219616204690833e-05,
+      "loss": 1.3955,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3240938166311302e-05,
+      "loss": 1.208,
+      "step": 621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.326226012793177e-05,
+      "loss": 1.5977,
+      "step": 622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.328358208955224e-05,
+      "loss": 1.5234,
+      "step": 623
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3304904051172709e-05,
+      "loss": 1.5967,
+      "step": 624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3326226012793178e-05,
+      "loss": 1.4668,
+      "step": 625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3347547974413649e-05,
+      "loss": 1.5723,
+      "step": 626
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3368869936034116e-05,
+      "loss": 1.1758,
+      "step": 627
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3390191897654585e-05,
+      "loss": 1.9609,
+      "step": 628
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3411513859275054e-05,
+      "loss": 1.5059,
+      "step": 629
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3432835820895525e-05,
+      "loss": 1.5254,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3454157782515992e-05,
+      "loss": 1.501,
+      "step": 631
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3475479744136461e-05,
+      "loss": 1.873,
+      "step": 632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.349680170575693e-05,
+      "loss": 1.5293,
+      "step": 633
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3518123667377401e-05,
+      "loss": 1.5547,
+      "step": 634
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3539445628997869e-05,
+      "loss": 2.1055,
+      "step": 635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3560767590618338e-05,
+      "loss": 1.4863,
+      "step": 636
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3582089552238807e-05,
+      "loss": 1.5225,
+      "step": 637
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3603411513859277e-05,
+      "loss": 1.8672,
+      "step": 638
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3624733475479745e-05,
+      "loss": 1.2773,
+      "step": 639
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3646055437100214e-05,
+      "loss": 1.5156,
+      "step": 640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3667377398720683e-05,
+      "loss": 1.2129,
+      "step": 641
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3688699360341154e-05,
+      "loss": 1.625,
+      "step": 642
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3710021321961621e-05,
+      "loss": 1.7891,
+      "step": 643
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.373134328358209e-05,
+      "loss": 1.2617,
+      "step": 644
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3752665245202559e-05,
+      "loss": 1.4785,
+      "step": 645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.377398720682303e-05,
+      "loss": 1.502,
+      "step": 646
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3795309168443499e-05,
+      "loss": 1.6387,
+      "step": 647
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3816631130063966e-05,
+      "loss": 1.3555,
+      "step": 648
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3837953091684435e-05,
+      "loss": 1.3398,
+      "step": 649
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3859275053304906e-05,
+      "loss": 1.7715,
+      "step": 650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3880597014925375e-05,
+      "loss": 1.5801,
+      "step": 651
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3901918976545842e-05,
+      "loss": 1.6348,
+      "step": 652
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3923240938166312e-05,
+      "loss": 1.332,
+      "step": 653
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3944562899786782e-05,
+      "loss": 1.5625,
+      "step": 654
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3965884861407251e-05,
+      "loss": 1.2715,
+      "step": 655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3987206823027719e-05,
+      "loss": 1.4023,
+      "step": 656
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.400852878464819e-05,
+      "loss": 1.9688,
+      "step": 657
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4029850746268658e-05,
+      "loss": 1.292,
+      "step": 658
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4051172707889128e-05,
+      "loss": 1.9434,
+      "step": 659
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4072494669509595e-05,
+      "loss": 1.2803,
+      "step": 660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4093816631130066e-05,
+      "loss": 1.4375,
+      "step": 661
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4115138592750535e-05,
+      "loss": 1.6738,
+      "step": 662
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4136460554371004e-05,
+      "loss": 1.8594,
+      "step": 663
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4157782515991471e-05,
+      "loss": 1.5586,
+      "step": 664
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4179104477611942e-05,
+      "loss": 1.2529,
+      "step": 665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4200426439232411e-05,
+      "loss": 1.5684,
+      "step": 666
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.422174840085288e-05,
+      "loss": 1.332,
+      "step": 667
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4243070362473347e-05,
+      "loss": 1.3457,
+      "step": 668
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4264392324093818e-05,
+      "loss": 1.541,
+      "step": 669
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4285714285714287e-05,
+      "loss": 1.5312,
+      "step": 670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4307036247334756e-05,
+      "loss": 1.2998,
+      "step": 671
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4328358208955224e-05,
+      "loss": 1.8398,
+      "step": 672
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4349680170575694e-05,
+      "loss": 1.1826,
+      "step": 673
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4371002132196163e-05,
+      "loss": 1.623,
+      "step": 674
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4392324093816632e-05,
+      "loss": 2.1152,
+      "step": 675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.44136460554371e-05,
+      "loss": 1.5254,
+      "step": 676
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.443496801705757e-05,
+      "loss": 1.5049,
+      "step": 677
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.445628997867804e-05,
+      "loss": 1.4473,
+      "step": 678
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4477611940298509e-05,
+      "loss": 1.2803,
+      "step": 679
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4498933901918976e-05,
+      "loss": 1.3535,
+      "step": 680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4520255863539447e-05,
+      "loss": 1.3184,
+      "step": 681
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4541577825159916e-05,
+      "loss": 1.4121,
+      "step": 682
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4562899786780385e-05,
+      "loss": 1.4629,
+      "step": 683
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4584221748400854e-05,
+      "loss": 1.7578,
+      "step": 684
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4605543710021325e-05,
+      "loss": 1.3467,
+      "step": 685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4626865671641792e-05,
+      "loss": 1.7461,
+      "step": 686
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4648187633262261e-05,
+      "loss": 1.3311,
+      "step": 687
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4669509594882732e-05,
+      "loss": 1.627,
+      "step": 688
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.46908315565032e-05,
+      "loss": 1.3574,
+      "step": 689
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4712153518123668e-05,
+      "loss": 1.6377,
+      "step": 690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4733475479744137e-05,
+      "loss": 1.3594,
+      "step": 691
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4754797441364608e-05,
+      "loss": 1.5176,
+      "step": 692
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4776119402985077e-05,
+      "loss": 1.7812,
+      "step": 693
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4797441364605544e-05,
+      "loss": 1.1934,
+      "step": 694
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4818763326226013e-05,
+      "loss": 1.6113,
+      "step": 695
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4840085287846484e-05,
+      "loss": 1.7051,
+      "step": 696
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4861407249466953e-05,
+      "loss": 1.376,
+      "step": 697
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.488272921108742e-05,
+      "loss": 1.6689,
+      "step": 698
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.490405117270789e-05,
+      "loss": 1.123,
+      "step": 699
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.492537313432836e-05,
+      "loss": 1.5654,
+      "step": 700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.494669509594883e-05,
+      "loss": 1.2617,
+      "step": 701
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4968017057569297e-05,
+      "loss": 1.4121,
+      "step": 702
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4989339019189766e-05,
+      "loss": 1.3857,
+      "step": 703
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5010660980810237e-05,
+      "loss": 1.8535,
+      "step": 704
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5031982942430706e-05,
+      "loss": 1.7656,
+      "step": 705
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5053304904051173e-05,
+      "loss": 1.5527,
+      "step": 706
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5074626865671642e-05,
+      "loss": 1.5859,
+      "step": 707
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5095948827292113e-05,
+      "loss": 1.4082,
+      "step": 708
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5117270788912582e-05,
+      "loss": 1.2695,
+      "step": 709
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.513859275053305e-05,
+      "loss": 1.916,
+      "step": 710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5159914712153518e-05,
+      "loss": 1.5625,
+      "step": 711
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5181236673773989e-05,
+      "loss": 1.1904,
+      "step": 712
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5202558635394458e-05,
+      "loss": 1.6992,
+      "step": 713
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5223880597014925e-05,
+      "loss": 1.9023,
+      "step": 714
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5245202558635394e-05,
+      "loss": 1.4727,
+      "step": 715
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5266524520255865e-05,
+      "loss": 1.6172,
+      "step": 716
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5287846481876334e-05,
+      "loss": 1.6855,
+      "step": 717
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5309168443496803e-05,
+      "loss": 1.4668,
+      "step": 718
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5330490405117272e-05,
+      "loss": 1.541,
+      "step": 719
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.535181236673774e-05,
+      "loss": 1.7227,
+      "step": 720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.537313432835821e-05,
+      "loss": 1.4277,
+      "step": 721
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.539445628997868e-05,
+      "loss": 1.8008,
+      "step": 722
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.541577825159915e-05,
+      "loss": 1.2061,
+      "step": 723
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5437100213219618e-05,
+      "loss": 1.7793,
+      "step": 724
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5458422174840087e-05,
+      "loss": 1.4473,
+      "step": 725
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5479744136460556e-05,
+      "loss": 1.8008,
+      "step": 726
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5501066098081025e-05,
+      "loss": 1.6582,
+      "step": 727
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5522388059701494e-05,
+      "loss": 1.627,
+      "step": 728
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5543710021321963e-05,
+      "loss": 1.5586,
+      "step": 729
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5565031982942432e-05,
+      "loss": 1.4844,
+      "step": 730
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.55863539445629e-05,
+      "loss": 1.334,
+      "step": 731
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.560767590618337e-05,
+      "loss": 1.3848,
+      "step": 732
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.562899786780384e-05,
+      "loss": 1.2539,
+      "step": 733
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5650319829424308e-05,
+      "loss": 1.0684,
+      "step": 734
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5671641791044777e-05,
+      "loss": 1.1299,
+      "step": 735
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5692963752665246e-05,
+      "loss": 1.4316,
+      "step": 736
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5714285714285715e-05,
+      "loss": 1.3027,
+      "step": 737
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5735607675906184e-05,
+      "loss": 1.3477,
+      "step": 738
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5756929637526653e-05,
+      "loss": 1.918,
+      "step": 739
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5778251599147122e-05,
+      "loss": 1.502,
+      "step": 740
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.579957356076759e-05,
+      "loss": 1.7129,
+      "step": 741
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.582089552238806e-05,
+      "loss": 1.5723,
+      "step": 742
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.584221748400853e-05,
+      "loss": 1.5146,
+      "step": 743
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5863539445629e-05,
+      "loss": 1.4609,
+      "step": 744
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5884861407249468e-05,
+      "loss": 1.4805,
+      "step": 745
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5906183368869937e-05,
+      "loss": 1.6875,
+      "step": 746
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5927505330490406e-05,
+      "loss": 1.791,
+      "step": 747
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5948827292110875e-05,
+      "loss": 1.3281,
+      "step": 748
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5970149253731344e-05,
+      "loss": 1.7715,
+      "step": 749
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5991471215351813e-05,
+      "loss": 1.7793,
+      "step": 750
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6012793176972282e-05,
+      "loss": 1.5293,
+      "step": 751
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.603411513859275e-05,
+      "loss": 1.6191,
+      "step": 752
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.605543710021322e-05,
+      "loss": 1.6289,
+      "step": 753
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6076759061833693e-05,
+      "loss": 1.6104,
+      "step": 754
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6098081023454158e-05,
+      "loss": 1.7852,
+      "step": 755
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6119402985074627e-05,
+      "loss": 1.8203,
+      "step": 756
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6140724946695096e-05,
+      "loss": 1.1611,
+      "step": 757
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.616204690831557e-05,
+      "loss": 1.5703,
+      "step": 758
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6183368869936034e-05,
+      "loss": 1.9199,
+      "step": 759
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6204690831556504e-05,
+      "loss": 1.6426,
+      "step": 760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6226012793176973e-05,
+      "loss": 1.584,
+      "step": 761
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6247334754797445e-05,
+      "loss": 1.8223,
+      "step": 762
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.626865671641791e-05,
+      "loss": 1.2832,
+      "step": 763
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.628997867803838e-05,
+      "loss": 1.4863,
+      "step": 764
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.631130063965885e-05,
+      "loss": 1.2471,
+      "step": 765
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.633262260127932e-05,
+      "loss": 1.9082,
+      "step": 766
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6353944562899787e-05,
+      "loss": 1.5586,
+      "step": 767
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6375266524520256e-05,
+      "loss": 1.3955,
+      "step": 768
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6396588486140725e-05,
+      "loss": 1.4111,
+      "step": 769
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6417910447761197e-05,
+      "loss": 1.4072,
+      "step": 770
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6439232409381663e-05,
+      "loss": 1.4365,
+      "step": 771
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6460554371002132e-05,
+      "loss": 1.6328,
+      "step": 772
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.64818763326226e-05,
+      "loss": 1.582,
+      "step": 773
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6503198294243074e-05,
+      "loss": 1.2666,
+      "step": 774
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.652452025586354e-05,
+      "loss": 1.7734,
+      "step": 775
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.654584221748401e-05,
+      "loss": 1.8066,
+      "step": 776
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6567164179104477e-05,
+      "loss": 1.9023,
+      "step": 777
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.658848614072495e-05,
+      "loss": 1.6465,
+      "step": 778
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6609808102345416e-05,
+      "loss": 1.0752,
+      "step": 779
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6631130063965885e-05,
+      "loss": 1.6738,
+      "step": 780
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6652452025586354e-05,
+      "loss": 1.5518,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6673773987206826e-05,
+      "loss": 1.9922,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6695095948827295e-05,
+      "loss": 1.6396,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.671641791044776e-05,
+      "loss": 1.6406,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6737739872068233e-05,
+      "loss": 1.5342,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6759061833688702e-05,
+      "loss": 1.5645,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.678038379530917e-05,
+      "loss": 1.4707,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6801705756929637e-05,
+      "loss": 1.375,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.682302771855011e-05,
+      "loss": 1.0029,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.684434968017058e-05,
+      "loss": 1.3818,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6865671641791048e-05,
+      "loss": 1.9668,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6886993603411513e-05,
+      "loss": 1.4238,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6908315565031986e-05,
+      "loss": 1.5664,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6929637526652455e-05,
+      "loss": 1.541,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6950959488272924e-05,
+      "loss": 1.2549,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.697228144989339e-05,
+      "loss": 1.4824,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6993603411513862e-05,
+      "loss": 1.7578,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.701492537313433e-05,
+      "loss": 1.6289,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.70362473347548e-05,
+      "loss": 1.4131,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7057569296375266e-05,
+      "loss": 1.752,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7078891257995738e-05,
+      "loss": 1.5781,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7100213219616207e-05,
+      "loss": 1.6211,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7121535181236676e-05,
+      "loss": 1.8438,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7142857142857142e-05,
+      "loss": 1.5664,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7164179104477614e-05,
+      "loss": 1.5273,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7185501066098083e-05,
+      "loss": 1.1318,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7206823027718552e-05,
+      "loss": 1.7227,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7228144989339018e-05,
+      "loss": 1.4492,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.724946695095949e-05,
+      "loss": 1.7715,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.727078891257996e-05,
+      "loss": 1.7305,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.729211087420043e-05,
+      "loss": 1.0859,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7313432835820894e-05,
+      "loss": 1.4277,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7334754797441367e-05,
+      "loss": 1.583,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7356076759061836e-05,
+      "loss": 1.1855,
+      "step": 814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7377398720682305e-05,
+      "loss": 1.9453,
+      "step": 815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7398720682302774e-05,
+      "loss": 1.3408,
+      "step": 816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7420042643923243e-05,
+      "loss": 1.8906,
+      "step": 817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7441364605543712e-05,
+      "loss": 1.7129,
+      "step": 818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.746268656716418e-05,
+      "loss": 1.4082,
+      "step": 819
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.748400852878465e-05,
+      "loss": 1.2754,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.750533049040512e-05,
+      "loss": 1.6426,
+      "step": 821
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7526652452025588e-05,
+      "loss": 1.498,
+      "step": 822
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7547974413646057e-05,
+      "loss": 1.6035,
+      "step": 823
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7569296375266526e-05,
+      "loss": 1.3984,
+      "step": 824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7590618336886995e-05,
+      "loss": 1.9629,
+      "step": 825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7611940298507464e-05,
+      "loss": 1.3447,
+      "step": 826
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7633262260127933e-05,
+      "loss": 1.5859,
+      "step": 827
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7654584221748402e-05,
+      "loss": 1.5977,
+      "step": 828
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.767590618336887e-05,
+      "loss": 1.8086,
+      "step": 829
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.769722814498934e-05,
+      "loss": 1.4961,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.771855010660981e-05,
+      "loss": 1.8184,
+      "step": 831
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.773987206823028e-05,
+      "loss": 1.2705,
+      "step": 832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7761194029850748e-05,
+      "loss": 1.6445,
+      "step": 833
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7782515991471217e-05,
+      "loss": 1.4883,
+      "step": 834
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7803837953091686e-05,
+      "loss": 1.9785,
+      "step": 835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7825159914712155e-05,
+      "loss": 1.6152,
+      "step": 836
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7846481876332624e-05,
+      "loss": 1.2295,
+      "step": 837
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7867803837953093e-05,
+      "loss": 1.5898,
+      "step": 838
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7889125799573562e-05,
+      "loss": 1.375,
+      "step": 839
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.791044776119403e-05,
+      "loss": 1.6064,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.79317697228145e-05,
+      "loss": 1.7285,
+      "step": 841
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.795309168443497e-05,
+      "loss": 1.502,
+      "step": 842
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7974413646055438e-05,
+      "loss": 1.4766,
+      "step": 843
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7995735607675907e-05,
+      "loss": 1.4375,
+      "step": 844
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8017057569296376e-05,
+      "loss": 1.4941,
+      "step": 845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8038379530916845e-05,
+      "loss": 1.4482,
+      "step": 846
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8059701492537314e-05,
+      "loss": 1.6797,
+      "step": 847
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8081023454157784e-05,
+      "loss": 1.4141,
+      "step": 848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8102345415778253e-05,
+      "loss": 1.1572,
+      "step": 849
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.812366737739872e-05,
+      "loss": 1.8555,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.814498933901919e-05,
+      "loss": 1.1836,
+      "step": 851
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.816631130063966e-05,
+      "loss": 1.1191,
+      "step": 852
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.818763326226013e-05,
+      "loss": 1.3184,
+      "step": 853
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8208955223880598e-05,
+      "loss": 1.5918,
+      "step": 854
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8230277185501067e-05,
+      "loss": 1.4473,
+      "step": 855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8251599147121536e-05,
+      "loss": 1.9512,
+      "step": 856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8272921108742005e-05,
+      "loss": 1.5586,
+      "step": 857
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8294243070362474e-05,
+      "loss": 1.3545,
+      "step": 858
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8315565031982943e-05,
+      "loss": 1.4287,
+      "step": 859
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8336886993603412e-05,
+      "loss": 1.5195,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.835820895522388e-05,
+      "loss": 1.7549,
+      "step": 861
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.837953091684435e-05,
+      "loss": 1.7539,
+      "step": 862
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8400852878464823e-05,
+      "loss": 1.3311,
+      "step": 863
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.842217484008529e-05,
+      "loss": 1.4395,
+      "step": 864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8443496801705757e-05,
+      "loss": 1.5518,
+      "step": 865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8464818763326226e-05,
+      "loss": 1.5879,
+      "step": 866
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.84861407249467e-05,
+      "loss": 1.4805,
+      "step": 867
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8507462686567165e-05,
+      "loss": 1.6963,
+      "step": 868
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8528784648187634e-05,
+      "loss": 1.5137,
+      "step": 869
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8550106609808103e-05,
+      "loss": 1.3652,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8571428571428575e-05,
+      "loss": 1.3906,
+      "step": 871
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.859275053304904e-05,
+      "loss": 1.5898,
+      "step": 872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.861407249466951e-05,
+      "loss": 1.248,
+      "step": 873
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.863539445628998e-05,
+      "loss": 1.4834,
+      "step": 874
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.865671641791045e-05,
+      "loss": 1.5488,
+      "step": 875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8678038379530917e-05,
+      "loss": 1.6016,
+      "step": 876
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8699360341151386e-05,
+      "loss": 1.1719,
+      "step": 877
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.872068230277186e-05,
+      "loss": 1.6064,
+      "step": 878
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8742004264392328e-05,
+      "loss": 1.8262,
+      "step": 879
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8763326226012793e-05,
+      "loss": 1.8613,
+      "step": 880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8784648187633262e-05,
+      "loss": 1.2793,
+      "step": 881
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8805970149253735e-05,
+      "loss": 1.166,
+      "step": 882
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8827292110874204e-05,
+      "loss": 1.208,
+      "step": 883
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.884861407249467e-05,
+      "loss": 1.3662,
+      "step": 884
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.886993603411514e-05,
+      "loss": 1.4707,
+      "step": 885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.889125799573561e-05,
+      "loss": 1.457,
+      "step": 886
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.891257995735608e-05,
+      "loss": 1.8379,
+      "step": 887
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8933901918976546e-05,
+      "loss": 1.7871,
+      "step": 888
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8955223880597015e-05,
+      "loss": 1.5605,
+      "step": 889
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8976545842217487e-05,
+      "loss": 1.6914,
+      "step": 890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8997867803837956e-05,
+      "loss": 1.1953,
+      "step": 891
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9019189765458422e-05,
+      "loss": 1.5762,
+      "step": 892
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.904051172707889e-05,
+      "loss": 1.6875,
+      "step": 893
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9061833688699363e-05,
+      "loss": 1.5293,
+      "step": 894
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9083155650319832e-05,
+      "loss": 1.9082,
+      "step": 895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9104477611940298e-05,
+      "loss": 1.3047,
+      "step": 896
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9125799573560767e-05,
+      "loss": 1.957,
+      "step": 897
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.914712153518124e-05,
+      "loss": 1.8008,
+      "step": 898
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.916844349680171e-05,
+      "loss": 1.2656,
+      "step": 899
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9189765458422178e-05,
+      "loss": 1.7109,
+      "step": 900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9211087420042643e-05,
+      "loss": 1.4531,
+      "step": 901
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9232409381663116e-05,
+      "loss": 1.7471,
+      "step": 902
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9253731343283585e-05,
+      "loss": 1.7578,
+      "step": 903
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9275053304904054e-05,
+      "loss": 1.0596,
+      "step": 904
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.929637526652452e-05,
+      "loss": 1.8398,
+      "step": 905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9317697228144992e-05,
+      "loss": 1.3359,
+      "step": 906
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.933901918976546e-05,
+      "loss": 1.6543,
+      "step": 907
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.936034115138593e-05,
+      "loss": 1.7637,
+      "step": 908
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.93816631130064e-05,
+      "loss": 2.1211,
+      "step": 909
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9402985074626868e-05,
+      "loss": 1.8027,
+      "step": 910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9424307036247337e-05,
+      "loss": 1.7324,
+      "step": 911
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9445628997867806e-05,
+      "loss": 1.7129,
+      "step": 912
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9466950959488275e-05,
+      "loss": 2.3145,
+      "step": 913
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9488272921108744e-05,
+      "loss": 1.3848,
+      "step": 914
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9509594882729213e-05,
+      "loss": 1.665,
+      "step": 915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9530916844349682e-05,
+      "loss": 1.5938,
+      "step": 916
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.955223880597015e-05,
+      "loss": 1.5547,
+      "step": 917
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.957356076759062e-05,
+      "loss": 1.5654,
+      "step": 918
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.959488272921109e-05,
+      "loss": 1.5176,
+      "step": 919
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.961620469083156e-05,
+      "loss": 1.1338,
+      "step": 920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9637526652452028e-05,
+      "loss": 1.3809,
+      "step": 921
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9658848614072497e-05,
+      "loss": 1.5713,
+      "step": 922
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9680170575692966e-05,
+      "loss": 1.6445,
+      "step": 923
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9701492537313435e-05,
+      "loss": 1.6279,
+      "step": 924
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9722814498933904e-05,
+      "loss": 1.4678,
+      "step": 925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9744136460554373e-05,
+      "loss": 1.668,
+      "step": 926
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9765458422174842e-05,
+      "loss": 1.5273,
+      "step": 927
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978678038379531e-05,
+      "loss": 1.8281,
+      "step": 928
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980810234541578e-05,
+      "loss": 2.0117,
+      "step": 929
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982942430703625e-05,
+      "loss": 1.4775,
+      "step": 930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9850746268656718e-05,
+      "loss": 1.5566,
+      "step": 931
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9872068230277187e-05,
+      "loss": 1.4844,
+      "step": 932
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9893390191897656e-05,
+      "loss": 1.3857,
+      "step": 933
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9914712153518125e-05,
+      "loss": 1.7676,
+      "step": 934
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9936034115138594e-05,
+      "loss": 1.2324,
+      "step": 935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9957356076759064e-05,
+      "loss": 1.499,
+      "step": 936
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9978678038379533e-05,
+      "loss": 1.5381,
+      "step": 937
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2e-05,
+      "loss": 1.2949,
+      "step": 938
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999946291807e-05,
+      "loss": 1.6191,
+      "step": 939
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999785167218e-05,
+      "loss": 1.2305,
+      "step": 940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999516626244e-05,
+      "loss": 1.6406,
+      "step": 941
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999140668878e-05,
+      "loss": 1.6914,
+      "step": 942
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999865729513e-05,
+      "loss": 1.5244,
+      "step": 943
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999998066505005e-05,
+      "loss": 1.9258,
+      "step": 944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999997368298514e-05,
+      "loss": 1.5332,
+      "step": 945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999996562675653e-05,
+      "loss": 1.5254,
+      "step": 946
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999564963644e-05,
+      "loss": 1.8418,
+      "step": 947
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999994629180883e-05,
+      "loss": 1.6406,
+      "step": 948
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999993501308988e-05,
+      "loss": 1.5488,
+      "step": 949
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999992266020773e-05,
+      "loss": 2.0352,
+      "step": 950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999092331625e-05,
+      "loss": 1.0703,
+      "step": 951
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999989473195432e-05,
+      "loss": 1.4004,
+      "step": 952
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999987915658333e-05,
+      "loss": 1.4902,
+      "step": 953
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999986250704974e-05,
+      "loss": 1.3604,
+      "step": 954
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999984478335373e-05,
+      "loss": 1.4902,
+      "step": 955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999982598549543e-05,
+      "loss": 1.4736,
+      "step": 956
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998061134751e-05,
+      "loss": 1.3184,
+      "step": 957
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999978516729293e-05,
+      "loss": 1.2705,
+      "step": 958
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999976314694915e-05,
+      "loss": 1.6523,
+      "step": 959
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999974005244395e-05,
+      "loss": 1.3975,
+      "step": 960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999971588377768e-05,
+      "loss": 1.5527,
+      "step": 961
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999969064095054e-05,
+      "loss": 1.7637,
+      "step": 962
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999966432396278e-05,
+      "loss": 1.4746,
+      "step": 963
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999963693281472e-05,
+      "loss": 1.707,
+      "step": 964
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996084675066e-05,
+      "loss": 1.8066,
+      "step": 965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999957892803884e-05,
+      "loss": 1.418,
+      "step": 966
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999954831441164e-05,
+      "loss": 1.291,
+      "step": 967
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999995166266254e-05,
+      "loss": 1.6016,
+      "step": 968
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999948386468038e-05,
+      "loss": 1.5312,
+      "step": 969
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999945002857704e-05,
+      "loss": 1.8125,
+      "step": 970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999941511831564e-05,
+      "loss": 1.0977,
+      "step": 971
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999937913389663e-05,
+      "loss": 1.4766,
+      "step": 972
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999934207532038e-05,
+      "loss": 1.5039,
+      "step": 973
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999993039425873e-05,
+      "loss": 1.6289,
+      "step": 974
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999926473569774e-05,
+      "loss": 1.7852,
+      "step": 975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999922445465213e-05,
+      "loss": 1.4434,
+      "step": 976
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999918309945097e-05,
+      "loss": 1.2539,
+      "step": 977
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999914067009467e-05,
+      "loss": 1.6602,
+      "step": 978
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999909716658367e-05,
+      "loss": 1.9492,
+      "step": 979
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999905258891844e-05,
+      "loss": 1.2559,
+      "step": 980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999990069370995e-05,
+      "loss": 1.793,
+      "step": 981
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999896021112727e-05,
+      "loss": 1.4512,
+      "step": 982
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999989124110023e-05,
+      "loss": 1.6064,
+      "step": 983
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999988635367251e-05,
+      "loss": 1.5527,
+      "step": 984
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999988135882962e-05,
+      "loss": 1.4873,
+      "step": 985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999876256571612e-05,
+      "loss": 1.6309,
+      "step": 986
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999987104689854e-05,
+      "loss": 1.6533,
+      "step": 987
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999986572981046e-05,
+      "loss": 1.0898,
+      "step": 988
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999860305307434e-05,
+      "loss": 1.7031,
+      "step": 989
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999854773389516e-05,
+      "loss": 1.3164,
+      "step": 990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999849134056762e-05,
+      "loss": 1.3076,
+      "step": 991
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999843387309243e-05,
+      "loss": 1.4785,
+      "step": 992
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999983753314701e-05,
+      "loss": 1.335,
+      "step": 993
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999983157157013e-05,
+      "loss": 1.3867,
+      "step": 994
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999982550257867e-05,
+      "loss": 1.4902,
+      "step": 995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999981932617269e-05,
+      "loss": 1.2783,
+      "step": 996
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999981304235226e-05,
+      "loss": 1.5957,
+      "step": 997
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999806651117446e-05,
+      "loss": 1.5918,
+      "step": 998
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999800152468317e-05,
+      "loss": 1.3643,
+      "step": 999
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999979354640494e-05,
+      "loss": 1.4102,
+      "step": 1000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999786832927395e-05,
+      "loss": 1.5703,
+      "step": 1001
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999978001203574e-05,
+      "loss": 1.3262,
+      "step": 1002
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999977308373006e-05,
+      "loss": 1.4873,
+      "step": 1003
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999976604801043e-05,
+      "loss": 1.6152,
+      "step": 1004
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999975890487692e-05,
+      "loss": 1.208,
+      "step": 1005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99997516543296e-05,
+      "loss": 1.6973,
+      "step": 1006
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999744296368563e-05,
+      "loss": 1.459,
+      "step": 1007
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999736830993876e-05,
+      "loss": 1.6328,
+      "step": 1008
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999729258205628e-05,
+      "loss": 1.1152,
+      "step": 1009
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999721578003894e-05,
+      "loss": 1.7559,
+      "step": 1010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999971379038876e-05,
+      "loss": 1.4727,
+      "step": 1011
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999970589536031e-05,
+      "loss": 1.6152,
+      "step": 1012
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999697892918622e-05,
+      "loss": 1.332,
+      "step": 1013
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999968978306379e-05,
+      "loss": 1.2031,
+      "step": 1014
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99996815657959e-05,
+      "loss": 1.5547,
+      "step": 1015
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999673241115037e-05,
+      "loss": 1.4668,
+      "step": 1016
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999664809021293e-05,
+      "loss": 1.2656,
+      "step": 1017
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999656269514758e-05,
+      "loss": 1.7832,
+      "step": 1018
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999964762259552e-05,
+      "loss": 1.375,
+      "step": 1019
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999638868263677e-05,
+      "loss": 1.6602,
+      "step": 1020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999630006519324e-05,
+      "loss": 1.1699,
+      "step": 1021
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999962103736255e-05,
+      "loss": 1.4395,
+      "step": 1022
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999611960793457e-05,
+      "loss": 1.3506,
+      "step": 1023
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999960277681214e-05,
+      "loss": 1.4922,
+      "step": 1024
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999593485418697e-05,
+      "loss": 1.5078,
+      "step": 1025
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999584086613226e-05,
+      "loss": 1.6816,
+      "step": 1026
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999574580395835e-05,
+      "loss": 1.7598,
+      "step": 1027
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999956496676662e-05,
+      "loss": 1.4238,
+      "step": 1028
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999555245725685e-05,
+      "loss": 1.2236,
+      "step": 1029
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999545417273136e-05,
+      "loss": 1.2891,
+      "step": 1030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999953548140908e-05,
+      "loss": 1.4043,
+      "step": 1031
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999952543813362e-05,
+      "loss": 1.0508,
+      "step": 1032
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999515287446866e-05,
+      "loss": 1.5488,
+      "step": 1033
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999505029348927e-05,
+      "loss": 1.2119,
+      "step": 1034
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999494663839915e-05,
+      "loss": 1.7969,
+      "step": 1035
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999484190919936e-05,
+      "loss": 1.5332,
+      "step": 1036
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999473610589107e-05,
+      "loss": 1.4736,
+      "step": 1037
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999946292284754e-05,
+      "loss": 1.2393,
+      "step": 1038
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999945212769535e-05,
+      "loss": 1.5977,
+      "step": 1039
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999441225132652e-05,
+      "loss": 1.2344,
+      "step": 1040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999430215159568e-05,
+      "loss": 1.5645,
+      "step": 1041
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999941909777621e-05,
+      "loss": 1.5127,
+      "step": 1042
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99994078729827e-05,
+      "loss": 1.8867,
+      "step": 1043
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999939654077916e-05,
+      "loss": 1.6084,
+      "step": 1044
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999385101165708e-05,
+      "loss": 1.7754,
+      "step": 1045
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999373554142466e-05,
+      "loss": 1.3047,
+      "step": 1046
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999361899709567e-05,
+      "loss": 1.1797,
+      "step": 1047
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999350137867128e-05,
+      "loss": 1.1719,
+      "step": 1048
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999338268615277e-05,
+      "loss": 1.5547,
+      "step": 1049
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999326291954145e-05,
+      "loss": 1.2412,
+      "step": 1050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999314207883856e-05,
+      "loss": 1.3594,
+      "step": 1051
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999302016404544e-05,
+      "loss": 1.584,
+      "step": 1052
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999928971751633e-05,
+      "loss": 1.417,
+      "step": 1053
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999927731121936e-05,
+      "loss": 1.3164,
+      "step": 1054
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999926479751376e-05,
+      "loss": 1.7637,
+      "step": 1055
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999252176399665e-05,
+      "loss": 1.377,
+      "step": 1056
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999239447877212e-05,
+      "loss": 1.4727,
+      "step": 1057
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999226611946536e-05,
+      "loss": 1.627,
+      "step": 1058
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999213668607777e-05,
+      "loss": 1.9316,
+      "step": 1059
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999200617861068e-05,
+      "loss": 1.2246,
+      "step": 1060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999187459706557e-05,
+      "loss": 1.29,
+      "step": 1061
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999174194144385e-05,
+      "loss": 1.4531,
+      "step": 1062
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999160821174685e-05,
+      "loss": 1.1533,
+      "step": 1063
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999147340797613e-05,
+      "loss": 1.4297,
+      "step": 1064
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999133753013303e-05,
+      "loss": 1.7168,
+      "step": 1065
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999912005782191e-05,
+      "loss": 1.0742,
+      "step": 1066
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999106255223576e-05,
+      "loss": 1.666,
+      "step": 1067
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999909234521845e-05,
+      "loss": 1.2637,
+      "step": 1068
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999907832780668e-05,
+      "loss": 1.4609,
+      "step": 1069
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999906420298842e-05,
+      "loss": 0.8472,
+      "step": 1070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999904997076382e-05,
+      "loss": 1.4883,
+      "step": 1071
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999035631133036e-05,
+      "loss": 1.4023,
+      "step": 1072
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999021184096215e-05,
+      "loss": 2.043,
+      "step": 1073
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999006629653518e-05,
+      "loss": 1.2334,
+      "step": 1074
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99989919678051e-05,
+      "loss": 1.3867,
+      "step": 1075
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998977198551116e-05,
+      "loss": 1.335,
+      "step": 1076
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998962321891726e-05,
+      "loss": 1.6875,
+      "step": 1077
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999894733782709e-05,
+      "loss": 1.8477,
+      "step": 1078
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998932246357375e-05,
+      "loss": 1.4668,
+      "step": 1079
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998917047482735e-05,
+      "loss": 1.6338,
+      "step": 1080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999890174120333e-05,
+      "loss": 1.4463,
+      "step": 1081
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998886327519337e-05,
+      "loss": 1.7305,
+      "step": 1082
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999887080643091e-05,
+      "loss": 1.251,
+      "step": 1083
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998855177938227e-05,
+      "loss": 1.543,
+      "step": 1084
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998839442041446e-05,
+      "loss": 1.4355,
+      "step": 1085
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998823598740738e-05,
+      "loss": 1.6211,
+      "step": 1086
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998807648036273e-05,
+      "loss": 1.3301,
+      "step": 1087
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998791589928227e-05,
+      "loss": 1.5801,
+      "step": 1088
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999877542441677e-05,
+      "loss": 1.9551,
+      "step": 1089
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998759151502075e-05,
+      "loss": 1.5117,
+      "step": 1090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998742771184315e-05,
+      "loss": 1.2866,
+      "step": 1091
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999872628346367e-05,
+      "loss": 1.4834,
+      "step": 1092
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998709688340308e-05,
+      "loss": 1.208,
+      "step": 1093
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998692985814424e-05,
+      "loss": 1.2988,
+      "step": 1094
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999867617588618e-05,
+      "loss": 1.5928,
+      "step": 1095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998659258555767e-05,
+      "loss": 1.3057,
+      "step": 1096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998642233823364e-05,
+      "loss": 1.5684,
+      "step": 1097
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998625101689155e-05,
+      "loss": 1.4551,
+      "step": 1098
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999860786215332e-05,
+      "loss": 1.5781,
+      "step": 1099
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998590515216048e-05,
+      "loss": 1.25,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998573060877527e-05,
+      "loss": 1.5654,
+      "step": 1101
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998555499137937e-05,
+      "loss": 1.3818,
+      "step": 1102
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998537829997474e-05,
+      "loss": 1.5645,
+      "step": 1103
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998520053456324e-05,
+      "loss": 1.4238,
+      "step": 1104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998502169514678e-05,
+      "loss": 1.5117,
+      "step": 1105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998484178172735e-05,
+      "loss": 1.4121,
+      "step": 1106
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999846607943068e-05,
+      "loss": 1.0674,
+      "step": 1107
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998447873288705e-05,
+      "loss": 1.207,
+      "step": 1108
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998429559747016e-05,
+      "loss": 1.167,
+      "step": 1109
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99984111388058e-05,
+      "loss": 1.4873,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998392610465264e-05,
+      "loss": 1.6367,
+      "step": 1111
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998373974725598e-05,
+      "loss": 1.3359,
+      "step": 1112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999835523158701e-05,
+      "loss": 1.4971,
+      "step": 1113
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998336381049694e-05,
+      "loss": 1.4521,
+      "step": 1114
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998317423113856e-05,
+      "loss": 1.7676,
+      "step": 1115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998298357779703e-05,
+      "loss": 1.7109,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998279185047433e-05,
+      "loss": 1.5,
+      "step": 1117
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998259904917257e-05,
+      "loss": 1.2773,
+      "step": 1118
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999824051738938e-05,
+      "loss": 1.7012,
+      "step": 1119
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998221022464014e-05,
+      "loss": 1.5098,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999820142014136e-05,
+      "loss": 1.6113,
+      "step": 1121
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999818171042164e-05,
+      "loss": 1.373,
+      "step": 1122
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998161893305054e-05,
+      "loss": 1.3926,
+      "step": 1123
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998141968791822e-05,
+      "loss": 1.3896,
+      "step": 1124
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999812193688216e-05,
+      "loss": 1.5332,
+      "step": 1125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998101797576273e-05,
+      "loss": 1.2637,
+      "step": 1126
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998081550874388e-05,
+      "loss": 1.4922,
+      "step": 1127
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998061196776715e-05,
+      "loss": 1.7363,
+      "step": 1128
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999804073528348e-05,
+      "loss": 1.5469,
+      "step": 1129
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998020166394898e-05,
+      "loss": 1.4277,
+      "step": 1130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997999490111188e-05,
+      "loss": 1.6328,
+      "step": 1131
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997978706432577e-05,
+      "loss": 1.4023,
+      "step": 1132
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997957815359288e-05,
+      "loss": 1.4902,
+      "step": 1133
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999793681689154e-05,
+      "loss": 1.6914,
+      "step": 1134
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997915711029564e-05,
+      "loss": 1.5176,
+      "step": 1135
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997894497773586e-05,
+      "loss": 1.3174,
+      "step": 1136
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999787317712383e-05,
+      "loss": 1.1982,
+      "step": 1137
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999785174908053e-05,
+      "loss": 1.4131,
+      "step": 1138
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999783021364391e-05,
+      "loss": 1.3096,
+      "step": 1139
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997808570814204e-05,
+      "loss": 1.2695,
+      "step": 1140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997786820591654e-05,
+      "loss": 1.5078,
+      "step": 1141
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999776496297648e-05,
+      "loss": 1.5098,
+      "step": 1142
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999774299796892e-05,
+      "loss": 0.998,
+      "step": 1143
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997720925569217e-05,
+      "loss": 1.4883,
+      "step": 1144
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99976987457776e-05,
+      "loss": 1.5732,
+      "step": 1145
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997676458594308e-05,
+      "loss": 1.1602,
+      "step": 1146
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997654064019588e-05,
+      "loss": 1.7324,
+      "step": 1147
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997631562053674e-05,
+      "loss": 1.4658,
+      "step": 1148
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997608952696806e-05,
+      "loss": 1.3672,
+      "step": 1149
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999758623594923e-05,
+      "loss": 1.2012,
+      "step": 1150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997563411811192e-05,
+      "loss": 1.5098,
+      "step": 1151
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997540480282937e-05,
+      "loss": 1.3584,
+      "step": 1152
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997517441364707e-05,
+      "loss": 1.5684,
+      "step": 1153
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999749429505675e-05,
+      "loss": 1.3916,
+      "step": 1154
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997471041359318e-05,
+      "loss": 1.6172,
+      "step": 1155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997447680272663e-05,
+      "loss": 1.3027,
+      "step": 1156
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997424211797026e-05,
+      "loss": 0.7646,
+      "step": 1157
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997400635932668e-05,
+      "loss": 1.6055,
+      "step": 1158
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999737695267984e-05,
+      "loss": 1.1055,
+      "step": 1159
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997353162038796e-05,
+      "loss": 2.0176,
+      "step": 1160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999732926400979e-05,
+      "loss": 1.4443,
+      "step": 1161
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999730525859308e-05,
+      "loss": 1.041,
+      "step": 1162
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997281145788926e-05,
+      "loss": 1.8828,
+      "step": 1163
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997256925597583e-05,
+      "loss": 1.5059,
+      "step": 1164
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997232598019315e-05,
+      "loss": 1.6191,
+      "step": 1165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997208163054382e-05,
+      "loss": 1.7344,
+      "step": 1166
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997183620703042e-05,
+      "loss": 1.0371,
+      "step": 1167
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997158970965564e-05,
+      "loss": 1.5,
+      "step": 1168
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997134213842214e-05,
+      "loss": 1.5635,
+      "step": 1169
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997109349333254e-05,
+      "loss": 1.373,
+      "step": 1170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997084377438954e-05,
+      "loss": 1.5498,
+      "step": 1171
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999705929815958e-05,
+      "loss": 1.4375,
+      "step": 1172
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99970341114954e-05,
+      "loss": 1.6641,
+      "step": 1173
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997008817446686e-05,
+      "loss": 1.5508,
+      "step": 1174
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996983416013712e-05,
+      "loss": 1.2354,
+      "step": 1175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999695790719675e-05,
+      "loss": 1.2842,
+      "step": 1176
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999693229099607e-05,
+      "loss": 1.6074,
+      "step": 1177
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996906567411957e-05,
+      "loss": 1.5801,
+      "step": 1178
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996880736444675e-05,
+      "loss": 1.4668,
+      "step": 1179
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999685479809451e-05,
+      "loss": 1.8965,
+      "step": 1180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996828752361733e-05,
+      "loss": 1.5664,
+      "step": 1181
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996802599246635e-05,
+      "loss": 1.3789,
+      "step": 1182
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996776338749484e-05,
+      "loss": 1.3838,
+      "step": 1183
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996749970870572e-05,
+      "loss": 1.0156,
+      "step": 1184
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996723495610176e-05,
+      "loss": 1.5059,
+      "step": 1185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996696912968587e-05,
+      "loss": 1.5781,
+      "step": 1186
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996670222946088e-05,
+      "loss": 1.5791,
+      "step": 1187
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996643425542958e-05,
+      "loss": 1.7012,
+      "step": 1188
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996616520759496e-05,
+      "loss": 1.5488,
+      "step": 1189
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996589508595984e-05,
+      "loss": 1.4541,
+      "step": 1190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996562389052716e-05,
+      "loss": 1.3115,
+      "step": 1191
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999653516212998e-05,
+      "loss": 1.3613,
+      "step": 1192
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999650782782807e-05,
+      "loss": 1.3965,
+      "step": 1193
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999648038614728e-05,
+      "loss": 1.6797,
+      "step": 1194
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996452837087906e-05,
+      "loss": 1.6699,
+      "step": 1195
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999642518065024e-05,
+      "loss": 1.7344,
+      "step": 1196
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996397416834585e-05,
+      "loss": 1.208,
+      "step": 1197
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996369545641232e-05,
+      "loss": 2.1621,
+      "step": 1198
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999634156707049e-05,
+      "loss": 1.292,
+      "step": 1199
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996313481122645e-05,
+      "loss": 1.7207,
+      "step": 1200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996285287798013e-05,
+      "loss": 1.2705,
+      "step": 1201
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999625698709689e-05,
+      "loss": 1.4268,
+      "step": 1202
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999622857901958e-05,
+      "loss": 1.4023,
+      "step": 1203
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999620006356639e-05,
+      "loss": 1.8066,
+      "step": 1204
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996171440737627e-05,
+      "loss": 1.4814,
+      "step": 1205
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996142710533596e-05,
+      "loss": 1.7246,
+      "step": 1206
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996113872954607e-05,
+      "loss": 1.3848,
+      "step": 1207
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999608492800097e-05,
+      "loss": 1.4453,
+      "step": 1208
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996055875672995e-05,
+      "loss": 1.5293,
+      "step": 1209
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996026715970995e-05,
+      "loss": 1.5156,
+      "step": 1210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999599744889528e-05,
+      "loss": 1.2148,
+      "step": 1211
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995968074446172e-05,
+      "loss": 1.4004,
+      "step": 1212
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995938592623976e-05,
+      "loss": 1.1953,
+      "step": 1213
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995909003429017e-05,
+      "loss": 1.5703,
+      "step": 1214
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999587930686161e-05,
+      "loss": 1.334,
+      "step": 1215
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995849502922072e-05,
+      "loss": 1.2549,
+      "step": 1216
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999581959161073e-05,
+      "loss": 1.2744,
+      "step": 1217
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995789572927897e-05,
+      "loss": 1.625,
+      "step": 1218
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99957594468739e-05,
+      "loss": 1.3623,
+      "step": 1219
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999572921344906e-05,
+      "loss": 1.6562,
+      "step": 1220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995698872653705e-05,
+      "loss": 1.3691,
+      "step": 1221
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999566842448816e-05,
+      "loss": 1.1963,
+      "step": 1222
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995637868952747e-05,
+      "loss": 1.498,
+      "step": 1223
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995607206047803e-05,
+      "loss": 1.6582,
+      "step": 1224
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999557643577365e-05,
+      "loss": 1.6484,
+      "step": 1225
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995545558130624e-05,
+      "loss": 1.6758,
+      "step": 1226
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995514573119054e-05,
+      "loss": 1.4873,
+      "step": 1227
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999548348073927e-05,
+      "loss": 1.4121,
+      "step": 1228
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999545228099161e-05,
+      "loss": 1.9316,
+      "step": 1229
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995420973876412e-05,
+      "loss": 1.4297,
+      "step": 1230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995389559394004e-05,
+      "loss": 1.4434,
+      "step": 1231
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999535803754473e-05,
+      "loss": 1.1387,
+      "step": 1232
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995326408328925e-05,
+      "loss": 1.4961,
+      "step": 1233
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999529467174693e-05,
+      "loss": 1.0986,
+      "step": 1234
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995262827799083e-05,
+      "loss": 2.1016,
+      "step": 1235
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995230876485736e-05,
+      "loss": 1.2686,
+      "step": 1236
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995198817807223e-05,
+      "loss": 1.7031,
+      "step": 1237
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995166651763885e-05,
+      "loss": 1.5059,
+      "step": 1238
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999513437835608e-05,
+      "loss": 1.3789,
+      "step": 1239
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995101997584147e-05,
+      "loss": 1.5234,
+      "step": 1240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995069509448433e-05,
+      "loss": 1.0186,
+      "step": 1241
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995036913949288e-05,
+      "loss": 1.7129,
+      "step": 1242
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995004211087063e-05,
+      "loss": 1.6211,
+      "step": 1243
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999497140086211e-05,
+      "loss": 1.7598,
+      "step": 1244
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994938483274778e-05,
+      "loss": 1.3789,
+      "step": 1245
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994905458325426e-05,
+      "loss": 1.6465,
+      "step": 1246
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994872326014407e-05,
+      "loss": 1.1514,
+      "step": 1247
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994839086342074e-05,
+      "loss": 1.7031,
+      "step": 1248
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994805739308785e-05,
+      "loss": 1.4678,
+      "step": 1249
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994772284914897e-05,
+      "loss": 1.5215,
+      "step": 1250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994738723160776e-05,
+      "loss": 1.2148,
+      "step": 1251
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994705054046774e-05,
+      "loss": 1.0479,
+      "step": 1252
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999467127757326e-05,
+      "loss": 1.6855,
+      "step": 1253
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999463739374059e-05,
+      "loss": 1.3574,
+      "step": 1254
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994603402549133e-05,
+      "loss": 1.668,
+      "step": 1255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994569303999252e-05,
+      "loss": 1.5742,
+      "step": 1256
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994535098091313e-05,
+      "loss": 1.2344,
+      "step": 1257
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994500784825685e-05,
+      "loss": 1.4775,
+      "step": 1258
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994466364202734e-05,
+      "loss": 1.6875,
+      "step": 1259
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994431836222833e-05,
+      "loss": 1.2295,
+      "step": 1260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994397200886353e-05,
+      "loss": 1.6152,
+      "step": 1261
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994362458193663e-05,
+      "loss": 1.5654,
+      "step": 1262
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994327608145133e-05,
+      "loss": 1.3965,
+      "step": 1263
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999429265074115e-05,
+      "loss": 1.542,
+      "step": 1264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994257585982077e-05,
+      "loss": 1.5859,
+      "step": 1265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994222413868293e-05,
+      "loss": 1.2109,
+      "step": 1266
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999418713440018e-05,
+      "loss": 1.4111,
+      "step": 1267
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994151747578115e-05,
+      "loss": 1.3711,
+      "step": 1268
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999411625340248e-05,
+      "loss": 1.7129,
+      "step": 1269
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999408065187365e-05,
+      "loss": 1.6357,
+      "step": 1270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994044942992015e-05,
+      "loss": 1.3867,
+      "step": 1271
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994009126757952e-05,
+      "loss": 1.5078,
+      "step": 1272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993973203171852e-05,
+      "loss": 1.6543,
+      "step": 1273
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993937172234094e-05,
+      "loss": 1.1475,
+      "step": 1274
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999390103394507e-05,
+      "loss": 1.3535,
+      "step": 1275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999386478830517e-05,
+      "loss": 1.1738,
+      "step": 1276
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999382843531478e-05,
+      "loss": 1.2881,
+      "step": 1277
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993791974974286e-05,
+      "loss": 1.2061,
+      "step": 1278
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999375540728409e-05,
+      "loss": 1.2441,
+      "step": 1279
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993718732244574e-05,
+      "loss": 1.4316,
+      "step": 1280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999368194985614e-05,
+      "loss": 1.1367,
+      "step": 1281
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999364506011918e-05,
+      "loss": 1.0264,
+      "step": 1282
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993608063034094e-05,
+      "loss": 1.5098,
+      "step": 1283
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999357095860127e-05,
+      "loss": 1.96,
+      "step": 1284
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993533746821117e-05,
+      "loss": 1.5742,
+      "step": 1285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993496427694027e-05,
+      "loss": 1.502,
+      "step": 1286
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993459001220404e-05,
+      "loss": 1.334,
+      "step": 1287
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993421467400656e-05,
+      "loss": 1.0918,
+      "step": 1288
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993383826235176e-05,
+      "loss": 1.0771,
+      "step": 1289
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993346077724373e-05,
+      "loss": 1.2393,
+      "step": 1290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999330822186865e-05,
+      "loss": 1.6133,
+      "step": 1291
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993270258668417e-05,
+      "loss": 1.5703,
+      "step": 1292
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993232188124084e-05,
+      "loss": 1.8008,
+      "step": 1293
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999319401023605e-05,
+      "loss": 1.1846,
+      "step": 1294
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999315572500474e-05,
+      "loss": 1.1367,
+      "step": 1295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999311733243055e-05,
+      "loss": 1.709,
+      "step": 1296
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99930788325139e-05,
+      "loss": 1.8105,
+      "step": 1297
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993040225255205e-05,
+      "loss": 1.3906,
+      "step": 1298
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993001510654875e-05,
+      "loss": 1.5293,
+      "step": 1299
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999296268871333e-05,
+      "loss": 1.6133,
+      "step": 1300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992923759430986e-05,
+      "loss": 1.8848,
+      "step": 1301
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999288472280826e-05,
+      "loss": 1.5957,
+      "step": 1302
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999284557884557e-05,
+      "loss": 1.8418,
+      "step": 1303
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999280632754334e-05,
+      "loss": 1.7578,
+      "step": 1304
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999276696890199e-05,
+      "loss": 1.4922,
+      "step": 1305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992727502921943e-05,
+      "loss": 1.5557,
+      "step": 1306
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992687929603623e-05,
+      "loss": 1.2275,
+      "step": 1307
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992648248947452e-05,
+      "loss": 1.4941,
+      "step": 1308
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999260846095386e-05,
+      "loss": 1.5957,
+      "step": 1309
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992568565623277e-05,
+      "loss": 1.2744,
+      "step": 1310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992528562956124e-05,
+      "loss": 1.501,
+      "step": 1311
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992488452952833e-05,
+      "loss": 1.4941,
+      "step": 1312
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992448235613842e-05,
+      "loss": 1.165,
+      "step": 1313
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992407910939575e-05,
+      "loss": 1.5156,
+      "step": 1314
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992367478930465e-05,
+      "loss": 1.6582,
+      "step": 1315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992326939586953e-05,
+      "loss": 1.4922,
+      "step": 1316
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992286292909465e-05,
+      "loss": 1.3809,
+      "step": 1317
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992245538898446e-05,
+      "loss": 1.4893,
+      "step": 1318
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992204677554332e-05,
+      "loss": 1.0742,
+      "step": 1319
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992163708877557e-05,
+      "loss": 1.1338,
+      "step": 1320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999212263286857e-05,
+      "loss": 1.4766,
+      "step": 1321
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992081449527804e-05,
+      "loss": 1.1475,
+      "step": 1322
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992040158855704e-05,
+      "loss": 1.3633,
+      "step": 1323
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991998760852714e-05,
+      "loss": 1.5244,
+      "step": 1324
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999195725551928e-05,
+      "loss": 1.4629,
+      "step": 1325
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991915642855844e-05,
+      "loss": 1.249,
+      "step": 1326
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991873922862854e-05,
+      "loss": 1.6172,
+      "step": 1327
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991832095540764e-05,
+      "loss": 1.4355,
+      "step": 1328
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991790160890015e-05,
+      "loss": 1.832,
+      "step": 1329
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991748118911064e-05,
+      "loss": 1.6094,
+      "step": 1330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999170596960436e-05,
+      "loss": 1.6875,
+      "step": 1331
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999166371297035e-05,
+      "loss": 1.875,
+      "step": 1332
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991621349009498e-05,
+      "loss": 1.3262,
+      "step": 1333
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991578877722255e-05,
+      "loss": 1.293,
+      "step": 1334
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991536299109077e-05,
+      "loss": 1.8809,
+      "step": 1335
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991493613170417e-05,
+      "loss": 1.668,
+      "step": 1336
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991450819906744e-05,
+      "loss": 1.4141,
+      "step": 1337
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991407919318507e-05,
+      "loss": 1.3965,
+      "step": 1338
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991364911406172e-05,
+      "loss": 1.2793,
+      "step": 1339
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99913217961702e-05,
+      "loss": 1.457,
+      "step": 1340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991278573611053e-05,
+      "loss": 2.0117,
+      "step": 1341
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99912352437292e-05,
+      "loss": 1.5625,
+      "step": 1342
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99911918065251e-05,
+      "loss": 1.4531,
+      "step": 1343
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991148261999224e-05,
+      "loss": 1.7031,
+      "step": 1344
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991104610152035e-05,
+      "loss": 1.668,
+      "step": 1345
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991060850984007e-05,
+      "loss": 1.3359,
+      "step": 1346
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999101698449561e-05,
+      "loss": 1.6621,
+      "step": 1347
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990973010687315e-05,
+      "loss": 1.623,
+      "step": 1348
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990928929559586e-05,
+      "loss": 1.6172,
+      "step": 1349
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999088474111291e-05,
+      "loss": 1.4727,
+      "step": 1350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999084044534775e-05,
+      "loss": 1.4316,
+      "step": 1351
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999079604226459e-05,
+      "loss": 1.7432,
+      "step": 1352
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990751531863905e-05,
+      "loss": 1.2363,
+      "step": 1353
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990706914146164e-05,
+      "loss": 1.4717,
+      "step": 1354
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990662189111863e-05,
+      "loss": 1.6543,
+      "step": 1355
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999061735676147e-05,
+      "loss": 0.9795,
+      "step": 1356
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990572417095467e-05,
+      "loss": 1.3086,
+      "step": 1357
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990527370114345e-05,
+      "loss": 1.5059,
+      "step": 1358
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999048221581858e-05,
+      "loss": 1.5195,
+      "step": 1359
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999043695420866e-05,
+      "loss": 1.5723,
+      "step": 1360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990391585285072e-05,
+      "loss": 1.7988,
+      "step": 1361
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99903461090483e-05,
+      "loss": 1.9102,
+      "step": 1362
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990300525498835e-05,
+      "loss": 1.6816,
+      "step": 1363
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990254834637167e-05,
+      "loss": 1.1035,
+      "step": 1364
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990209036463784e-05,
+      "loss": 1.2334,
+      "step": 1365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990163130979184e-05,
+      "loss": 0.9941,
+      "step": 1366
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990117118183858e-05,
+      "loss": 1.7881,
+      "step": 1367
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990070998078293e-05,
+      "loss": 1.75,
+      "step": 1368
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999002477066299e-05,
+      "loss": 1.2227,
+      "step": 1369
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998997843593845e-05,
+      "loss": 1.6309,
+      "step": 1370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998993199390516e-05,
+      "loss": 1.4443,
+      "step": 1371
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989885444563632e-05,
+      "loss": 1.5898,
+      "step": 1372
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989838787914354e-05,
+      "loss": 1.0576,
+      "step": 1373
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989792023957833e-05,
+      "loss": 1.5547,
+      "step": 1374
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998974515269457e-05,
+      "loss": 1.0801,
+      "step": 1375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998969817412507e-05,
+      "loss": 1.2549,
+      "step": 1376
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989651088249834e-05,
+      "loss": 1.2852,
+      "step": 1377
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989603895069372e-05,
+      "loss": 1.4482,
+      "step": 1378
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989556594584192e-05,
+      "loss": 1.251,
+      "step": 1379
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99895091867948e-05,
+      "loss": 1.1934,
+      "step": 1380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99894616717017e-05,
+      "loss": 1.5332,
+      "step": 1381
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989414049305412e-05,
+      "loss": 1.2441,
+      "step": 1382
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989366319606438e-05,
+      "loss": 1.3467,
+      "step": 1383
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989318482605297e-05,
+      "loss": 1.832,
+      "step": 1384
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989270538302502e-05,
+      "loss": 1.4902,
+      "step": 1385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989222486698564e-05,
+      "loss": 1.6953,
+      "step": 1386
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989174327794006e-05,
+      "loss": 1.3047,
+      "step": 1387
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998912606158934e-05,
+      "loss": 1.5859,
+      "step": 1388
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989077688085085e-05,
+      "loss": 1.5625,
+      "step": 1389
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989029207281763e-05,
+      "loss": 1.7969,
+      "step": 1390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998898061917989e-05,
+      "loss": 1.4434,
+      "step": 1391
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988931923779996e-05,
+      "loss": 1.208,
+      "step": 1392
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99888831210826e-05,
+      "loss": 1.1191,
+      "step": 1393
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998883421108822e-05,
+      "loss": 1.5938,
+      "step": 1394
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998878519379739e-05,
+      "loss": 1.501,
+      "step": 1395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988736069210634e-05,
+      "loss": 1.7266,
+      "step": 1396
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998868683732848e-05,
+      "loss": 1.7383,
+      "step": 1397
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988637498151452e-05,
+      "loss": 1.3955,
+      "step": 1398
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988588051680086e-05,
+      "loss": 1.291,
+      "step": 1399
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988538497914915e-05,
+      "loss": 1.4102,
+      "step": 1400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988488836856464e-05,
+      "loss": 1.5615,
+      "step": 1401
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988439068505267e-05,
+      "loss": 1.4727,
+      "step": 1402
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988389192861867e-05,
+      "loss": 1.4004,
+      "step": 1403
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998833920992679e-05,
+      "loss": 1.2734,
+      "step": 1404
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988289119700578e-05,
+      "loss": 1.5664,
+      "step": 1405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998823892218377e-05,
+      "loss": 1.4199,
+      "step": 1406
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99881886173769e-05,
+      "loss": 1.2344,
+      "step": 1407
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988138205280515e-05,
+      "loss": 1.3389,
+      "step": 1408
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988087685895155e-05,
+      "loss": 1.3906,
+      "step": 1409
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988037059221356e-05,
+      "loss": 1.3232,
+      "step": 1410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998798632525967e-05,
+      "loss": 1.5059,
+      "step": 1411
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998793548401064e-05,
+      "loss": 1.2607,
+      "step": 1412
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987884535474803e-05,
+      "loss": 1.6367,
+      "step": 1413
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987833479652724e-05,
+      "loss": 1.9824,
+      "step": 1414
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987782316544933e-05,
+      "loss": 1.5449,
+      "step": 1415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987731046151993e-05,
+      "loss": 1.6719,
+      "step": 1416
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998767966847445e-05,
+      "loss": 1.2002,
+      "step": 1417
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987628183512854e-05,
+      "loss": 2.25,
+      "step": 1418
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998757659126776e-05,
+      "loss": 0.8086,
+      "step": 1419
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998752489173972e-05,
+      "loss": 1.3418,
+      "step": 1420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998747308492929e-05,
+      "loss": 1.3975,
+      "step": 1421
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987421170837033e-05,
+      "loss": 1.3945,
+      "step": 1422
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987369149463496e-05,
+      "loss": 1.5391,
+      "step": 1423
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987317020809245e-05,
+      "loss": 1.291,
+      "step": 1424
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998726478487484e-05,
+      "loss": 1.8418,
+      "step": 1425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998721244166084e-05,
+      "loss": 1.3037,
+      "step": 1426
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998715999116781e-05,
+      "loss": 1.3809,
+      "step": 1427
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987107433396303e-05,
+      "loss": 1.5371,
+      "step": 1428
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987054768346898e-05,
+      "loss": 1.5664,
+      "step": 1429
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998700199602015e-05,
+      "loss": 1.3125,
+      "step": 1430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986949116416633e-05,
+      "loss": 1.3975,
+      "step": 1431
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998689612953691e-05,
+      "loss": 1.5156,
+      "step": 1432
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986843035381553e-05,
+      "loss": 1.4004,
+      "step": 1433
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998678983395113e-05,
+      "loss": 1.7578,
+      "step": 1434
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986736525246216e-05,
+      "loss": 1.4531,
+      "step": 1435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986683109267382e-05,
+      "loss": 1.6426,
+      "step": 1436
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99866295860152e-05,
+      "loss": 1.3887,
+      "step": 1437
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986575955490245e-05,
+      "loss": 1.3779,
+      "step": 1438
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986522217693098e-05,
+      "loss": 0.9863,
+      "step": 1439
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986468372624327e-05,
+      "loss": 1.8945,
+      "step": 1440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986414420284522e-05,
+      "loss": 1.2793,
+      "step": 1441
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986360360674252e-05,
+      "loss": 1.3281,
+      "step": 1442
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986306193794106e-05,
+      "loss": 1.5312,
+      "step": 1443
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986251919644658e-05,
+      "loss": 1.3594,
+      "step": 1444
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986197538226497e-05,
+      "loss": 1.3594,
+      "step": 1445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986143049540205e-05,
+      "loss": 1.4668,
+      "step": 1446
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998608845358637e-05,
+      "loss": 1.5605,
+      "step": 1447
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998603375036557e-05,
+      "loss": 1.5527,
+      "step": 1448
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99859789398784e-05,
+      "loss": 1.8027,
+      "step": 1449
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985924022125453e-05,
+      "loss": 0.9893,
+      "step": 1450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998586899710731e-05,
+      "loss": 1.0137,
+      "step": 1451
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985813864824562e-05,
+      "loss": 1.6406,
+      "step": 1452
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985758625277807e-05,
+      "loss": 1.0566,
+      "step": 1453
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985703278467637e-05,
+      "loss": 1.4072,
+      "step": 1454
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985647824394642e-05,
+      "loss": 1.4141,
+      "step": 1455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985592263059423e-05,
+      "loss": 0.8496,
+      "step": 1456
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985536594462575e-05,
+      "loss": 1.4727,
+      "step": 1457
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985480818604698e-05,
+      "loss": 1.2617,
+      "step": 1458
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985424935486385e-05,
+      "loss": 0.9102,
+      "step": 1459
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985368945108242e-05,
+      "loss": 1.0586,
+      "step": 1460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998531284747087e-05,
+      "loss": 1.4023,
+      "step": 1461
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985256642574863e-05,
+      "loss": 1.4355,
+      "step": 1462
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998520033042084e-05,
+      "loss": 1.2334,
+      "step": 1463
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985143911009394e-05,
+      "loss": 1.7598,
+      "step": 1464
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985087384341136e-05,
+      "loss": 1.3379,
+      "step": 1465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998503075041667e-05,
+      "loss": 1.4609,
+      "step": 1466
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998497400923661e-05,
+      "loss": 1.4688,
+      "step": 1467
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984917160801564e-05,
+      "loss": 1.499,
+      "step": 1468
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984860205112135e-05,
+      "loss": 1.8906,
+      "step": 1469
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998480314216894e-05,
+      "loss": 1.4824,
+      "step": 1470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984745971972596e-05,
+      "loss": 1.1299,
+      "step": 1471
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984688694523712e-05,
+      "loss": 1.2666,
+      "step": 1472
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99846313098229e-05,
+      "loss": 1.7471,
+      "step": 1473
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984573817870787e-05,
+      "loss": 1.7812,
+      "step": 1474
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984516218667985e-05,
+      "loss": 1.4766,
+      "step": 1475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984458512215108e-05,
+      "loss": 2.0195,
+      "step": 1476
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984400698512782e-05,
+      "loss": 0.9795,
+      "step": 1477
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998434277756163e-05,
+      "loss": 1.5488,
+      "step": 1478
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984284749362263e-05,
+      "loss": 1.3545,
+      "step": 1479
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984226613915316e-05,
+      "loss": 1.6895,
+      "step": 1480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984168371221408e-05,
+      "loss": 1.332,
+      "step": 1481
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984110021281165e-05,
+      "loss": 1.2734,
+      "step": 1482
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984051564095215e-05,
+      "loss": 0.9102,
+      "step": 1483
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983992999664187e-05,
+      "loss": 1.4287,
+      "step": 1484
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983934327988708e-05,
+      "loss": 1.0864,
+      "step": 1485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983875549069408e-05,
+      "loss": 1.8018,
+      "step": 1486
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983816662906918e-05,
+      "loss": 1.6406,
+      "step": 1487
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983757669501874e-05,
+      "loss": 1.4131,
+      "step": 1488
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983698568854904e-05,
+      "loss": 1.5918,
+      "step": 1489
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998363936096665e-05,
+      "loss": 1.7871,
+      "step": 1490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983580045837745e-05,
+      "loss": 0.9502,
+      "step": 1491
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998352062346882e-05,
+      "loss": 1.7969,
+      "step": 1492
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983461093860523e-05,
+      "loss": 1.6816,
+      "step": 1493
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983401457013488e-05,
+      "loss": 1.417,
+      "step": 1494
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983341712928354e-05,
+      "loss": 1.3965,
+      "step": 1495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998328186160577e-05,
+      "loss": 1.4375,
+      "step": 1496
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998322190304637e-05,
+      "loss": 1.2617,
+      "step": 1497
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983161837250806e-05,
+      "loss": 1.5059,
+      "step": 1498
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983101664219718e-05,
+      "loss": 1.6992,
+      "step": 1499
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983041383953754e-05,
+      "loss": 1.3652,
+      "step": 1500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998298099645356e-05,
+      "loss": 1.6152,
+      "step": 1501
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998292050171979e-05,
+      "loss": 1.6143,
+      "step": 1502
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982859899753087e-05,
+      "loss": 1.7871,
+      "step": 1503
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982799190554104e-05,
+      "loss": 1.4385,
+      "step": 1504
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982738374123494e-05,
+      "loss": 1.6846,
+      "step": 1505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998267745046191e-05,
+      "loss": 1.4219,
+      "step": 1506
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998261641957001e-05,
+      "loss": 1.6328,
+      "step": 1507
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982555281448443e-05,
+      "loss": 1.4287,
+      "step": 1508
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982494036097872e-05,
+      "loss": 1.1963,
+      "step": 1509
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982432683518947e-05,
+      "loss": 1.7695,
+      "step": 1510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982371223712336e-05,
+      "loss": 1.3008,
+      "step": 1511
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982309656678692e-05,
+      "loss": 1.2266,
+      "step": 1512
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982247982418684e-05,
+      "loss": 1.5898,
+      "step": 1513
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982186200932964e-05,
+      "loss": 1.6113,
+      "step": 1514
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982124312222208e-05,
+      "loss": 1.5938,
+      "step": 1515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998206231628707e-05,
+      "loss": 1.125,
+      "step": 1516
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998200021312822e-05,
+      "loss": 1.541,
+      "step": 1517
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981938002746325e-05,
+      "loss": 1.5693,
+      "step": 1518
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981875685142055e-05,
+      "loss": 1.6211,
+      "step": 1519
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998181326031608e-05,
+      "loss": 1.5684,
+      "step": 1520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981750728269064e-05,
+      "loss": 1.2275,
+      "step": 1521
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981688089001685e-05,
+      "loss": 1.6992,
+      "step": 1522
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998162534251462e-05,
+      "loss": 1.4043,
+      "step": 1523
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998156248880853e-05,
+      "loss": 1.1895,
+      "step": 1524
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981499527884095e-05,
+      "loss": 1.3701,
+      "step": 1525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981436459742e-05,
+      "loss": 1.7441,
+      "step": 1526
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981373284382914e-05,
+      "loss": 1.6191,
+      "step": 1527
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981310001807516e-05,
+      "loss": 1.4414,
+      "step": 1528
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981246612016484e-05,
+      "loss": 1.4453,
+      "step": 1529
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981183115010508e-05,
+      "loss": 1.5312,
+      "step": 1530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981119510790262e-05,
+      "loss": 1.2168,
+      "step": 1531
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981055799356432e-05,
+      "loss": 1.2959,
+      "step": 1532
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980991980709704e-05,
+      "loss": 1.3027,
+      "step": 1533
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980928054850757e-05,
+      "loss": 1.0547,
+      "step": 1534
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980864021780285e-05,
+      "loss": 1.2031,
+      "step": 1535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980799881498973e-05,
+      "loss": 1.7627,
+      "step": 1536
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980735634007508e-05,
+      "loss": 1.6035,
+      "step": 1537
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998067127930658e-05,
+      "loss": 1.582,
+      "step": 1538
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980606817396887e-05,
+      "loss": 1.2676,
+      "step": 1539
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980542248279114e-05,
+      "loss": 1.6328,
+      "step": 1540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998047757195396e-05,
+      "loss": 1.8047,
+      "step": 1541
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980412788422115e-05,
+      "loss": 1.6953,
+      "step": 1542
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980347897684275e-05,
+      "loss": 1.4805,
+      "step": 1543
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980282899741138e-05,
+      "loss": 1.459,
+      "step": 1544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980217794593407e-05,
+      "loss": 1.2441,
+      "step": 1545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980152582241775e-05,
+      "loss": 1.4629,
+      "step": 1546
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980087262686946e-05,
+      "loss": 1.2549,
+      "step": 1547
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980021835929622e-05,
+      "loss": 1.2373,
+      "step": 1548
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979956301970503e-05,
+      "loss": 1.5508,
+      "step": 1549
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997989066081029e-05,
+      "loss": 1.2754,
+      "step": 1550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979824912449702e-05,
+      "loss": 1.9414,
+      "step": 1551
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997975905688943e-05,
+      "loss": 1.2197,
+      "step": 1552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979693094130184e-05,
+      "loss": 1.5469,
+      "step": 1553
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997962702417268e-05,
+      "loss": 1.4746,
+      "step": 1554
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997956084701762e-05,
+      "loss": 1.6641,
+      "step": 1555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997949456266572e-05,
+      "loss": 1.625,
+      "step": 1556
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997942817111769e-05,
+      "loss": 1.5068,
+      "step": 1557
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979361672374242e-05,
+      "loss": 1.3262,
+      "step": 1558
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979295066436094e-05,
+      "loss": 1.5107,
+      "step": 1559
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979228353303958e-05,
+      "loss": 1.2402,
+      "step": 1560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997916153297855e-05,
+      "loss": 0.8281,
+      "step": 1561
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997909460546059e-05,
+      "loss": 1.4326,
+      "step": 1562
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99790275707508e-05,
+      "loss": 1.3809,
+      "step": 1563
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978960428849895e-05,
+      "loss": 1.3477,
+      "step": 1564
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978893179758593e-05,
+      "loss": 1.4385,
+      "step": 1565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978825823477623e-05,
+      "loss": 1.2441,
+      "step": 1566
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997875836000771e-05,
+      "loss": 1.1934,
+      "step": 1567
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997869078934957e-05,
+      "loss": 1.2861,
+      "step": 1568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978623111503937e-05,
+      "loss": 1.7246,
+      "step": 1569
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978555326471535e-05,
+      "loss": 1.3418,
+      "step": 1570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997848743425309e-05,
+      "loss": 1.3633,
+      "step": 1571
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978419434849338e-05,
+      "loss": 1.3613,
+      "step": 1572
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978351328261e-05,
+      "loss": 1.436,
+      "step": 1573
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978283114488814e-05,
+      "loss": 1.3262,
+      "step": 1574
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997821479353351e-05,
+      "loss": 1.6406,
+      "step": 1575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978146365395824e-05,
+      "loss": 1.4062,
+      "step": 1576
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978077830076488e-05,
+      "loss": 1.4355,
+      "step": 1577
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997800918757624e-05,
+      "loss": 1.4883,
+      "step": 1578
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977940437895824e-05,
+      "loss": 1.4004,
+      "step": 1579
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977871581035963e-05,
+      "loss": 1.3613,
+      "step": 1580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997780261699741e-05,
+      "loss": 1.2871,
+      "step": 1581
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977733545780903e-05,
+      "loss": 1.5127,
+      "step": 1582
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977664367387176e-05,
+      "loss": 1.0615,
+      "step": 1583
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977595081816986e-05,
+      "loss": 1.1904,
+      "step": 1584
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977525689071065e-05,
+      "loss": 1.6602,
+      "step": 1585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977456189150164e-05,
+      "loss": 1.4932,
+      "step": 1586
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997738658205503e-05,
+      "loss": 1.6289,
+      "step": 1587
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977316867786406e-05,
+      "loss": 1.2129,
+      "step": 1588
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997724704634505e-05,
+      "loss": 1.543,
+      "step": 1589
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99771771177317e-05,
+      "loss": 1.5703,
+      "step": 1590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977107081947115e-05,
+      "loss": 1.374,
+      "step": 1591
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997703693899205e-05,
+      "loss": 1.3125,
+      "step": 1592
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976966688867246e-05,
+      "loss": 1.5156,
+      "step": 1593
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997689633157347e-05,
+      "loss": 1.498,
+      "step": 1594
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976825867111477e-05,
+      "loss": 1.7402,
+      "step": 1595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976755295482014e-05,
+      "loss": 1.499,
+      "step": 1596
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976684616685848e-05,
+      "loss": 1.6484,
+      "step": 1597
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976613830723734e-05,
+      "loss": 1.2764,
+      "step": 1598
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976542937596435e-05,
+      "loss": 1.1006,
+      "step": 1599
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976471937304713e-05,
+      "loss": 1.6309,
+      "step": 1600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976400829849326e-05,
+      "loss": 1.4355,
+      "step": 1601
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997632961523104e-05,
+      "loss": 1.3984,
+      "step": 1602
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976258293450626e-05,
+      "loss": 1.4199,
+      "step": 1603
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997618686450884e-05,
+      "loss": 1.1875,
+      "step": 1604
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997611532840646e-05,
+      "loss": 1.7422,
+      "step": 1605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976043685144244e-05,
+      "loss": 1.4268,
+      "step": 1606
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975971934722965e-05,
+      "loss": 1.2168,
+      "step": 1607
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975900077143397e-05,
+      "loss": 0.9893,
+      "step": 1608
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997582811240631e-05,
+      "loss": 1.3867,
+      "step": 1609
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975756040512477e-05,
+      "loss": 1.3379,
+      "step": 1610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997568386146267e-05,
+      "loss": 1.7012,
+      "step": 1611
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975611575257668e-05,
+      "loss": 1.2041,
+      "step": 1612
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975539181898247e-05,
+      "loss": 1.085,
+      "step": 1613
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975466681385184e-05,
+      "loss": 1.1309,
+      "step": 1614
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997539407371925e-05,
+      "loss": 1.5,
+      "step": 1615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997532135890124e-05,
+      "loss": 1.2109,
+      "step": 1616
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975248536931925e-05,
+      "loss": 1.8691,
+      "step": 1617
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975175607812086e-05,
+      "loss": 1.5332,
+      "step": 1618
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975102571542515e-05,
+      "loss": 1.4834,
+      "step": 1619
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975029428123988e-05,
+      "loss": 1.9336,
+      "step": 1620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974956177557293e-05,
+      "loss": 1.5059,
+      "step": 1621
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997488281984322e-05,
+      "loss": 1.3887,
+      "step": 1622
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974809354982555e-05,
+      "loss": 1.0146,
+      "step": 1623
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974735782976083e-05,
+      "loss": 1.5781,
+      "step": 1624
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974662103824602e-05,
+      "loss": 1.2383,
+      "step": 1625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974588317528902e-05,
+      "loss": 1.2695,
+      "step": 1626
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974514424089767e-05,
+      "loss": 1.2314,
+      "step": 1627
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974440423508002e-05,
+      "loss": 1.6797,
+      "step": 1628
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974366315784396e-05,
+      "loss": 1.8496,
+      "step": 1629
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997429210091974e-05,
+      "loss": 1.3096,
+      "step": 1630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974217778914845e-05,
+      "loss": 1.6973,
+      "step": 1631
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99741433497705e-05,
+      "loss": 0.8506,
+      "step": 1632
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99740688134875e-05,
+      "loss": 1.8086,
+      "step": 1633
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973994170066654e-05,
+      "loss": 1.5547,
+      "step": 1634
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973919419508765e-05,
+      "loss": 1.3926,
+      "step": 1635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997384456181463e-05,
+      "loss": 1.6094,
+      "step": 1636
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973769596985057e-05,
+      "loss": 1.2246,
+      "step": 1637
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973694525020848e-05,
+      "loss": 1.3359,
+      "step": 1638
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997361934592281e-05,
+      "loss": 1.583,
+      "step": 1639
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973544059691755e-05,
+      "loss": 1.2793,
+      "step": 1640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973468666328484e-05,
+      "loss": 1.3613,
+      "step": 1641
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973393165833813e-05,
+      "loss": 0.9404,
+      "step": 1642
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997331755820855e-05,
+      "loss": 1.4492,
+      "step": 1643
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997324184345351e-05,
+      "loss": 1.2217,
+      "step": 1644
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99731660215695e-05,
+      "loss": 1.3848,
+      "step": 1645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973090092557346e-05,
+      "loss": 1.3535,
+      "step": 1646
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973014056417854e-05,
+      "loss": 1.3184,
+      "step": 1647
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972937913151842e-05,
+      "loss": 1.3135,
+      "step": 1648
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997286166276013e-05,
+      "loss": 1.2939,
+      "step": 1649
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972785305243537e-05,
+      "loss": 1.2207,
+      "step": 1650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997270884060288e-05,
+      "loss": 1.2227,
+      "step": 1651
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972632268838988e-05,
+      "loss": 1.5254,
+      "step": 1652
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972555589952677e-05,
+      "loss": 1.3301,
+      "step": 1653
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997247880394477e-05,
+      "loss": 1.5059,
+      "step": 1654
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997240191081609e-05,
+      "loss": 1.9844,
+      "step": 1655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972324910567473e-05,
+      "loss": 1.4346,
+      "step": 1656
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972247803199738e-05,
+      "loss": 1.3779,
+      "step": 1657
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972170588713715e-05,
+      "loss": 1.251,
+      "step": 1658
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972093267110233e-05,
+      "loss": 1.1797,
+      "step": 1659
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972015838390123e-05,
+      "loss": 1.4434,
+      "step": 1660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971938302554217e-05,
+      "loss": 1.3623,
+      "step": 1661
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971860659603348e-05,
+      "loss": 1.3301,
+      "step": 1662
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997178290953835e-05,
+      "loss": 1.458,
+      "step": 1663
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971705052360057e-05,
+      "loss": 1.623,
+      "step": 1664
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997162708806931e-05,
+      "loss": 1.3369,
+      "step": 1665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997154901666694e-05,
+      "loss": 1.4199,
+      "step": 1666
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971470838153787e-05,
+      "loss": 1.4727,
+      "step": 1667
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971392552530692e-05,
+      "loss": 1.3037,
+      "step": 1668
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99713141597985e-05,
+      "loss": 1.2285,
+      "step": 1669
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971235659958045e-05,
+      "loss": 1.418,
+      "step": 1670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971157053010173e-05,
+      "loss": 1.6777,
+      "step": 1671
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971078338955734e-05,
+      "loss": 1.4316,
+      "step": 1672
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970999517795567e-05,
+      "loss": 1.1992,
+      "step": 1673
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970920589530517e-05,
+      "loss": 1.1875,
+      "step": 1674
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970841554161444e-05,
+      "loss": 1.3857,
+      "step": 1675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997076241168918e-05,
+      "loss": 1.4688,
+      "step": 1676
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997068316211459e-05,
+      "loss": 1.7803,
+      "step": 1677
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970603805438517e-05,
+      "loss": 1.4844,
+      "step": 1678
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970524341661815e-05,
+      "loss": 1.5439,
+      "step": 1679
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970444770785337e-05,
+      "loss": 1.5332,
+      "step": 1680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997036509280994e-05,
+      "loss": 1.4932,
+      "step": 1681
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997028530773648e-05,
+      "loss": 1.6602,
+      "step": 1682
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997020541556581e-05,
+      "loss": 1.3555,
+      "step": 1683
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970125416298793e-05,
+      "loss": 1.6699,
+      "step": 1684
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970045309936284e-05,
+      "loss": 1.499,
+      "step": 1685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969965096479144e-05,
+      "loss": 1.5439,
+      "step": 1686
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996988477592824e-05,
+      "loss": 1.1934,
+      "step": 1687
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996980434828443e-05,
+      "loss": 1.7148,
+      "step": 1688
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969723813548575e-05,
+      "loss": 1.416,
+      "step": 1689
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996964317172155e-05,
+      "loss": 1.832,
+      "step": 1690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969562422804213e-05,
+      "loss": 1.5898,
+      "step": 1691
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969481566797433e-05,
+      "loss": 1.3623,
+      "step": 1692
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969400603702078e-05,
+      "loss": 0.9951,
+      "step": 1693
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996931953351902e-05,
+      "loss": 1.5391,
+      "step": 1694
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969238356249128e-05,
+      "loss": 1.3027,
+      "step": 1695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969157071893278e-05,
+      "loss": 1.2842,
+      "step": 1696
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969075680452334e-05,
+      "loss": 1.9512,
+      "step": 1697
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968994181927178e-05,
+      "loss": 1.4004,
+      "step": 1698
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968912576318687e-05,
+      "loss": 1.6172,
+      "step": 1699
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968830863627732e-05,
+      "loss": 1.1279,
+      "step": 1700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996874904385519e-05,
+      "loss": 1.5449,
+      "step": 1701
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968667117001942e-05,
+      "loss": 1.6738,
+      "step": 1702
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968585083068873e-05,
+      "loss": 1.2256,
+      "step": 1703
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968502942056856e-05,
+      "loss": 1.123,
+      "step": 1704
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996842069396678e-05,
+      "loss": 1.6465,
+      "step": 1705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968338338799525e-05,
+      "loss": 1.3696,
+      "step": 1706
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968255876555976e-05,
+      "loss": 1.4453,
+      "step": 1707
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968173307237016e-05,
+      "loss": 1.9492,
+      "step": 1708
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968090630843538e-05,
+      "loss": 1.2051,
+      "step": 1709
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968007847376428e-05,
+      "loss": 1.209,
+      "step": 1710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996792495683657e-05,
+      "loss": 1.2744,
+      "step": 1711
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967841959224864e-05,
+      "loss": 1.3223,
+      "step": 1712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967758854542193e-05,
+      "loss": 1.2422,
+      "step": 1713
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996767564278945e-05,
+      "loss": 1.332,
+      "step": 1714
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996759232396754e-05,
+      "loss": 1.1475,
+      "step": 1715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967508898077345e-05,
+      "loss": 1.4805,
+      "step": 1716
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967425365119765e-05,
+      "loss": 1.8086,
+      "step": 1717
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99673417250957e-05,
+      "loss": 1.498,
+      "step": 1718
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967257978006044e-05,
+      "loss": 1.5,
+      "step": 1719
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99671741238517e-05,
+      "loss": 1.9424,
+      "step": 1720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967090162633568e-05,
+      "loss": 1.4082,
+      "step": 1721
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996700609435255e-05,
+      "loss": 1.2471,
+      "step": 1722
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996692191900955e-05,
+      "loss": 1.8828,
+      "step": 1723
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966837636605465e-05,
+      "loss": 1.3711,
+      "step": 1724
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966753247141213e-05,
+      "loss": 0.8975,
+      "step": 1725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996666875061769e-05,
+      "loss": 1.5234,
+      "step": 1726
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966584147035808e-05,
+      "loss": 1.1709,
+      "step": 1727
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966499436396477e-05,
+      "loss": 1.2002,
+      "step": 1728
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966414618700604e-05,
+      "loss": 1.7988,
+      "step": 1729
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966329693949098e-05,
+      "loss": 1.417,
+      "step": 1730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996624466214288e-05,
+      "loss": 1.4668,
+      "step": 1731
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966159523282853e-05,
+      "loss": 1.3379,
+      "step": 1732
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996607427736994e-05,
+      "loss": 1.8262,
+      "step": 1733
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996598892440505e-05,
+      "loss": 1.4697,
+      "step": 1734
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965903464389106e-05,
+      "loss": 1.6348,
+      "step": 1735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996581789732302e-05,
+      "loss": 1.2725,
+      "step": 1736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965732223207715e-05,
+      "loss": 1.2607,
+      "step": 1737
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965646442044106e-05,
+      "loss": 1.3662,
+      "step": 1738
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965560553833124e-05,
+      "loss": 1.1631,
+      "step": 1739
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965474558575683e-05,
+      "loss": 1.4688,
+      "step": 1740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965388456272712e-05,
+      "loss": 1.7051,
+      "step": 1741
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965302246925136e-05,
+      "loss": 1.0869,
+      "step": 1742
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965215930533873e-05,
+      "loss": 1.2979,
+      "step": 1743
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996512950709986e-05,
+      "loss": 1.3086,
+      "step": 1744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965042976624022e-05,
+      "loss": 1.125,
+      "step": 1745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964956339107285e-05,
+      "loss": 1.4941,
+      "step": 1746
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964869594550584e-05,
+      "loss": 1.3887,
+      "step": 1747
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964782742954848e-05,
+      "loss": 1.3125,
+      "step": 1748
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996469578432101e-05,
+      "loss": 1.5156,
+      "step": 1749
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996460871865001e-05,
+      "loss": 1.3691,
+      "step": 1750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964521545942776e-05,
+      "loss": 1.1191,
+      "step": 1751
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964434266200247e-05,
+      "loss": 1.3096,
+      "step": 1752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996434687942336e-05,
+      "loss": 1.4229,
+      "step": 1753
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964259385613053e-05,
+      "loss": 1.0234,
+      "step": 1754
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964171784770268e-05,
+      "loss": 1.2441,
+      "step": 1755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964084076895945e-05,
+      "loss": 1.6543,
+      "step": 1756
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963996261991027e-05,
+      "loss": 1.0547,
+      "step": 1757
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963908340056456e-05,
+      "loss": 1.1357,
+      "step": 1758
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963820311093173e-05,
+      "loss": 1.416,
+      "step": 1759
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996373217510213e-05,
+      "loss": 1.207,
+      "step": 1760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996364393208427e-05,
+      "loss": 1.5264,
+      "step": 1761
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963555582040546e-05,
+      "loss": 1.5352,
+      "step": 1762
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963467124971898e-05,
+      "loss": 1.5029,
+      "step": 1763
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963378560879284e-05,
+      "loss": 1.5723,
+      "step": 1764
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996328988976365e-05,
+      "loss": 0.9121,
+      "step": 1765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996320111162595e-05,
+      "loss": 1.3545,
+      "step": 1766
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996311222646714e-05,
+      "loss": 1.5352,
+      "step": 1767
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963023234288176e-05,
+      "loss": 1.791,
+      "step": 1768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996293413509001e-05,
+      "loss": 1.2168,
+      "step": 1769
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962844928873598e-05,
+      "loss": 1.0801,
+      "step": 1770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99627556156399e-05,
+      "loss": 1.4902,
+      "step": 1771
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962666195389875e-05,
+      "loss": 1.5391,
+      "step": 1772
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996257666812449e-05,
+      "loss": 1.1387,
+      "step": 1773
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962487033844697e-05,
+      "loss": 1.3359,
+      "step": 1774
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962397292551464e-05,
+      "loss": 1.1543,
+      "step": 1775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996230744424575e-05,
+      "loss": 1.1719,
+      "step": 1776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962217488928532e-05,
+      "loss": 1.3398,
+      "step": 1777
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996212742660076e-05,
+      "loss": 1.4629,
+      "step": 1778
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962037257263412e-05,
+      "loss": 1.291,
+      "step": 1779
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961946980917457e-05,
+      "loss": 1.2236,
+      "step": 1780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996185659756386e-05,
+      "loss": 1.832,
+      "step": 1781
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961766107203596e-05,
+      "loss": 1.4434,
+      "step": 1782
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996167550983763e-05,
+      "loss": 1.2383,
+      "step": 1783
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996158480546694e-05,
+      "loss": 1.6953,
+      "step": 1784
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961493994092506e-05,
+      "loss": 1.3887,
+      "step": 1785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961403075715292e-05,
+      "loss": 1.1758,
+      "step": 1786
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961312050336286e-05,
+      "loss": 1.5225,
+      "step": 1787
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961220917956455e-05,
+      "loss": 1.1973,
+      "step": 1788
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961129678576783e-05,
+      "loss": 1.3223,
+      "step": 1789
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961038332198256e-05,
+      "loss": 1.6113,
+      "step": 1790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996094687882184e-05,
+      "loss": 1.4062,
+      "step": 1791
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960855318448536e-05,
+      "loss": 1.3672,
+      "step": 1792
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996076365107931e-05,
+      "loss": 1.5977,
+      "step": 1793
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996067187671516e-05,
+      "loss": 1.2002,
+      "step": 1794
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960579995357065e-05,
+      "loss": 1.2969,
+      "step": 1795
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960488007006016e-05,
+      "loss": 1.4316,
+      "step": 1796
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960395911662997e-05,
+      "loss": 1.1299,
+      "step": 1797
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960303709328996e-05,
+      "loss": 1.3486,
+      "step": 1798
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996021140000501e-05,
+      "loss": 1.5684,
+      "step": 1799
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960118983692025e-05,
+      "loss": 1.1641,
+      "step": 1800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960026460391036e-05,
+      "loss": 1.1851,
+      "step": 1801
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959933830103034e-05,
+      "loss": 1.3535,
+      "step": 1802
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995984109282902e-05,
+      "loss": 1.5898,
+      "step": 1803
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959748248569986e-05,
+      "loss": 1.709,
+      "step": 1804
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995965529732693e-05,
+      "loss": 1.1201,
+      "step": 1805
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995956223910085e-05,
+      "loss": 1.7471,
+      "step": 1806
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959469073892745e-05,
+      "loss": 1.3594,
+      "step": 1807
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995937580170362e-05,
+      "loss": 1.6719,
+      "step": 1808
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995928242253447e-05,
+      "loss": 1.1406,
+      "step": 1809
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99591889363863e-05,
+      "loss": 1.5176,
+      "step": 1810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995909534326012e-05,
+      "loss": 1.2168,
+      "step": 1811
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959001643156927e-05,
+      "loss": 1.1084,
+      "step": 1812
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958907836077734e-05,
+      "loss": 1.6152,
+      "step": 1813
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958813922023547e-05,
+      "loss": 1.1621,
+      "step": 1814
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958719900995374e-05,
+      "loss": 1.5469,
+      "step": 1815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958625772994225e-05,
+      "loss": 1.3701,
+      "step": 1816
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958531538021112e-05,
+      "loss": 1.5146,
+      "step": 1817
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995843719607704e-05,
+      "loss": 1.3496,
+      "step": 1818
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958342747163036e-05,
+      "loss": 1.2334,
+      "step": 1819
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958248191280107e-05,
+      "loss": 1.5156,
+      "step": 1820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958153528429266e-05,
+      "loss": 1.4688,
+      "step": 1821
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995805875861153e-05,
+      "loss": 1.5029,
+      "step": 1822
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957963881827925e-05,
+      "loss": 1.4688,
+      "step": 1823
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957868898079466e-05,
+      "loss": 1.2559,
+      "step": 1824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957773807367167e-05,
+      "loss": 1.1406,
+      "step": 1825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957678609692056e-05,
+      "loss": 1.582,
+      "step": 1826
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957583305055155e-05,
+      "loss": 1.2451,
+      "step": 1827
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995748789345749e-05,
+      "loss": 1.2773,
+      "step": 1828
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957392374900077e-05,
+      "loss": 1.5361,
+      "step": 1829
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957296749383952e-05,
+      "loss": 1.1621,
+      "step": 1830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957201016910137e-05,
+      "loss": 0.9365,
+      "step": 1831
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995710517747966e-05,
+      "loss": 1.2275,
+      "step": 1832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957009231093553e-05,
+      "loss": 1.1895,
+      "step": 1833
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956913177752846e-05,
+      "loss": 1.5703,
+      "step": 1834
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995681701745857e-05,
+      "loss": 1.5859,
+      "step": 1835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956720750211755e-05,
+      "loss": 1.4707,
+      "step": 1836
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956624376013443e-05,
+      "loss": 1.5293,
+      "step": 1837
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956527894864662e-05,
+      "loss": 1.3594,
+      "step": 1838
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956431306766452e-05,
+      "loss": 1.6953,
+      "step": 1839
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956334611719846e-05,
+      "loss": 1.5381,
+      "step": 1840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956237809725888e-05,
+      "loss": 1.8652,
+      "step": 1841
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956140900785617e-05,
+      "loss": 1.2734,
+      "step": 1842
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995604388490007e-05,
+      "loss": 1.1152,
+      "step": 1843
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955946762070293e-05,
+      "loss": 1.1201,
+      "step": 1844
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995584953229733e-05,
+      "loss": 1.5918,
+      "step": 1845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995575219558222e-05,
+      "loss": 1.4648,
+      "step": 1846
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955654751926015e-05,
+      "loss": 1.3135,
+      "step": 1847
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995555720132976e-05,
+      "loss": 1.5508,
+      "step": 1848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99554595437945e-05,
+      "loss": 1.2344,
+      "step": 1849
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995536177932128e-05,
+      "loss": 1.5898,
+      "step": 1850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995526390791116e-05,
+      "loss": 1.4834,
+      "step": 1851
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995516592956519e-05,
+      "loss": 1.7773,
+      "step": 1852
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955067844284418e-05,
+      "loss": 1.1777,
+      "step": 1853
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954969652069896e-05,
+      "loss": 1.4258,
+      "step": 1854
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954871352922686e-05,
+      "loss": 1.2646,
+      "step": 1855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954772946843838e-05,
+      "loss": 1.8203,
+      "step": 1856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995467443383441e-05,
+      "loss": 1.2725,
+      "step": 1857
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954575813895462e-05,
+      "loss": 1.4824,
+      "step": 1858
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954477087028052e-05,
+      "loss": 1.2148,
+      "step": 1859
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995437825323324e-05,
+      "loss": 1.4072,
+      "step": 1860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954279312512086e-05,
+      "loss": 1.625,
+      "step": 1861
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995418026486566e-05,
+      "loss": 1.2012,
+      "step": 1862
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995408111029502e-05,
+      "loss": 1.3135,
+      "step": 1863
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995398184880123e-05,
+      "loss": 1.2158,
+      "step": 1864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953882480385364e-05,
+      "loss": 1.1914,
+      "step": 1865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953783005048478e-05,
+      "loss": 1.2407,
+      "step": 1866
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995368342279165e-05,
+      "loss": 1.4199,
+      "step": 1867
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953583733615942e-05,
+      "loss": 1.8066,
+      "step": 1868
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995348393752243e-05,
+      "loss": 1.4258,
+      "step": 1869
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953384034512185e-05,
+      "loss": 1.4141,
+      "step": 1870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995328402458628e-05,
+      "loss": 1.627,
+      "step": 1871
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953183907745793e-05,
+      "loss": 1.8652,
+      "step": 1872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953083683991794e-05,
+      "loss": 1.332,
+      "step": 1873
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952983353325358e-05,
+      "loss": 1.6396,
+      "step": 1874
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952882915747568e-05,
+      "loss": 1.6465,
+      "step": 1875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99527823712595e-05,
+      "loss": 1.5117,
+      "step": 1876
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952681719862237e-05,
+      "loss": 1.2637,
+      "step": 1877
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952580961556855e-05,
+      "loss": 1.5654,
+      "step": 1878
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952480096344442e-05,
+      "loss": 1.2842,
+      "step": 1879
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952379124226077e-05,
+      "loss": 1.2383,
+      "step": 1880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995227804520285e-05,
+      "loss": 1.7793,
+      "step": 1881
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995217685927584e-05,
+      "loss": 1.5293,
+      "step": 1882
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952075566446134e-05,
+      "loss": 1.1455,
+      "step": 1883
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995197416671483e-05,
+      "loss": 1.5176,
+      "step": 1884
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951872660083006e-05,
+      "loss": 1.332,
+      "step": 1885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951771046551757e-05,
+      "loss": 1.627,
+      "step": 1886
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995166932612217e-05,
+      "loss": 1.4609,
+      "step": 1887
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951567498795348e-05,
+      "loss": 1.166,
+      "step": 1888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951465564572377e-05,
+      "loss": 1.5039,
+      "step": 1889
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951363523454353e-05,
+      "loss": 1.0881,
+      "step": 1890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951261375442373e-05,
+      "loss": 1.4199,
+      "step": 1891
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951159120537533e-05,
+      "loss": 1.1641,
+      "step": 1892
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951056758740932e-05,
+      "loss": 1.2559,
+      "step": 1893
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995095429005367e-05,
+      "loss": 1.3486,
+      "step": 1894
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950851714476846e-05,
+      "loss": 1.2793,
+      "step": 1895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950749032011566e-05,
+      "loss": 1.6895,
+      "step": 1896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950646242658928e-05,
+      "loss": 1.3848,
+      "step": 1897
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950543346420042e-05,
+      "loss": 1.1035,
+      "step": 1898
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950440343296007e-05,
+      "loss": 1.4492,
+      "step": 1899
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995033723328793e-05,
+      "loss": 1.5547,
+      "step": 1900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950234016396922e-05,
+      "loss": 1.0693,
+      "step": 1901
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950130692624092e-05,
+      "loss": 1.4404,
+      "step": 1902
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950027261970547e-05,
+      "loss": 1.4375,
+      "step": 1903
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99499237244374e-05,
+      "loss": 1.2744,
+      "step": 1904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994982008002576e-05,
+      "loss": 1.4038,
+      "step": 1905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949716328736746e-05,
+      "loss": 1.4863,
+      "step": 1906
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949612470571467e-05,
+      "loss": 1.293,
+      "step": 1907
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994950850553104e-05,
+      "loss": 1.083,
+      "step": 1908
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949404433616583e-05,
+      "loss": 1.4258,
+      "step": 1909
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949300254829217e-05,
+      "loss": 1.4961,
+      "step": 1910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994919596917005e-05,
+      "loss": 1.3555,
+      "step": 1911
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949091576640217e-05,
+      "loss": 1.1309,
+      "step": 1912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948987077240832e-05,
+      "loss": 1.166,
+      "step": 1913
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994888247097301e-05,
+      "loss": 1.4004,
+      "step": 1914
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994877775783789e-05,
+      "loss": 1.6211,
+      "step": 1915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948672937836588e-05,
+      "loss": 1.4551,
+      "step": 1916
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948568010970225e-05,
+      "loss": 1.418,
+      "step": 1917
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994846297723994e-05,
+      "loss": 1.3252,
+      "step": 1918
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994835783664685e-05,
+      "loss": 1.5781,
+      "step": 1919
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948252589192095e-05,
+      "loss": 1.4336,
+      "step": 1920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948147234876797e-05,
+      "loss": 1.1758,
+      "step": 1921
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994804177370209e-05,
+      "loss": 1.5029,
+      "step": 1922
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994793620566911e-05,
+      "loss": 1.2544,
+      "step": 1923
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947830530778988e-05,
+      "loss": 1.1992,
+      "step": 1924
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994772474903286e-05,
+      "loss": 1.2783,
+      "step": 1925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947618860431858e-05,
+      "loss": 1.2363,
+      "step": 1926
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947512864977128e-05,
+      "loss": 1.4902,
+      "step": 1927
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947406762669802e-05,
+      "loss": 1.3525,
+      "step": 1928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994730055351102e-05,
+      "loss": 1.3613,
+      "step": 1929
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947194237501928e-05,
+      "loss": 1.3574,
+      "step": 1930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947087814643663e-05,
+      "loss": 1.2314,
+      "step": 1931
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946981284937372e-05,
+      "loss": 1.0059,
+      "step": 1932
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946874648384195e-05,
+      "loss": 1.0977,
+      "step": 1933
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946767904985278e-05,
+      "loss": 1.5039,
+      "step": 1934
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994666105474177e-05,
+      "loss": 1.3975,
+      "step": 1935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994655409765482e-05,
+      "loss": 1.7627,
+      "step": 1936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994644703372557e-05,
+      "loss": 1.457,
+      "step": 1937
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946339862955177e-05,
+      "loss": 1.3838,
+      "step": 1938
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994623258534479e-05,
+      "loss": 1.4561,
+      "step": 1939
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946125200895565e-05,
+      "loss": 0.999,
+      "step": 1940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946017709608646e-05,
+      "loss": 0.9873,
+      "step": 1941
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945910111485198e-05,
+      "loss": 1.2402,
+      "step": 1942
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994580240652637e-05,
+      "loss": 1.6133,
+      "step": 1943
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994569459473332e-05,
+      "loss": 1.5059,
+      "step": 1944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945586676107207e-05,
+      "loss": 1.3652,
+      "step": 1945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945478650649192e-05,
+      "loss": 1.7246,
+      "step": 1946
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945370518360435e-05,
+      "loss": 1.5879,
+      "step": 1947
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945262279242093e-05,
+      "loss": 1.2275,
+      "step": 1948
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945153933295333e-05,
+      "loss": 1.4424,
+      "step": 1949
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994504548052132e-05,
+      "loss": 1.248,
+      "step": 1950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944936920921215e-05,
+      "loss": 1.5381,
+      "step": 1951
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944828254496185e-05,
+      "loss": 1.21,
+      "step": 1952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944719481247398e-05,
+      "loss": 1.3496,
+      "step": 1953
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944610601176023e-05,
+      "loss": 1.5322,
+      "step": 1954
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994450161428323e-05,
+      "loss": 1.2061,
+      "step": 1955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994439252057019e-05,
+      "loss": 1.5479,
+      "step": 1956
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994428332003807e-05,
+      "loss": 1.418,
+      "step": 1957
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994417401268805e-05,
+      "loss": 1.7246,
+      "step": 1958
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944064598521296e-05,
+      "loss": 1.2583,
+      "step": 1959
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943955077538993e-05,
+      "loss": 1.3516,
+      "step": 1960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943845449742308e-05,
+      "loss": 1.6426,
+      "step": 1961
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943735715132427e-05,
+      "loss": 1.2725,
+      "step": 1962
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943625873710523e-05,
+      "loss": 1.6777,
+      "step": 1963
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943515925477777e-05,
+      "loss": 0.9902,
+      "step": 1964
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943405870435373e-05,
+      "loss": 1.417,
+      "step": 1965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943295708584486e-05,
+      "loss": 1.2314,
+      "step": 1966
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994318543992631e-05,
+      "loss": 1.2012,
+      "step": 1967
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994307506446202e-05,
+      "loss": 1.3672,
+      "step": 1968
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942964582192806e-05,
+      "loss": 1.9609,
+      "step": 1969
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942853993119852e-05,
+      "loss": 1.2129,
+      "step": 1970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994274329724435e-05,
+      "loss": 1.332,
+      "step": 1971
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994263249456749e-05,
+      "loss": 1.4131,
+      "step": 1972
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942521585090453e-05,
+      "loss": 1.5273,
+      "step": 1973
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994241056881444e-05,
+      "loss": 1.3281,
+      "step": 1974
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942299445740637e-05,
+      "loss": 1.7988,
+      "step": 1975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942188215870244e-05,
+      "loss": 1.4141,
+      "step": 1976
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994207687920445e-05,
+      "loss": 1.0713,
+      "step": 1977
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941965435744455e-05,
+      "loss": 1.2676,
+      "step": 1978
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941853885491456e-05,
+      "loss": 1.3418,
+      "step": 1979
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941742228446647e-05,
+      "loss": 1.1123,
+      "step": 1980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994163046461123e-05,
+      "loss": 1.4277,
+      "step": 1981
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994151859398641e-05,
+      "loss": 1.3398,
+      "step": 1982
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941406616573377e-05,
+      "loss": 1.5625,
+      "step": 1983
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941294532373345e-05,
+      "loss": 1.4121,
+      "step": 1984
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941182341387514e-05,
+      "loss": 1.0801,
+      "step": 1985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994107004361709e-05,
+      "loss": 1.707,
+      "step": 1986
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994095763906328e-05,
+      "loss": 1.5996,
+      "step": 1987
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940845127727283e-05,
+      "loss": 1.6582,
+      "step": 1988
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994073250961032e-05,
+      "loss": 1.5254,
+      "step": 1989
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994061978471359e-05,
+      "loss": 1.4316,
+      "step": 1990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940506953038317e-05,
+      "loss": 1.1221,
+      "step": 1991
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99403940145857e-05,
+      "loss": 1.7363,
+      "step": 1992
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940280969356958e-05,
+      "loss": 1.7285,
+      "step": 1993
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99401678173533e-05,
+      "loss": 1.4785,
+      "step": 1994
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994005455857595e-05,
+      "loss": 1.2988,
+      "step": 1995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993994119302612e-05,
+      "loss": 1.2881,
+      "step": 1996
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939827720705033e-05,
+      "loss": 1.4258,
+      "step": 1997
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939714141613897e-05,
+      "loss": 1.7637,
+      "step": 1998
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993960045575394e-05,
+      "loss": 1.4395,
+      "step": 1999
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939486663126376e-05,
+      "loss": 1.8574,
+      "step": 2000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993937276373244e-05,
+      "loss": 1.6699,
+      "step": 2001
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939258757573345e-05,
+      "loss": 1.5625,
+      "step": 2002
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939144644650315e-05,
+      "loss": 1.1172,
+      "step": 2003
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939030424964584e-05,
+      "loss": 1.3926,
+      "step": 2004
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938916098517372e-05,
+      "loss": 1.6094,
+      "step": 2005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993880166530991e-05,
+      "loss": 1.1348,
+      "step": 2006
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993868712534343e-05,
+      "loss": 1.5,
+      "step": 2007
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938572478619154e-05,
+      "loss": 1.1099,
+      "step": 2008
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993845772513832e-05,
+      "loss": 1.3076,
+      "step": 2009
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993834286490216e-05,
+      "loss": 1.3994,
+      "step": 2010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938227897911904e-05,
+      "loss": 1.4512,
+      "step": 2011
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938112824168795e-05,
+      "loss": 1.7676,
+      "step": 2012
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993799764367406e-05,
+      "loss": 1.3955,
+      "step": 2013
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937882356428942e-05,
+      "loss": 1.3643,
+      "step": 2014
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937766962434675e-05,
+      "loss": 1.3853,
+      "step": 2015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99376514616925e-05,
+      "loss": 1.1592,
+      "step": 2016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993753585420366e-05,
+      "loss": 1.4375,
+      "step": 2017
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937420139969397e-05,
+      "loss": 1.4209,
+      "step": 2018
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937304318990952e-05,
+      "loss": 1.4766,
+      "step": 2019
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993718839126957e-05,
+      "loss": 1.4658,
+      "step": 2020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937072356806492e-05,
+      "loss": 1.3555,
+      "step": 2021
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993695621560297e-05,
+      "loss": 1.29,
+      "step": 2022
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993683996766025e-05,
+      "loss": 1.4531,
+      "step": 2023
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993672361297958e-05,
+      "loss": 1.3711,
+      "step": 2024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993660715156221e-05,
+      "loss": 1.6172,
+      "step": 2025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936490583409393e-05,
+      "loss": 1.4473,
+      "step": 2026
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993637390852238e-05,
+      "loss": 1.6133,
+      "step": 2027
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936257126902422e-05,
+      "loss": 1.7871,
+      "step": 2028
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936140238550773e-05,
+      "loss": 1.3477,
+      "step": 2029
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993602324346869e-05,
+      "loss": 1.6738,
+      "step": 2030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9935906141657434e-05,
+      "loss": 1.418,
+      "step": 2031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935788933118258e-05,
+      "loss": 1.3457,
+      "step": 2032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935671617852422e-05,
+      "loss": 1.3906,
+      "step": 2033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935554195861187e-05,
+      "loss": 1.1904,
+      "step": 2034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935436667145813e-05,
+      "loss": 1.5137,
+      "step": 2035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935319031707564e-05,
+      "loss": 1.2842,
+      "step": 2036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935201289547704e-05,
+      "loss": 1.709,
+      "step": 2037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935083440667493e-05,
+      "loss": 1.6455,
+      "step": 2038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934965485068202e-05,
+      "loss": 1.5723,
+      "step": 2039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934847422751094e-05,
+      "loss": 1.7148,
+      "step": 2040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934729253717444e-05,
+      "loss": 1.0117,
+      "step": 2041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934610977968514e-05,
+      "loss": 1.1445,
+      "step": 2042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934492595505577e-05,
+      "loss": 1.334,
+      "step": 2043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934374106329905e-05,
+      "loss": 1.3066,
+      "step": 2044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934255510442773e-05,
+      "loss": 1.6504,
+      "step": 2045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934136807845454e-05,
+      "loss": 1.5361,
+      "step": 2046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934017998539216e-05,
+      "loss": 1.8223,
+      "step": 2047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933899082525346e-05,
+      "loss": 1.4463,
+      "step": 2048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993378005980511e-05,
+      "loss": 1.7754,
+      "step": 2049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933660930379803e-05,
+      "loss": 1.1846,
+      "step": 2050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933541694250684e-05,
+      "loss": 1.3623,
+      "step": 2051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993342235141905e-05,
+      "loss": 1.2354,
+      "step": 2052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933302901886177e-05,
+      "loss": 1.8027,
+      "step": 2053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933183345653347e-05,
+      "loss": 0.8877,
+      "step": 2054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933063682721847e-05,
+      "loss": 1.3936,
+      "step": 2055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993294391309296e-05,
+      "loss": 1.2666,
+      "step": 2056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993282403676797e-05,
+      "loss": 1.1543,
+      "step": 2057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993270405374817e-05,
+      "loss": 1.249,
+      "step": 2058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932583964034848e-05,
+      "loss": 1.1426,
+      "step": 2059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932463767629295e-05,
+      "loss": 1.4844,
+      "step": 2060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932343464532797e-05,
+      "loss": 1.1924,
+      "step": 2061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932223054746648e-05,
+      "loss": 1.0068,
+      "step": 2062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932102538272146e-05,
+      "loss": 1.5352,
+      "step": 2063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931981915110577e-05,
+      "loss": 1.5186,
+      "step": 2064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931861185263247e-05,
+      "loss": 1.7148,
+      "step": 2065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931740348731447e-05,
+      "loss": 1.29,
+      "step": 2066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931619405516474e-05,
+      "loss": 1.2285,
+      "step": 2067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993149835561963e-05,
+      "loss": 1.3711,
+      "step": 2068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931377199042213e-05,
+      "loss": 1.5254,
+      "step": 2069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931255935785524e-05,
+      "loss": 1.709,
+      "step": 2070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993113456585087e-05,
+      "loss": 0.9502,
+      "step": 2071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993101308923955e-05,
+      "loss": 1.5664,
+      "step": 2072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930891505952872e-05,
+      "loss": 1.5352,
+      "step": 2073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930769815992138e-05,
+      "loss": 1.4121,
+      "step": 2074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993064801935866e-05,
+      "loss": 1.3926,
+      "step": 2075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930526116053745e-05,
+      "loss": 1.2949,
+      "step": 2076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99304041060787e-05,
+      "loss": 1.416,
+      "step": 2077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930281989434835e-05,
+      "loss": 1.3262,
+      "step": 2078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930159766123468e-05,
+      "loss": 1.584,
+      "step": 2079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930037436145903e-05,
+      "loss": 1.6367,
+      "step": 2080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992991499950346e-05,
+      "loss": 1.2158,
+      "step": 2081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929792456197456e-05,
+      "loss": 1.2891,
+      "step": 2082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929669806229204e-05,
+      "loss": 1.2051,
+      "step": 2083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929547049600018e-05,
+      "loss": 1.3594,
+      "step": 2084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992942418631122e-05,
+      "loss": 1.583,
+      "step": 2085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929301216364133e-05,
+      "loss": 1.3008,
+      "step": 2086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929178139760076e-05,
+      "loss": 1.2812,
+      "step": 2087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929054956500368e-05,
+      "loss": 1.6172,
+      "step": 2088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928931666586333e-05,
+      "loss": 1.3604,
+      "step": 2089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928808270019297e-05,
+      "loss": 1.1992,
+      "step": 2090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928684766800588e-05,
+      "loss": 1.29,
+      "step": 2091
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928561156931526e-05,
+      "loss": 1.4639,
+      "step": 2092
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928437440413444e-05,
+      "loss": 1.2764,
+      "step": 2093
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928313617247667e-05,
+      "loss": 1.4883,
+      "step": 2094
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928189687435528e-05,
+      "loss": 1.3174,
+      "step": 2095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992806565097836e-05,
+      "loss": 1.7549,
+      "step": 2096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992794150787749e-05,
+      "loss": 1.5312,
+      "step": 2097
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992781725813426e-05,
+      "loss": 1.4238,
+      "step": 2098
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927692901749992e-05,
+      "loss": 1.666,
+      "step": 2099
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927568438726032e-05,
+      "loss": 1.5293,
+      "step": 2100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927443869063718e-05,
+      "loss": 1.4629,
+      "step": 2101
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927319192764377e-05,
+      "loss": 1.5898,
+      "step": 2102
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927194409829358e-05,
+      "loss": 1.2832,
+      "step": 2103
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927069520259996e-05,
+      "loss": 1.3623,
+      "step": 2104
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992694452405764e-05,
+      "loss": 1.541,
+      "step": 2105
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926819421223622e-05,
+      "loss": 1.4795,
+      "step": 2106
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926694211759298e-05,
+      "loss": 1.3164,
+      "step": 2107
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926568895666e-05,
+      "loss": 1.1748,
+      "step": 2108
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926443472945082e-05,
+      "loss": 1.1826,
+      "step": 2109
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992631794359789e-05,
+      "loss": 1.3418,
+      "step": 2110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926192307625775e-05,
+      "loss": 1.4775,
+      "step": 2111
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992606656503008e-05,
+      "loss": 1.335,
+      "step": 2112
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925940715812163e-05,
+      "loss": 1.5674,
+      "step": 2113
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992581475997337e-05,
+      "loss": 1.4375,
+      "step": 2114
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925688697515056e-05,
+      "loss": 1.7168,
+      "step": 2115
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925562528438574e-05,
+      "loss": 1.3672,
+      "step": 2116
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925436252745283e-05,
+      "loss": 1.2305,
+      "step": 2117
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925309870436537e-05,
+      "loss": 1.3867,
+      "step": 2118
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925183381513692e-05,
+      "loss": 1.3164,
+      "step": 2119
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992505678597811e-05,
+      "loss": 1.4355,
+      "step": 2120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924930083831147e-05,
+      "loss": 1.3555,
+      "step": 2121
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924803275074167e-05,
+      "loss": 1.4551,
+      "step": 2122
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992467635970853e-05,
+      "loss": 1.6797,
+      "step": 2123
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924549337735603e-05,
+      "loss": 1.3965,
+      "step": 2124
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924422209156747e-05,
+      "loss": 1.2197,
+      "step": 2125
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992429497397333e-05,
+      "loss": 1.1572,
+      "step": 2126
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924167632186714e-05,
+      "loss": 1.4111,
+      "step": 2127
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992404018379827e-05,
+      "loss": 1.6406,
+      "step": 2128
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992391262880937e-05,
+      "loss": 1.4336,
+      "step": 2129
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923784967221376e-05,
+      "loss": 1.3477,
+      "step": 2130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992365719903567e-05,
+      "loss": 1.6357,
+      "step": 2131
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923529324253614e-05,
+      "loss": 1.3447,
+      "step": 2132
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923401342876588e-05,
+      "loss": 1.6973,
+      "step": 2133
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992327325490597e-05,
+      "loss": 1.54,
+      "step": 2134
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923145060343127e-05,
+      "loss": 1.1333,
+      "step": 2135
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992301675918944e-05,
+      "loss": 0.9844,
+      "step": 2136
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922888351446288e-05,
+      "loss": 1.3262,
+      "step": 2137
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992275983711505e-05,
+      "loss": 1.1602,
+      "step": 2138
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922631216197102e-05,
+      "loss": 1.3047,
+      "step": 2139
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992250248869384e-05,
+      "loss": 1.2871,
+      "step": 2140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922373654606625e-05,
+      "loss": 1.2656,
+      "step": 2141
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992224471393686e-05,
+      "loss": 1.7324,
+      "step": 2142
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992211566668592e-05,
+      "loss": 1.9551,
+      "step": 2143
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992198651285519e-05,
+      "loss": 1.4785,
+      "step": 2144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921857252446063e-05,
+      "loss": 1.1377,
+      "step": 2145
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992172788545993e-05,
+      "loss": 1.1582,
+      "step": 2146
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921598411898167e-05,
+      "loss": 1.3965,
+      "step": 2147
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921468831762178e-05,
+      "loss": 1.0654,
+      "step": 2148
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992133914505335e-05,
+      "loss": 1.4424,
+      "step": 2149
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921209351773077e-05,
+      "loss": 1.3965,
+      "step": 2150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921079451922754e-05,
+      "loss": 1.0791,
+      "step": 2151
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920949445503774e-05,
+      "loss": 1.3135,
+      "step": 2152
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920819332517535e-05,
+      "loss": 1.4004,
+      "step": 2153
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920689112965434e-05,
+      "loss": 1.2832,
+      "step": 2154
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992055878684887e-05,
+      "loss": 1.5586,
+      "step": 2155
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920428354169246e-05,
+      "loss": 1.2363,
+      "step": 2156
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920297814927957e-05,
+      "loss": 1.3027,
+      "step": 2157
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920167169126407e-05,
+      "loss": 1.3828,
+      "step": 2158
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920036416766e-05,
+      "loss": 1.5391,
+      "step": 2159
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919905557848146e-05,
+      "loss": 1.2744,
+      "step": 2160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991977459237424e-05,
+      "loss": 1.2598,
+      "step": 2161
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919643520345698e-05,
+      "loss": 1.8809,
+      "step": 2162
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919512341763924e-05,
+      "loss": 1.5176,
+      "step": 2163
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919381056630328e-05,
+      "loss": 1.3193,
+      "step": 2164
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991924966494632e-05,
+      "loss": 1.0791,
+      "step": 2165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919118166713306e-05,
+      "loss": 1.7695,
+      "step": 2166
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918986561932708e-05,
+      "loss": 1.293,
+      "step": 2167
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918854850605936e-05,
+      "loss": 1.8691,
+      "step": 2168
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99187230327344e-05,
+      "loss": 1.3672,
+      "step": 2169
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918591108319523e-05,
+      "loss": 1.0859,
+      "step": 2170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991845907736272e-05,
+      "loss": 1.1436,
+      "step": 2171
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918326939865407e-05,
+      "loss": 1.5518,
+      "step": 2172
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918194695829003e-05,
+      "loss": 1.2363,
+      "step": 2173
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918062345254933e-05,
+      "loss": 1.3018,
+      "step": 2174
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917929888144613e-05,
+      "loss": 1.0342,
+      "step": 2175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991779732449947e-05,
+      "loss": 1.1299,
+      "step": 2176
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917664654320926e-05,
+      "loss": 1.543,
+      "step": 2177
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917531877610408e-05,
+      "loss": 1.3184,
+      "step": 2178
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991739899436934e-05,
+      "loss": 1.2031,
+      "step": 2179
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917266004599152e-05,
+      "loss": 1.0796,
+      "step": 2180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917132908301267e-05,
+      "loss": 1.2178,
+      "step": 2181
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916999705477123e-05,
+      "loss": 1.4668,
+      "step": 2182
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916866396128144e-05,
+      "loss": 1.4375,
+      "step": 2183
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916732980255762e-05,
+      "loss": 1.21,
+      "step": 2184
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916599457861416e-05,
+      "loss": 1.4648,
+      "step": 2185
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916465828946533e-05,
+      "loss": 1.1816,
+      "step": 2186
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916332093512555e-05,
+      "loss": 1.2451,
+      "step": 2187
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916198251560917e-05,
+      "loss": 1.5039,
+      "step": 2188
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916064303093053e-05,
+      "loss": 1.3682,
+      "step": 2189
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915930248110403e-05,
+      "loss": 1.4277,
+      "step": 2190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915796086614407e-05,
+      "loss": 1.0205,
+      "step": 2191
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915661818606508e-05,
+      "loss": 1.0742,
+      "step": 2192
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915527444088153e-05,
+      "loss": 1.4141,
+      "step": 2193
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915392963060775e-05,
+      "loss": 1.6348,
+      "step": 2194
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991525837552582e-05,
+      "loss": 1.3232,
+      "step": 2195
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915123681484743e-05,
+      "loss": 1.1367,
+      "step": 2196
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991498888093898e-05,
+      "loss": 1.041,
+      "step": 2197
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914853973889988e-05,
+      "loss": 1.3916,
+      "step": 2198
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914718960339213e-05,
+      "loss": 1.1465,
+      "step": 2199
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99145838402881e-05,
+      "loss": 1.3301,
+      "step": 2200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914448613738107e-05,
+      "loss": 1.2637,
+      "step": 2201
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991431328069068e-05,
+      "loss": 1.3857,
+      "step": 2202
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991417784114728e-05,
+      "loss": 1.2573,
+      "step": 2203
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914042295109363e-05,
+      "loss": 1.2695,
+      "step": 2204
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913906642578376e-05,
+      "loss": 1.3613,
+      "step": 2205
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913770883555782e-05,
+      "loss": 1.7812,
+      "step": 2206
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991363501804304e-05,
+      "loss": 1.5801,
+      "step": 2207
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913499046041603e-05,
+      "loss": 1.0625,
+      "step": 2208
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991336296755294e-05,
+      "loss": 1.5049,
+      "step": 2209
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913226782578506e-05,
+      "loss": 1.8594,
+      "step": 2210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991309049111977e-05,
+      "loss": 1.292,
+      "step": 2211
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991295409317819e-05,
+      "loss": 1.666,
+      "step": 2212
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912817588755234e-05,
+      "loss": 1.4277,
+      "step": 2213
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912680977852367e-05,
+      "loss": 1.4551,
+      "step": 2214
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991254426047106e-05,
+      "loss": 1.5586,
+      "step": 2215
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991240743661278e-05,
+      "loss": 1.4258,
+      "step": 2216
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912270506278993e-05,
+      "loss": 1.4707,
+      "step": 2217
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912133469471174e-05,
+      "loss": 1.3477,
+      "step": 2218
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911996326190793e-05,
+      "loss": 1.5068,
+      "step": 2219
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911859076439322e-05,
+      "loss": 1.2783,
+      "step": 2220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911721720218243e-05,
+      "loss": 1.6543,
+      "step": 2221
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991158425752902e-05,
+      "loss": 1.1855,
+      "step": 2222
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991144668837314e-05,
+      "loss": 1.1357,
+      "step": 2223
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991130901275207e-05,
+      "loss": 1.1689,
+      "step": 2224
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911171230667296e-05,
+      "loss": 1.0088,
+      "step": 2225
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911033342120297e-05,
+      "loss": 1.7832,
+      "step": 2226
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910895347112556e-05,
+      "loss": 1.3398,
+      "step": 2227
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991075724564555e-05,
+      "loss": 1.2627,
+      "step": 2228
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991061903772077e-05,
+      "loss": 1.3203,
+      "step": 2229
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910480723339693e-05,
+      "loss": 1.4746,
+      "step": 2230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991034230250381e-05,
+      "loss": 1.2988,
+      "step": 2231
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910203775214602e-05,
+      "loss": 1.0703,
+      "step": 2232
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910065141473566e-05,
+      "loss": 1.4863,
+      "step": 2233
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990992640128218e-05,
+      "loss": 1.2461,
+      "step": 2234
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909787554641942e-05,
+      "loss": 1.4189,
+      "step": 2235
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909648601554347e-05,
+      "loss": 1.4521,
+      "step": 2236
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909509542020877e-05,
+      "loss": 1.0791,
+      "step": 2237
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909370376043032e-05,
+      "loss": 1.3203,
+      "step": 2238
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909231103622305e-05,
+      "loss": 1.2236,
+      "step": 2239
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990909172476019e-05,
+      "loss": 1.3789,
+      "step": 2240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908952239458194e-05,
+      "loss": 1.2959,
+      "step": 2241
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908812647717805e-05,
+      "loss": 1.4043,
+      "step": 2242
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908672949540524e-05,
+      "loss": 1.4766,
+      "step": 2243
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908533144927853e-05,
+      "loss": 0.9199,
+      "step": 2244
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908393233881296e-05,
+      "loss": 0.9492,
+      "step": 2245
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908253216402354e-05,
+      "loss": 1.5332,
+      "step": 2246
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990811309249253e-05,
+      "loss": 1.707,
+      "step": 2247
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907972862153332e-05,
+      "loss": 1.3965,
+      "step": 2248
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990783252538626e-05,
+      "loss": 1.4238,
+      "step": 2249
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907692082192832e-05,
+      "loss": 1.4531,
+      "step": 2250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907551532574546e-05,
+      "loss": 1.2939,
+      "step": 2251
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990741087653292e-05,
+      "loss": 1.5723,
+      "step": 2252
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907270114069456e-05,
+      "loss": 1.4922,
+      "step": 2253
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907129245185674e-05,
+      "loss": 1.6406,
+      "step": 2254
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906988269883083e-05,
+      "loss": 1.5195,
+      "step": 2255
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99068471881632e-05,
+      "loss": 1.0713,
+      "step": 2256
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990670600002754e-05,
+      "loss": 1.4824,
+      "step": 2257
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906564705477616e-05,
+      "loss": 1.3652,
+      "step": 2258
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990642330451495e-05,
+      "loss": 1.2539,
+      "step": 2259
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906281797141057e-05,
+      "loss": 1.3184,
+      "step": 2260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990614018335746e-05,
+      "loss": 0.9863,
+      "step": 2261
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905998463165683e-05,
+      "loss": 1.584,
+      "step": 2262
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905856636567237e-05,
+      "loss": 1.373,
+      "step": 2263
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905714703563662e-05,
+      "loss": 1.3936,
+      "step": 2264
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990557266415647e-05,
+      "loss": 1.4209,
+      "step": 2265
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905430518347185e-05,
+      "loss": 2.0508,
+      "step": 2266
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905288266137347e-05,
+      "loss": 1.2754,
+      "step": 2267
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905145907528474e-05,
+      "loss": 1.0254,
+      "step": 2268
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905003442522095e-05,
+      "loss": 1.332,
+      "step": 2269
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904860871119746e-05,
+      "loss": 1.0488,
+      "step": 2270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904718193322953e-05,
+      "loss": 1.4453,
+      "step": 2271
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904575409133252e-05,
+      "loss": 1.7188,
+      "step": 2272
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904432518552177e-05,
+      "loss": 1.3418,
+      "step": 2273
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990428952158126e-05,
+      "loss": 1.1807,
+      "step": 2274
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990414641822204e-05,
+      "loss": 1.6914,
+      "step": 2275
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990400320847605e-05,
+      "loss": 1.5137,
+      "step": 2276
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990385989234483e-05,
+      "loss": 1.4453,
+      "step": 2277
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903716469829925e-05,
+      "loss": 1.0879,
+      "step": 2278
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990357294093287e-05,
+      "loss": 1.6133,
+      "step": 2279
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903429305655206e-05,
+      "loss": 1.2139,
+      "step": 2280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990328556399848e-05,
+      "loss": 1.5791,
+      "step": 2281
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903141715964233e-05,
+      "loss": 1.2061,
+      "step": 2282
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990299776155401e-05,
+      "loss": 1.4229,
+      "step": 2283
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902853700769355e-05,
+      "loss": 1.4922,
+      "step": 2284
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990270953361182e-05,
+      "loss": 1.4033,
+      "step": 2285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902565260082953e-05,
+      "loss": 1.6094,
+      "step": 2286
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902420880184305e-05,
+      "loss": 0.8901,
+      "step": 2287
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902276393917424e-05,
+      "loss": 1.5039,
+      "step": 2288
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902131801283863e-05,
+      "loss": 1.3574,
+      "step": 2289
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901987102285172e-05,
+      "loss": 1.4609,
+      "step": 2290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901842296922912e-05,
+      "loss": 1.4297,
+      "step": 2291
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990169738519863e-05,
+      "loss": 1.668,
+      "step": 2292
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901552367113893e-05,
+      "loss": 1.1572,
+      "step": 2293
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901407242670248e-05,
+      "loss": 1.4238,
+      "step": 2294
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990126201186926e-05,
+      "loss": 1.1978,
+      "step": 2295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901116674712492e-05,
+      "loss": 1.6758,
+      "step": 2296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900971231201497e-05,
+      "loss": 1.3984,
+      "step": 2297
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900825681337842e-05,
+      "loss": 1.4863,
+      "step": 2298
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990068002512309e-05,
+      "loss": 1.2314,
+      "step": 2299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900534262558807e-05,
+      "loss": 1.2129,
+      "step": 2300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900388393646555e-05,
+      "loss": 1.5469,
+      "step": 2301
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900242418387904e-05,
+      "loss": 1.3135,
+      "step": 2302
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900096336784425e-05,
+      "loss": 1.2334,
+      "step": 2303
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899950148837678e-05,
+      "loss": 1.2998,
+      "step": 2304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989980385454924e-05,
+      "loss": 1.6973,
+      "step": 2305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989965745392068e-05,
+      "loss": 1.5361,
+      "step": 2306
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989951094695357e-05,
+      "loss": 1.3291,
+      "step": 2307
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899364333649493e-05,
+      "loss": 1.2305,
+      "step": 2308
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899217614010007e-05,
+      "loss": 1.6855,
+      "step": 2309
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98990707880367e-05,
+      "loss": 1.1992,
+      "step": 2310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989892385573115e-05,
+      "loss": 1.6855,
+      "step": 2311
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898776817094928e-05,
+      "loss": 1.6738,
+      "step": 2312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898629672129614e-05,
+      "loss": 1.54,
+      "step": 2313
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898482420836797e-05,
+      "loss": 1.4434,
+      "step": 2314
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989833506321805e-05,
+      "loss": 1.2031,
+      "step": 2315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898187599274963e-05,
+      "loss": 1.4785,
+      "step": 2316
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989804002900911e-05,
+      "loss": 1.2881,
+      "step": 2317
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897892352422087e-05,
+      "loss": 1.334,
+      "step": 2318
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897744569515475e-05,
+      "loss": 1.4248,
+      "step": 2319
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897596680290862e-05,
+      "loss": 1.6406,
+      "step": 2320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897448684749835e-05,
+      "loss": 1.4062,
+      "step": 2321
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897300582893986e-05,
+      "loss": 1.2471,
+      "step": 2322
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897152374724908e-05,
+      "loss": 1.2783,
+      "step": 2323
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989700406024419e-05,
+      "loss": 1.3066,
+      "step": 2324
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989685563945342e-05,
+      "loss": 1.3281,
+      "step": 2325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896707112354202e-05,
+      "loss": 1.2754,
+      "step": 2326
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896558478948126e-05,
+      "loss": 1.7148,
+      "step": 2327
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989640973923679e-05,
+      "loss": 1.3721,
+      "step": 2328
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989626089322179e-05,
+      "loss": 1.1729,
+      "step": 2329
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896111940904724e-05,
+      "loss": 1.3428,
+      "step": 2330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895962882287197e-05,
+      "loss": 1.2373,
+      "step": 2331
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989581371737081e-05,
+      "loss": 1.4414,
+      "step": 2332
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895664446157158e-05,
+      "loss": 1.4316,
+      "step": 2333
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989551506864785e-05,
+      "loss": 1.5371,
+      "step": 2334
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895365584844493e-05,
+      "loss": 1.2344,
+      "step": 2335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895215994748683e-05,
+      "loss": 1.5938,
+      "step": 2336
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989506629836204e-05,
+      "loss": 1.6836,
+      "step": 2337
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989491649568616e-05,
+      "loss": 1.4229,
+      "step": 2338
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894766586722658e-05,
+      "loss": 1.6816,
+      "step": 2339
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894616571473144e-05,
+      "loss": 1.4756,
+      "step": 2340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989446644993923e-05,
+      "loss": 1.2354,
+      "step": 2341
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894316222122526e-05,
+      "loss": 1.7871,
+      "step": 2342
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894165888024647e-05,
+      "loss": 1.4805,
+      "step": 2343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989401544764721e-05,
+      "loss": 1.502,
+      "step": 2344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893864900991823e-05,
+      "loss": 1.8486,
+      "step": 2345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893714248060118e-05,
+      "loss": 1.2871,
+      "step": 2346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98935634888537e-05,
+      "loss": 1.2881,
+      "step": 2347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989341262337419e-05,
+      "loss": 1.9844,
+      "step": 2348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893261651623213e-05,
+      "loss": 1.3398,
+      "step": 2349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989311057360239e-05,
+      "loss": 1.2852,
+      "step": 2350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989295938931334e-05,
+      "loss": 1.2656,
+      "step": 2351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989280809875769e-05,
+      "loss": 1.4199,
+      "step": 2352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989265670193707e-05,
+      "loss": 1.5527,
+      "step": 2353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892505198853097e-05,
+      "loss": 1.2246,
+      "step": 2354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892353589507404e-05,
+      "loss": 1.4727,
+      "step": 2355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892201873901617e-05,
+      "loss": 1.4648,
+      "step": 2356
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892050052037366e-05,
+      "loss": 1.5244,
+      "step": 2357
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891898123916285e-05,
+      "loss": 1.3359,
+      "step": 2358
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989174608954e-05,
+      "loss": 1.4736,
+      "step": 2359
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891593948910152e-05,
+      "loss": 1.0947,
+      "step": 2360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891441702028363e-05,
+      "loss": 1.1963,
+      "step": 2361
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891289348896283e-05,
+      "loss": 1.7109,
+      "step": 2362
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989113688951554e-05,
+      "loss": 1.5723,
+      "step": 2363
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890984323887773e-05,
+      "loss": 1.4639,
+      "step": 2364
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989083165201462e-05,
+      "loss": 1.0889,
+      "step": 2365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890678873897724e-05,
+      "loss": 1.1797,
+      "step": 2366
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890525989538726e-05,
+      "loss": 1.5361,
+      "step": 2367
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890372998939265e-05,
+      "loss": 1.2422,
+      "step": 2368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890219902100982e-05,
+      "loss": 1.249,
+      "step": 2369
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989006669902553e-05,
+      "loss": 1.502,
+      "step": 2370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889913389714546e-05,
+      "loss": 1.583,
+      "step": 2371
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889759974169684e-05,
+      "loss": 1.4658,
+      "step": 2372
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889606452392586e-05,
+      "loss": 1.4102,
+      "step": 2373
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889452824384908e-05,
+      "loss": 1.5098,
+      "step": 2374
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988929909014829e-05,
+      "loss": 1.3848,
+      "step": 2375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889145249684393e-05,
+      "loss": 1.248,
+      "step": 2376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888991302994864e-05,
+      "loss": 1.7285,
+      "step": 2377
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988883725008136e-05,
+      "loss": 1.2256,
+      "step": 2378
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888683090945536e-05,
+      "loss": 1.3193,
+      "step": 2379
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888528825589045e-05,
+      "loss": 1.4893,
+      "step": 2380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888374454013542e-05,
+      "loss": 1.2021,
+      "step": 2381
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988821997622069e-05,
+      "loss": 1.2031,
+      "step": 2382
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988806539221215e-05,
+      "loss": 1.1484,
+      "step": 2383
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887910701989574e-05,
+      "loss": 0.9873,
+      "step": 2384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887755905554633e-05,
+      "loss": 1.21,
+      "step": 2385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887601002908984e-05,
+      "loss": 1.6621,
+      "step": 2386
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887445994054294e-05,
+      "loss": 0.8896,
+      "step": 2387
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887290878992223e-05,
+      "loss": 1.1123,
+      "step": 2388
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887135657724445e-05,
+      "loss": 1.3643,
+      "step": 2389
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988698033025262e-05,
+      "loss": 1.3252,
+      "step": 2390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988682489657842e-05,
+      "loss": 1.3613,
+      "step": 2391
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988666935670352e-05,
+      "loss": 1.3281,
+      "step": 2392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988651371062958e-05,
+      "loss": 1.6035,
+      "step": 2393
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886357958358276e-05,
+      "loss": 1.3223,
+      "step": 2394
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988620209989128e-05,
+      "loss": 1.2793,
+      "step": 2395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886046135230272e-05,
+      "loss": 1.5479,
+      "step": 2396
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988589006437692e-05,
+      "loss": 1.1289,
+      "step": 2397
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885733887332907e-05,
+      "loss": 1.4453,
+      "step": 2398
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988557760409991e-05,
+      "loss": 1.5391,
+      "step": 2399
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885421214679598e-05,
+      "loss": 1.6035,
+      "step": 2400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885264719073664e-05,
+      "loss": 1.459,
+      "step": 2401
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988510811728378e-05,
+      "loss": 1.1641,
+      "step": 2402
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884951409311632e-05,
+      "loss": 1.0547,
+      "step": 2403
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98847945951589e-05,
+      "loss": 1.2256,
+      "step": 2404
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884637674827277e-05,
+      "loss": 1.3262,
+      "step": 2405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988448064831844e-05,
+      "loss": 1.3691,
+      "step": 2406
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884323515634076e-05,
+      "loss": 0.9824,
+      "step": 2407
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884166276775878e-05,
+      "loss": 1.4756,
+      "step": 2408
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884008931745532e-05,
+      "loss": 1.4648,
+      "step": 2409
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883851480544726e-05,
+      "loss": 1.4883,
+      "step": 2410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883693923175153e-05,
+      "loss": 1.1162,
+      "step": 2411
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883536259638508e-05,
+      "loss": 1.3672,
+      "step": 2412
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988337848993648e-05,
+      "loss": 1.0508,
+      "step": 2413
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988322061407077e-05,
+      "loss": 1.1504,
+      "step": 2414
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988306263204307e-05,
+      "loss": 1.2793,
+      "step": 2415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988290454385507e-05,
+      "loss": 0.835,
+      "step": 2416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882746349508485e-05,
+      "loss": 1.3301,
+      "step": 2417
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882588049005002e-05,
+      "loss": 1.1475,
+      "step": 2418
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882429642346318e-05,
+      "loss": 1.1924,
+      "step": 2419
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882271129534148e-05,
+      "loss": 1.5938,
+      "step": 2420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882112510570183e-05,
+      "loss": 1.2393,
+      "step": 2421
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988195378545613e-05,
+      "loss": 1.4434,
+      "step": 2422
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881794954193698e-05,
+      "loss": 1.4922,
+      "step": 2423
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988163601678459e-05,
+      "loss": 1.417,
+      "step": 2424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881476973230513e-05,
+      "loss": 1.2437,
+      "step": 2425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881317823533176e-05,
+      "loss": 1.0654,
+      "step": 2426
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881158567694285e-05,
+      "loss": 1.6797,
+      "step": 2427
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880999205715557e-05,
+      "loss": 1.8965,
+      "step": 2428
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880839737598697e-05,
+      "loss": 1.4727,
+      "step": 2429
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880680163345425e-05,
+      "loss": 1.4492,
+      "step": 2430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988052048295745e-05,
+      "loss": 1.2998,
+      "step": 2431
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988036069643649e-05,
+      "loss": 1.2773,
+      "step": 2432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988020080378426e-05,
+      "loss": 1.5303,
+      "step": 2433
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988004080500248e-05,
+      "loss": 1.6465,
+      "step": 2434
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987988070009286e-05,
+      "loss": 1.1553,
+      "step": 2435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879720489057132e-05,
+      "loss": 1.4482,
+      "step": 2436
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987956017189701e-05,
+      "loss": 1.665,
+      "step": 2437
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987939974861422e-05,
+      "loss": 1.4316,
+      "step": 2438
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879239219210478e-05,
+      "loss": 1.4355,
+      "step": 2439
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879078583687515e-05,
+      "loss": 1.4863,
+      "step": 2440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987891784204705e-05,
+      "loss": 1.5078,
+      "step": 2441
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987875699429082e-05,
+      "loss": 1.5859,
+      "step": 2442
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878596040420547e-05,
+      "loss": 1.0039,
+      "step": 2443
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878434980437955e-05,
+      "loss": 1.584,
+      "step": 2444
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878273814344782e-05,
+      "loss": 1.3789,
+      "step": 2445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878112542142755e-05,
+      "loss": 1.4443,
+      "step": 2446
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877951163833605e-05,
+      "loss": 1.0664,
+      "step": 2447
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877789679419067e-05,
+      "loss": 1.417,
+      "step": 2448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987762808890088e-05,
+      "loss": 1.4023,
+      "step": 2449
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877466392280773e-05,
+      "loss": 1.2441,
+      "step": 2450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987730458956049e-05,
+      "loss": 1.499,
+      "step": 2451
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987714268074176e-05,
+      "loss": 1.6621,
+      "step": 2452
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987698066582633e-05,
+      "loss": 1.3965,
+      "step": 2453
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987681854481594e-05,
+      "loss": 1.4561,
+      "step": 2454
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876656317712323e-05,
+      "loss": 1.3906,
+      "step": 2455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876493984517233e-05,
+      "loss": 1.248,
+      "step": 2456
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876331545232404e-05,
+      "loss": 1.1348,
+      "step": 2457
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876168999859585e-05,
+      "loss": 1.4727,
+      "step": 2458
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876006348400522e-05,
+      "loss": 1.3545,
+      "step": 2459
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875843590856965e-05,
+      "loss": 1.0957,
+      "step": 2460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875680727230656e-05,
+      "loss": 1.6133,
+      "step": 2461
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987551775752335e-05,
+      "loss": 1.1855,
+      "step": 2462
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875354681736796e-05,
+      "loss": 1.499,
+      "step": 2463
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875191499872744e-05,
+      "loss": 1.4414,
+      "step": 2464
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875028211932948e-05,
+      "loss": 1.3389,
+      "step": 2465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987486481791916e-05,
+      "loss": 1.4492,
+      "step": 2466
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874701317833143e-05,
+      "loss": 1.6719,
+      "step": 2467
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987453771167664e-05,
+      "loss": 1.2012,
+      "step": 2468
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987437399945142e-05,
+      "loss": 1.1299,
+      "step": 2469
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874210181159237e-05,
+      "loss": 1.2266,
+      "step": 2470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987404625680185e-05,
+      "loss": 1.6855,
+      "step": 2471
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987388222638102e-05,
+      "loss": 1.4062,
+      "step": 2472
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873718089898512e-05,
+      "loss": 1.1719,
+      "step": 2473
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873553847356087e-05,
+      "loss": 1.0264,
+      "step": 2474
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873389498755505e-05,
+      "loss": 1.5479,
+      "step": 2475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873225044098537e-05,
+      "loss": 1.5527,
+      "step": 2476
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873060483386953e-05,
+      "loss": 1.6523,
+      "step": 2477
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987289581662251e-05,
+      "loss": 1.3066,
+      "step": 2478
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872731043806984e-05,
+      "loss": 1.2568,
+      "step": 2479
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872566164942144e-05,
+      "loss": 1.3555,
+      "step": 2480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872401180029757e-05,
+      "loss": 1.2441,
+      "step": 2481
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98722360890716e-05,
+      "loss": 1.4551,
+      "step": 2482
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987207089206945e-05,
+      "loss": 1.2168,
+      "step": 2483
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987190558902507e-05,
+      "loss": 1.3555,
+      "step": 2484
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871740179940248e-05,
+      "loss": 1.2695,
+      "step": 2485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987157466481675e-05,
+      "loss": 1.2871,
+      "step": 2486
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987140904365636e-05,
+      "loss": 1.3496,
+      "step": 2487
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871243316460855e-05,
+      "loss": 1.4355,
+      "step": 2488
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871077483232016e-05,
+      "loss": 1.3359,
+      "step": 2489
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870911543971626e-05,
+      "loss": 1.5957,
+      "step": 2490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870745498681465e-05,
+      "loss": 1.2773,
+      "step": 2491
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870579347363318e-05,
+      "loss": 1.2207,
+      "step": 2492
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870413090018966e-05,
+      "loss": 1.375,
+      "step": 2493
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98702467266502e-05,
+      "loss": 1.3789,
+      "step": 2494
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870080257258804e-05,
+      "loss": 1.292,
+      "step": 2495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869913681846567e-05,
+      "loss": 1.2715,
+      "step": 2496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986974700041528e-05,
+      "loss": 1.584,
+      "step": 2497
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986958021296673e-05,
+      "loss": 1.2578,
+      "step": 2498
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869413319502714e-05,
+      "loss": 1.0342,
+      "step": 2499
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869246320025018e-05,
+      "loss": 1.2803,
+      "step": 2500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986907921453544e-05,
+      "loss": 1.3105,
+      "step": 2501
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986891200303577e-05,
+      "loss": 1.2715,
+      "step": 2502
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868744685527815e-05,
+      "loss": 1.0461,
+      "step": 2503
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986857726201336e-05,
+      "loss": 1.4292,
+      "step": 2504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986840973249421e-05,
+      "loss": 0.8936,
+      "step": 2505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868242096972166e-05,
+      "loss": 1.3115,
+      "step": 2506
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868074355449024e-05,
+      "loss": 0.9023,
+      "step": 2507
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867906507926586e-05,
+      "loss": 1.1685,
+      "step": 2508
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867738554406656e-05,
+      "loss": 1.3926,
+      "step": 2509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986757049489104e-05,
+      "loss": 1.2012,
+      "step": 2510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867402329381546e-05,
+      "loss": 1.6348,
+      "step": 2511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867234057879972e-05,
+      "loss": 1.7598,
+      "step": 2512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986706568038813e-05,
+      "loss": 1.1221,
+      "step": 2513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986689719690783e-05,
+      "loss": 1.2236,
+      "step": 2514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866728607440883e-05,
+      "loss": 1.0098,
+      "step": 2515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986655991198909e-05,
+      "loss": 1.4434,
+      "step": 2516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866391110554278e-05,
+      "loss": 1.1846,
+      "step": 2517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986622220313825e-05,
+      "loss": 1.1904,
+      "step": 2518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986605318974282e-05,
+      "loss": 1.498,
+      "step": 2519
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986588407036981e-05,
+      "loss": 1.5908,
+      "step": 2520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986571484502103e-05,
+      "loss": 0.9053,
+      "step": 2521
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865545513698304e-05,
+      "loss": 1.6152,
+      "step": 2522
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865376076403445e-05,
+      "loss": 1.2119,
+      "step": 2523
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986520653313828e-05,
+      "loss": 1.1709,
+      "step": 2524
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865036883904622e-05,
+      "loss": 1.6084,
+      "step": 2525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864867128704298e-05,
+      "loss": 1.0381,
+      "step": 2526
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986469726753913e-05,
+      "loss": 1.5645,
+      "step": 2527
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864527300410944e-05,
+      "loss": 1.3398,
+      "step": 2528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864357227321567e-05,
+      "loss": 1.3594,
+      "step": 2529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986418704827282e-05,
+      "loss": 1.0337,
+      "step": 2530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864016763266537e-05,
+      "loss": 1.1982,
+      "step": 2531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863846372304543e-05,
+      "loss": 1.2285,
+      "step": 2532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863675875388672e-05,
+      "loss": 1.3066,
+      "step": 2533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863505272520753e-05,
+      "loss": 1.541,
+      "step": 2534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863334563702618e-05,
+      "loss": 1.2891,
+      "step": 2535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863163748936105e-05,
+      "loss": 1.46,
+      "step": 2536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862992828223043e-05,
+      "loss": 1.3672,
+      "step": 2537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862821801565272e-05,
+      "loss": 1.457,
+      "step": 2538
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862650668964628e-05,
+      "loss": 1.2275,
+      "step": 2539
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862479430422947e-05,
+      "loss": 1.0161,
+      "step": 2540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862308085942073e-05,
+      "loss": 1.5898,
+      "step": 2541
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862136635523842e-05,
+      "loss": 1.3125,
+      "step": 2542
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98619650791701e-05,
+      "loss": 1.4883,
+      "step": 2543
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861793416882685e-05,
+      "loss": 1.4824,
+      "step": 2544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861621648663446e-05,
+      "loss": 1.0908,
+      "step": 2545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861449774514225e-05,
+      "loss": 1.4355,
+      "step": 2546
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861277794436863e-05,
+      "loss": 1.542,
+      "step": 2547
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986110570843322e-05,
+      "loss": 1.3281,
+      "step": 2548
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860933516505135e-05,
+      "loss": 1.0391,
+      "step": 2549
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986076121865446e-05,
+      "loss": 1.3867,
+      "step": 2550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860588814883044e-05,
+      "loss": 1.3086,
+      "step": 2551
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986041630519274e-05,
+      "loss": 1.334,
+      "step": 2552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860243689585407e-05,
+      "loss": 0.9912,
+      "step": 2553
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986007096806289e-05,
+      "loss": 1.0518,
+      "step": 2554
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985989814062705e-05,
+      "loss": 1.7168,
+      "step": 2555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985972520727974e-05,
+      "loss": 1.5137,
+      "step": 2556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859552168022823e-05,
+      "loss": 1.2002,
+      "step": 2557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985937902285815e-05,
+      "loss": 0.9985,
+      "step": 2558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859205771787587e-05,
+      "loss": 1.375,
+      "step": 2559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985903241481299e-05,
+      "loss": 1.1074,
+      "step": 2560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858858951936225e-05,
+      "loss": 1.2041,
+      "step": 2561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858685383159158e-05,
+      "loss": 1.2109,
+      "step": 2562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985851170848365e-05,
+      "loss": 1.6631,
+      "step": 2563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985833792791156e-05,
+      "loss": 1.3477,
+      "step": 2564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858164041444764e-05,
+      "loss": 1.5488,
+      "step": 2565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857990049085127e-05,
+      "loss": 1.1299,
+      "step": 2566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857815950834517e-05,
+      "loss": 1.2461,
+      "step": 2567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857641746694807e-05,
+      "loss": 1.2783,
+      "step": 2568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857467436667864e-05,
+      "loss": 1.1318,
+      "step": 2569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857293020755565e-05,
+      "loss": 1.1387,
+      "step": 2570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857118498959776e-05,
+      "loss": 1.3672,
+      "step": 2571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856943871282383e-05,
+      "loss": 1.2344,
+      "step": 2572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985676913772525e-05,
+      "loss": 1.3828,
+      "step": 2573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856594298290265e-05,
+      "loss": 0.9365,
+      "step": 2574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856419352979294e-05,
+      "loss": 0.916,
+      "step": 2575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856244301794226e-05,
+      "loss": 1.6055,
+      "step": 2576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856069144736934e-05,
+      "loss": 1.4844,
+      "step": 2577
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985589388180931e-05,
+      "loss": 1.582,
+      "step": 2578
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855718513013226e-05,
+      "loss": 1.707,
+      "step": 2579
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985554303835057e-05,
+      "loss": 0.9268,
+      "step": 2580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985536745782323e-05,
+      "loss": 1.2285,
+      "step": 2581
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855191771433083e-05,
+      "loss": 1.2578,
+      "step": 2582
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855015979182027e-05,
+      "loss": 1.5645,
+      "step": 2583
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854840081071944e-05,
+      "loss": 1.0127,
+      "step": 2584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985466407710472e-05,
+      "loss": 1.2002,
+      "step": 2585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854487967282256e-05,
+      "loss": 1.1035,
+      "step": 2586
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854311751606433e-05,
+      "loss": 1.0244,
+      "step": 2587
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854135430079152e-05,
+      "loss": 0.9756,
+      "step": 2588
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853959002702304e-05,
+      "loss": 1.3438,
+      "step": 2589
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853782469477783e-05,
+      "loss": 1.0352,
+      "step": 2590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853605830407482e-05,
+      "loss": 1.5371,
+      "step": 2591
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853429085493307e-05,
+      "loss": 1.4346,
+      "step": 2592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853252234737154e-05,
+      "loss": 1.2129,
+      "step": 2593
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853075278140913e-05,
+      "loss": 1.2441,
+      "step": 2594
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98528982157065e-05,
+      "loss": 0.9863,
+      "step": 2595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852721047435805e-05,
+      "loss": 1.623,
+      "step": 2596
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852543773330738e-05,
+      "loss": 1.2783,
+      "step": 2597
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852366393393197e-05,
+      "loss": 1.4551,
+      "step": 2598
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852188907625097e-05,
+      "loss": 0.9775,
+      "step": 2599
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852011316028334e-05,
+      "loss": 1.5625,
+      "step": 2600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985183361860482e-05,
+      "loss": 1.3164,
+      "step": 2601
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851655815356468e-05,
+      "loss": 1.3164,
+      "step": 2602
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985147790628518e-05,
+      "loss": 1.5254,
+      "step": 2603
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985129989139287e-05,
+      "loss": 1.4414,
+      "step": 2604
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851121770681453e-05,
+      "loss": 1.4746,
+      "step": 2605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985094354415284e-05,
+      "loss": 1.0332,
+      "step": 2606
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850765211808944e-05,
+      "loss": 1.4062,
+      "step": 2607
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850586773651685e-05,
+      "loss": 1.2686,
+      "step": 2608
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985040822968298e-05,
+      "loss": 1.6641,
+      "step": 2609
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850229579904737e-05,
+      "loss": 1.2461,
+      "step": 2610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850050824318886e-05,
+      "loss": 1.2266,
+      "step": 2611
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984987196292734e-05,
+      "loss": 1.5098,
+      "step": 2612
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849692995732026e-05,
+      "loss": 0.792,
+      "step": 2613
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849513922734863e-05,
+      "loss": 1.4375,
+      "step": 2614
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849334743937775e-05,
+      "loss": 1.2891,
+      "step": 2615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849155459342687e-05,
+      "loss": 1.3633,
+      "step": 2616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848976068951525e-05,
+      "loss": 1.3418,
+      "step": 2617
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848796572766214e-05,
+      "loss": 1.6406,
+      "step": 2618
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848616970788685e-05,
+      "loss": 1.2363,
+      "step": 2619
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848437263020866e-05,
+      "loss": 1.5195,
+      "step": 2620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984825744946469e-05,
+      "loss": 1.6855,
+      "step": 2621
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848077530122083e-05,
+      "loss": 1.1738,
+      "step": 2622
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984789750499498e-05,
+      "loss": 1.3008,
+      "step": 2623
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847717374085314e-05,
+      "loss": 1.4688,
+      "step": 2624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847537137395023e-05,
+      "loss": 1.0635,
+      "step": 2625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847356794926044e-05,
+      "loss": 0.999,
+      "step": 2626
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984717634668031e-05,
+      "loss": 1.7891,
+      "step": 2627
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984699579265976e-05,
+      "loss": 1.4375,
+      "step": 2628
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984681513286633e-05,
+      "loss": 1.9551,
+      "step": 2629
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984663436730197e-05,
+      "loss": 1.4873,
+      "step": 2630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984645349596861e-05,
+      "loss": 1.5,
+      "step": 2631
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846272518868206e-05,
+      "loss": 1.208,
+      "step": 2632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846091436002693e-05,
+      "loss": 1.5283,
+      "step": 2633
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845910247374017e-05,
+      "loss": 0.8623,
+      "step": 2634
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845728952984126e-05,
+      "loss": 1.2686,
+      "step": 2635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984554755283497e-05,
+      "loss": 1.1992,
+      "step": 2636
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984536604692849e-05,
+      "loss": 1.4805,
+      "step": 2637
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984518443526664e-05,
+      "loss": 1.1338,
+      "step": 2638
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845002717851376e-05,
+      "loss": 1.3662,
+      "step": 2639
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984482089468464e-05,
+      "loss": 1.7539,
+      "step": 2640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844638965768393e-05,
+      "loss": 1.1836,
+      "step": 2641
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844456931104588e-05,
+      "loss": 1.085,
+      "step": 2642
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844274790695178e-05,
+      "loss": 1.6309,
+      "step": 2643
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844092544542117e-05,
+      "loss": 1.4033,
+      "step": 2644
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843910192647367e-05,
+      "loss": 1.377,
+      "step": 2645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843727735012886e-05,
+      "loss": 1.5703,
+      "step": 2646
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984354517164063e-05,
+      "loss": 1.6133,
+      "step": 2647
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984336250253257e-05,
+      "loss": 1.3594,
+      "step": 2648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843179727690656e-05,
+      "loss": 1.1094,
+      "step": 2649
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842996847116856e-05,
+      "loss": 1.2285,
+      "step": 2650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842813860813138e-05,
+      "loss": 1.5391,
+      "step": 2651
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984263076878147e-05,
+      "loss": 1.3848,
+      "step": 2652
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842447571023806e-05,
+      "loss": 1.5137,
+      "step": 2653
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842264267542125e-05,
+      "loss": 1.0659,
+      "step": 2654
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984208085833839e-05,
+      "loss": 1.2949,
+      "step": 2655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9841897343414576e-05,
+      "loss": 1.332,
+      "step": 2656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984171372277265e-05,
+      "loss": 1.3008,
+      "step": 2657
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841529996414586e-05,
+      "loss": 1.125,
+      "step": 2658
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841346164342357e-05,
+      "loss": 1.1035,
+      "step": 2659
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984116222655794e-05,
+      "loss": 1.7227,
+      "step": 2660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984097818306331e-05,
+      "loss": 1.2021,
+      "step": 2661
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840794033860442e-05,
+      "loss": 1.2549,
+      "step": 2662
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840609778951318e-05,
+      "loss": 1.1992,
+      "step": 2663
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984042541833791e-05,
+      "loss": 1.2871,
+      "step": 2664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840240952022203e-05,
+      "loss": 1.6309,
+      "step": 2665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984005638000618e-05,
+      "loss": 1.5957,
+      "step": 2666
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839871702291822e-05,
+      "loss": 1.5527,
+      "step": 2667
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839686918881112e-05,
+      "loss": 1.4336,
+      "step": 2668
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839502029776035e-05,
+      "loss": 1.5703,
+      "step": 2669
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839317034978576e-05,
+      "loss": 1.4141,
+      "step": 2670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839131934490725e-05,
+      "loss": 0.9346,
+      "step": 2671
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983894672831447e-05,
+      "loss": 1.7383,
+      "step": 2672
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838761416451796e-05,
+      "loss": 1.5391,
+      "step": 2673
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98385759989047e-05,
+      "loss": 1.4902,
+      "step": 2674
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983839047567517e-05,
+      "loss": 0.8784,
+      "step": 2675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838204846765197e-05,
+      "loss": 1.1475,
+      "step": 2676
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838019112176775e-05,
+      "loss": 1.5234,
+      "step": 2677
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983783327191191e-05,
+      "loss": 1.2715,
+      "step": 2678
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837647325972583e-05,
+      "loss": 1.334,
+      "step": 2679
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98374612743608e-05,
+      "loss": 1.7148,
+      "step": 2680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837275117078553e-05,
+      "loss": 1.2363,
+      "step": 2681
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837088854127852e-05,
+      "loss": 1.2998,
+      "step": 2682
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836902485510688e-05,
+      "loss": 1.2393,
+      "step": 2683
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836716011229068e-05,
+      "loss": 1.5605,
+      "step": 2684
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836529431284997e-05,
+      "loss": 1.1523,
+      "step": 2685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983634274568047e-05,
+      "loss": 1.4551,
+      "step": 2686
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836155954417503e-05,
+      "loss": 1.1338,
+      "step": 2687
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98359690574981e-05,
+      "loss": 1.748,
+      "step": 2688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983578205492426e-05,
+      "loss": 0.999,
+      "step": 2689
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835594946698e-05,
+      "loss": 1.1465,
+      "step": 2690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835407732821328e-05,
+      "loss": 1.6602,
+      "step": 2691
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835220413296253e-05,
+      "loss": 1.4199,
+      "step": 2692
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835032988124794e-05,
+      "loss": 1.416,
+      "step": 2693
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834845457308956e-05,
+      "loss": 1.1201,
+      "step": 2694
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834657820850756e-05,
+      "loss": 1.4092,
+      "step": 2695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983447007875221e-05,
+      "loss": 1.21,
+      "step": 2696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983428223101534e-05,
+      "loss": 1.2754,
+      "step": 2697
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834094277642154e-05,
+      "loss": 1.0264,
+      "step": 2698
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833906218634677e-05,
+      "loss": 1.4863,
+      "step": 2699
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833718053994924e-05,
+      "loss": 0.9539,
+      "step": 2700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983352978372492e-05,
+      "loss": 1.4355,
+      "step": 2701
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983334140782669e-05,
+      "loss": 1.3066,
+      "step": 2702
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983315292630225e-05,
+      "loss": 0.9844,
+      "step": 2703
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832964339153638e-05,
+      "loss": 1.2715,
+      "step": 2704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832775646382866e-05,
+      "loss": 1.0225,
+      "step": 2705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832586847991964e-05,
+      "loss": 1.5742,
+      "step": 2706
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832397943982965e-05,
+      "loss": 1.3906,
+      "step": 2707
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832208934357892e-05,
+      "loss": 1.5703,
+      "step": 2708
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983201981911878e-05,
+      "loss": 1.6152,
+      "step": 2709
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983183059826766e-05,
+      "loss": 0.957,
+      "step": 2710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831641271806558e-05,
+      "loss": 1.4824,
+      "step": 2711
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831451839737518e-05,
+      "loss": 1.4434,
+      "step": 2712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831262302062564e-05,
+      "loss": 1.3369,
+      "step": 2713
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831072658783745e-05,
+      "loss": 1.4473,
+      "step": 2714
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830882909903084e-05,
+      "loss": 1.4385,
+      "step": 2715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830693055422627e-05,
+      "loss": 1.2949,
+      "step": 2716
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830503095344415e-05,
+      "loss": 1.2559,
+      "step": 2717
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830313029670484e-05,
+      "loss": 1.249,
+      "step": 2718
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983012285840288e-05,
+      "loss": 1.0381,
+      "step": 2719
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982993258154364e-05,
+      "loss": 1.083,
+      "step": 2720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982974219909481e-05,
+      "loss": 1.3984,
+      "step": 2721
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982955171105844e-05,
+      "loss": 1.6377,
+      "step": 2722
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829361117436568e-05,
+      "loss": 1.3926,
+      "step": 2723
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829170418231247e-05,
+      "loss": 1.4707,
+      "step": 2724
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828979613444525e-05,
+      "loss": 1.3027,
+      "step": 2725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982878870307845e-05,
+      "loss": 1.4258,
+      "step": 2726
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828597687135072e-05,
+      "loss": 1.043,
+      "step": 2727
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828406565616443e-05,
+      "loss": 1.5508,
+      "step": 2728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982821533852462e-05,
+      "loss": 1.5645,
+      "step": 2729
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982802400586165e-05,
+      "loss": 1.2402,
+      "step": 2730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827832567629593e-05,
+      "loss": 1.3525,
+      "step": 2731
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827641023830505e-05,
+      "loss": 0.8838,
+      "step": 2732
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827449374466446e-05,
+      "loss": 1.2959,
+      "step": 2733
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982725761953947e-05,
+      "loss": 1.4531,
+      "step": 2734
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827065759051635e-05,
+      "loss": 1.5918,
+      "step": 2735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982687379300501e-05,
+      "loss": 1.6172,
+      "step": 2736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982668172140165e-05,
+      "loss": 1.625,
+      "step": 2737
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826489544243623e-05,
+      "loss": 1.0508,
+      "step": 2738
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826297261532988e-05,
+      "loss": 1.29,
+      "step": 2739
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826104873271813e-05,
+      "loss": 1.1416,
+      "step": 2740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825912379462164e-05,
+      "loss": 1.2109,
+      "step": 2741
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825719780106114e-05,
+      "loss": 1.4883,
+      "step": 2742
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982552707520572e-05,
+      "loss": 1.2754,
+      "step": 2743
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825334264763066e-05,
+      "loss": 1.4746,
+      "step": 2744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825141348780215e-05,
+      "loss": 1.1553,
+      "step": 2745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824948327259238e-05,
+      "loss": 1.252,
+      "step": 2746
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824755200202214e-05,
+      "loss": 1.0459,
+      "step": 2747
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824561967611214e-05,
+      "loss": 1.4443,
+      "step": 2748
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824368629488312e-05,
+      "loss": 1.4268,
+      "step": 2749
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982417518583559e-05,
+      "loss": 1.5098,
+      "step": 2750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982398163665512e-05,
+      "loss": 1.3057,
+      "step": 2751
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823787981948985e-05,
+      "loss": 1.415,
+      "step": 2752
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823594221719264e-05,
+      "loss": 1.1265,
+      "step": 2753
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982340035596804e-05,
+      "loss": 1.3252,
+      "step": 2754
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982320638469739e-05,
+      "loss": 1.5293,
+      "step": 2755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823012307909405e-05,
+      "loss": 1.4971,
+      "step": 2756
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822818125606162e-05,
+      "loss": 1.0986,
+      "step": 2757
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822623837789753e-05,
+      "loss": 1.6367,
+      "step": 2758
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822429444462265e-05,
+      "loss": 1.2295,
+      "step": 2759
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982223494562578e-05,
+      "loss": 1.2803,
+      "step": 2760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822040341282392e-05,
+      "loss": 1.5918,
+      "step": 2761
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821845631434192e-05,
+      "loss": 1.752,
+      "step": 2762
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982165081608327e-05,
+      "loss": 1.1494,
+      "step": 2763
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982145589523172e-05,
+      "loss": 0.9854,
+      "step": 2764
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821260868881634e-05,
+      "loss": 1.2241,
+      "step": 2765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821065737035104e-05,
+      "loss": 1.6133,
+      "step": 2766
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820870499694233e-05,
+      "loss": 1.2656,
+      "step": 2767
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820675156861113e-05,
+      "loss": 1.1104,
+      "step": 2768
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820479708537843e-05,
+      "loss": 1.4727,
+      "step": 2769
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820284154726526e-05,
+      "loss": 1.2383,
+      "step": 2770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820088495429256e-05,
+      "loss": 1.5078,
+      "step": 2771
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981989273064814e-05,
+      "loss": 1.4131,
+      "step": 2772
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981969686038528e-05,
+      "loss": 1.5312,
+      "step": 2773
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981950088464278e-05,
+      "loss": 1.6367,
+      "step": 2774
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819304803422744e-05,
+      "loss": 1.1367,
+      "step": 2775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981910861672728e-05,
+      "loss": 1.2021,
+      "step": 2776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981891232455849e-05,
+      "loss": 1.1045,
+      "step": 2777
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981871592691849e-05,
+      "loss": 1.3564,
+      "step": 2778
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818519423809386e-05,
+      "loss": 1.5293,
+      "step": 2779
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818322815233287e-05,
+      "loss": 1.0537,
+      "step": 2780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981812610119231e-05,
+      "loss": 1.2305,
+      "step": 2781
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981792928168856e-05,
+      "loss": 1.6152,
+      "step": 2782
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981773235672416e-05,
+      "loss": 1.3643,
+      "step": 2783
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817535326301222e-05,
+      "loss": 1.0996,
+      "step": 2784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981733819042186e-05,
+      "loss": 1.2861,
+      "step": 2785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817140949088193e-05,
+      "loss": 1.2471,
+      "step": 2786
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816943602302338e-05,
+      "loss": 1.5801,
+      "step": 2787
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981674615006642e-05,
+      "loss": 1.1035,
+      "step": 2788
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816548592382555e-05,
+      "loss": 1.3496,
+      "step": 2789
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981635092925287e-05,
+      "loss": 1.2588,
+      "step": 2790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816153160679487e-05,
+      "loss": 1.3262,
+      "step": 2791
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981595528666452e-05,
+      "loss": 0.9941,
+      "step": 2792
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981575730721011e-05,
+      "loss": 1.4756,
+      "step": 2793
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815559222318374e-05,
+      "loss": 1.0264,
+      "step": 2794
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815361031991442e-05,
+      "loss": 1.1445,
+      "step": 2795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815162736231445e-05,
+      "loss": 1.2188,
+      "step": 2796
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981496433504051e-05,
+      "loss": 1.4629,
+      "step": 2797
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814765828420767e-05,
+      "loss": 1.1338,
+      "step": 2798
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814567216374354e-05,
+      "loss": 1.3887,
+      "step": 2799
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814368498903402e-05,
+      "loss": 1.0898,
+      "step": 2800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814169676010042e-05,
+      "loss": 1.5195,
+      "step": 2801
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813970747696415e-05,
+      "loss": 1.6562,
+      "step": 2802
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813771713964655e-05,
+      "loss": 1.2305,
+      "step": 2803
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813572574816897e-05,
+      "loss": 1.0049,
+      "step": 2804
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813373330255285e-05,
+      "loss": 1.3691,
+      "step": 2805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813173980281958e-05,
+      "loss": 1.5957,
+      "step": 2806
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812974524899057e-05,
+      "loss": 1.3779,
+      "step": 2807
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812774964108723e-05,
+      "loss": 0.8271,
+      "step": 2808
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98125752979131e-05,
+      "loss": 1.3242,
+      "step": 2809
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981237552631434e-05,
+      "loss": 1.0781,
+      "step": 2810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812175649314574e-05,
+      "loss": 1.3945,
+      "step": 2811
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811975666915963e-05,
+      "loss": 1.126,
+      "step": 2812
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811775579120645e-05,
+      "loss": 1.6055,
+      "step": 2813
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811575385930776e-05,
+      "loss": 2.0254,
+      "step": 2814
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811375087348507e-05,
+      "loss": 1.2432,
+      "step": 2815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811174683375983e-05,
+      "loss": 1.498,
+      "step": 2816
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810974174015363e-05,
+      "loss": 1.6582,
+      "step": 2817
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810773559268797e-05,
+      "loss": 1.2373,
+      "step": 2818
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981057283913844e-05,
+      "loss": 1.1758,
+      "step": 2819
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810372013626452e-05,
+      "loss": 1.21,
+      "step": 2820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981017108273499e-05,
+      "loss": 1.1953,
+      "step": 2821
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809970046466204e-05,
+      "loss": 1.0283,
+      "step": 2822
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809768904822262e-05,
+      "loss": 1.499,
+      "step": 2823
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980956765780532e-05,
+      "loss": 1.2383,
+      "step": 2824
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809366305417544e-05,
+      "loss": 0.9912,
+      "step": 2825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980916484766109e-05,
+      "loss": 1.3779,
+      "step": 2826
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808963284538135e-05,
+      "loss": 1.8672,
+      "step": 2827
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808761616050833e-05,
+      "loss": 1.6016,
+      "step": 2828
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808559842201346e-05,
+      "loss": 1.2979,
+      "step": 2829
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808357962991853e-05,
+      "loss": 1.1973,
+      "step": 2830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980815597842452e-05,
+      "loss": 1.5762,
+      "step": 2831
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980795388850151e-05,
+      "loss": 1.1699,
+      "step": 2832
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807751693224998e-05,
+      "loss": 1.7734,
+      "step": 2833
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980754939259716e-05,
+      "loss": 1.0957,
+      "step": 2834
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980734698662016e-05,
+      "loss": 1.5928,
+      "step": 2835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807144475296183e-05,
+      "loss": 1.2197,
+      "step": 2836
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806941858627392e-05,
+      "loss": 1.0273,
+      "step": 2837
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806739136615977e-05,
+      "loss": 1.1621,
+      "step": 2838
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806536309264105e-05,
+      "loss": 1.4092,
+      "step": 2839
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806333376573956e-05,
+      "loss": 1.0691,
+      "step": 2840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806130338547714e-05,
+      "loss": 1.4209,
+      "step": 2841
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980592719518756e-05,
+      "loss": 1.1055,
+      "step": 2842
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805723946495675e-05,
+      "loss": 1.5684,
+      "step": 2843
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805520592474237e-05,
+      "loss": 1.3477,
+      "step": 2844
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805317133125436e-05,
+      "loss": 1.4414,
+      "step": 2845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805113568451458e-05,
+      "loss": 1.7207,
+      "step": 2846
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804909898454487e-05,
+      "loss": 1.1758,
+      "step": 2847
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980470612313671e-05,
+      "loss": 1.3809,
+      "step": 2848
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804502242500317e-05,
+      "loss": 1.252,
+      "step": 2849
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98042982565475e-05,
+      "loss": 1.1528,
+      "step": 2850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980409416528045e-05,
+      "loss": 1.3164,
+      "step": 2851
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803889968701358e-05,
+      "loss": 1.085,
+      "step": 2852
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803685666812414e-05,
+      "loss": 1.2822,
+      "step": 2853
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980348125961582e-05,
+      "loss": 1.4707,
+      "step": 2854
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803276747113764e-05,
+      "loss": 1.21,
+      "step": 2855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980307212930845e-05,
+      "loss": 1.2949,
+      "step": 2856
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802867406202066e-05,
+      "loss": 1.1328,
+      "step": 2857
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802662577796822e-05,
+      "loss": 1.457,
+      "step": 2858
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802457644094913e-05,
+      "loss": 1.3848,
+      "step": 2859
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980225260509854e-05,
+      "loss": 1.5156,
+      "step": 2860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802047460809904e-05,
+      "loss": 0.96,
+      "step": 2861
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980184221123122e-05,
+      "loss": 1.5449,
+      "step": 2862
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801636856364674e-05,
+      "loss": 1.5273,
+      "step": 2863
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801431396212486e-05,
+      "loss": 1.1816,
+      "step": 2864
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801225830776855e-05,
+      "loss": 1.1377,
+      "step": 2865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801020160059995e-05,
+      "loss": 1.3408,
+      "step": 2866
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800814384064112e-05,
+      "loss": 1.25,
+      "step": 2867
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800608502791415e-05,
+      "loss": 1.1816,
+      "step": 2868
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800402516244118e-05,
+      "loss": 1.5156,
+      "step": 2869
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800196424424435e-05,
+      "loss": 1.29,
+      "step": 2870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979999022733458e-05,
+      "loss": 1.3096,
+      "step": 2871
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799783924976763e-05,
+      "loss": 1.2002,
+      "step": 2872
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799577517353202e-05,
+      "loss": 1.2451,
+      "step": 2873
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979937100446612e-05,
+      "loss": 1.4453,
+      "step": 2874
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799164386317727e-05,
+      "loss": 1.7637,
+      "step": 2875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798957662910245e-05,
+      "loss": 1.3389,
+      "step": 2876
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798750834245895e-05,
+      "loss": 1.2783,
+      "step": 2877
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97985439003269e-05,
+      "loss": 1.4863,
+      "step": 2878
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798336861155482e-05,
+      "loss": 1.3164,
+      "step": 2879
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798129716733868e-05,
+      "loss": 1.5469,
+      "step": 2880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797922467064273e-05,
+      "loss": 1.0869,
+      "step": 2881
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797715112148937e-05,
+      "loss": 1.2363,
+      "step": 2882
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797507651990076e-05,
+      "loss": 0.9219,
+      "step": 2883
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797300086589923e-05,
+      "loss": 1.4307,
+      "step": 2884
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979709241595071e-05,
+      "loss": 1.7109,
+      "step": 2885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796884640074663e-05,
+      "loss": 1.2285,
+      "step": 2886
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796676758964015e-05,
+      "loss": 1.2725,
+      "step": 2887
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796468772621003e-05,
+      "loss": 0.9453,
+      "step": 2888
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796260681047855e-05,
+      "loss": 1.6133,
+      "step": 2889
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796052484246812e-05,
+      "loss": 1.0283,
+      "step": 2890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795844182220106e-05,
+      "loss": 1.1543,
+      "step": 2891
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795635774969976e-05,
+      "loss": 1.2051,
+      "step": 2892
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795427262498662e-05,
+      "loss": 1.3477,
+      "step": 2893
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795218644808404e-05,
+      "loss": 1.7051,
+      "step": 2894
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795009921901438e-05,
+      "loss": 1.3125,
+      "step": 2895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979480109378001e-05,
+      "loss": 1.4805,
+      "step": 2896
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794592160446363e-05,
+      "loss": 1.4531,
+      "step": 2897
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794383121902744e-05,
+      "loss": 1.1533,
+      "step": 2898
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794173978151395e-05,
+      "loss": 1.5664,
+      "step": 2899
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793964729194563e-05,
+      "loss": 1.3262,
+      "step": 2900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979375537503449e-05,
+      "loss": 1.5723,
+      "step": 2901
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793545915673437e-05,
+      "loss": 1.4629,
+      "step": 2902
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793336351113643e-05,
+      "loss": 1.3232,
+      "step": 2903
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793126681357365e-05,
+      "loss": 1.6328,
+      "step": 2904
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792916906406855e-05,
+      "loss": 1.1899,
+      "step": 2905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792707026264364e-05,
+      "loss": 1.4629,
+      "step": 2906
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792497040932145e-05,
+      "loss": 1.1123,
+      "step": 2907
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979228695041246e-05,
+      "loss": 1.3691,
+      "step": 2908
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979207675470756e-05,
+      "loss": 1.2861,
+      "step": 2909
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791866453819704e-05,
+      "loss": 1.3379,
+      "step": 2910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791656047751152e-05,
+      "loss": 1.1719,
+      "step": 2911
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791445536504165e-05,
+      "loss": 1.2793,
+      "step": 2912
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791234920081e-05,
+      "loss": 1.1045,
+      "step": 2913
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791024198483922e-05,
+      "loss": 1.6621,
+      "step": 2914
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790813371715197e-05,
+      "loss": 1.5176,
+      "step": 2915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979060243977709e-05,
+      "loss": 1.2949,
+      "step": 2916
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790391402671858e-05,
+      "loss": 1.6816,
+      "step": 2917
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790180260401778e-05,
+      "loss": 1.459,
+      "step": 2918
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978996901296911e-05,
+      "loss": 1.0234,
+      "step": 2919
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789757660376133e-05,
+      "loss": 1.2559,
+      "step": 2920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789546202625107e-05,
+      "loss": 1.1367,
+      "step": 2921
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978933463971831e-05,
+      "loss": 1.3096,
+      "step": 2922
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789122971658013e-05,
+      "loss": 0.9902,
+      "step": 2923
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788911198446487e-05,
+      "loss": 1.1445,
+      "step": 2924
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788699320086012e-05,
+      "loss": 0.8809,
+      "step": 2925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788487336578858e-05,
+      "loss": 1.0488,
+      "step": 2926
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788275247927306e-05,
+      "loss": 1.2549,
+      "step": 2927
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978806305413363e-05,
+      "loss": 1.0381,
+      "step": 2928
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787850755200116e-05,
+      "loss": 1.3809,
+      "step": 2929
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978763835112904e-05,
+      "loss": 1.3877,
+      "step": 2930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978742584192268e-05,
+      "loss": 1.6191,
+      "step": 2931
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787213227583328e-05,
+      "loss": 1.3418,
+      "step": 2932
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787000508113265e-05,
+      "loss": 1.1865,
+      "step": 2933
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786787683514768e-05,
+      "loss": 1.1855,
+      "step": 2934
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786574753790127e-05,
+      "loss": 1.3975,
+      "step": 2935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786361718941637e-05,
+      "loss": 1.6621,
+      "step": 2936
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786148578971578e-05,
+      "loss": 1.3242,
+      "step": 2937
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785935333882243e-05,
+      "loss": 0.9844,
+      "step": 2938
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978572198367592e-05,
+      "loss": 1.1826,
+      "step": 2939
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785508528354908e-05,
+      "loss": 1.5107,
+      "step": 2940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785294967921485e-05,
+      "loss": 1.543,
+      "step": 2941
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978508130237796e-05,
+      "loss": 1.1738,
+      "step": 2942
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978486753172662e-05,
+      "loss": 1.3721,
+      "step": 2943
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784653655969764e-05,
+      "loss": 1.6387,
+      "step": 2944
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784439675109687e-05,
+      "loss": 1.0073,
+      "step": 2945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784225589148692e-05,
+      "loss": 1.0791,
+      "step": 2946
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784011398089076e-05,
+      "loss": 1.3691,
+      "step": 2947
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783797101933138e-05,
+      "loss": 0.8584,
+      "step": 2948
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783582700683183e-05,
+      "loss": 1.4805,
+      "step": 2949
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978336819434151e-05,
+      "loss": 1.3867,
+      "step": 2950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978315358291043e-05,
+      "loss": 1.0518,
+      "step": 2951
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978293886639224e-05,
+      "loss": 1.5352,
+      "step": 2952
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978272404478925e-05,
+      "loss": 1.1973,
+      "step": 2953
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782509118103773e-05,
+      "loss": 1.458,
+      "step": 2954
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978229408633811e-05,
+      "loss": 1.1074,
+      "step": 2955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978207894949457e-05,
+      "loss": 1.2324,
+      "step": 2956
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781863707575474e-05,
+      "loss": 1.5605,
+      "step": 2957
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781648360583127e-05,
+      "loss": 1.1826,
+      "step": 2958
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978143290851984e-05,
+      "loss": 1.1826,
+      "step": 2959
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781217351387928e-05,
+      "loss": 1.7559,
+      "step": 2960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781001689189708e-05,
+      "loss": 1.1006,
+      "step": 2961
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97807859219275e-05,
+      "loss": 1.1748,
+      "step": 2962
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780570049603618e-05,
+      "loss": 1.0957,
+      "step": 2963
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978035407222038e-05,
+      "loss": 1.6816,
+      "step": 2964
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780137989780108e-05,
+      "loss": 0.9106,
+      "step": 2965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779921802285123e-05,
+      "loss": 1.4189,
+      "step": 2966
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779705509737746e-05,
+      "loss": 1.54,
+      "step": 2967
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97794891121403e-05,
+      "loss": 1.6289,
+      "step": 2968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9779272609495113e-05,
+      "loss": 1.1094,
+      "step": 2969
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977905600180451e-05,
+      "loss": 1.3516,
+      "step": 2970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977883928907081e-05,
+      "loss": 1.4424,
+      "step": 2971
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778622471296353e-05,
+      "loss": 1.6025,
+      "step": 2972
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778405548483456e-05,
+      "loss": 1.415,
+      "step": 2973
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977818852063446e-05,
+      "loss": 1.5488,
+      "step": 2974
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777971387751683e-05,
+      "loss": 1.5195,
+      "step": 2975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777754149837472e-05,
+      "loss": 1.8418,
+      "step": 2976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977753680689415e-05,
+      "loss": 1.042,
+      "step": 2977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977731935892406e-05,
+      "loss": 1.1855,
+      "step": 2978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977710180592953e-05,
+      "loss": 1.3516,
+      "step": 2979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97768841479129e-05,
+      "loss": 1.4561,
+      "step": 2980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977666638487651e-05,
+      "loss": 1.2256,
+      "step": 2981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776448516822696e-05,
+      "loss": 1.2275,
+      "step": 2982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776230543753797e-05,
+      "loss": 1.2852,
+      "step": 2983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977601246567216e-05,
+      "loss": 1.2461,
+      "step": 2984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775794282580123e-05,
+      "loss": 1.1553,
+      "step": 2985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775575994480034e-05,
+      "loss": 1.3184,
+      "step": 2986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977535760137423e-05,
+      "loss": 1.3018,
+      "step": 2987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775139103265063e-05,
+      "loss": 1.2598,
+      "step": 2988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977492050015488e-05,
+      "loss": 1.1338,
+      "step": 2989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977470179204603e-05,
+      "loss": 1.1162,
+      "step": 2990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977448297894086e-05,
+      "loss": 1.5605,
+      "step": 2991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774264060841715e-05,
+      "loss": 1.3643,
+      "step": 2992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774045037750955e-05,
+      "loss": 1.0918,
+      "step": 2993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977382590967093e-05,
+      "loss": 1.5664,
+      "step": 2994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773606676603995e-05,
+      "loss": 1.2524,
+      "step": 2995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97733873385525e-05,
+      "loss": 1.3164,
+      "step": 2996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977316789551881e-05,
+      "loss": 1.4014,
+      "step": 2997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977294834750527e-05,
+      "loss": 1.7012,
+      "step": 2998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977272869451425e-05,
+      "loss": 1.5977,
+      "step": 2999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772508936548103e-05,
+      "loss": 0.8848,
+      "step": 3000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772289073609193e-05,
+      "loss": 1.4834,
+      "step": 3001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772069105699878e-05,
+      "loss": 1.2393,
+      "step": 3002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977184903282252e-05,
+      "loss": 1.293,
+      "step": 3003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771628854979487e-05,
+      "loss": 1.3408,
+      "step": 3004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771408572173145e-05,
+      "loss": 1.2852,
+      "step": 3005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771188184405858e-05,
+      "loss": 1.4121,
+      "step": 3006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770967691679987e-05,
+      "loss": 1.3359,
+      "step": 3007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977074709399791e-05,
+      "loss": 1.4365,
+      "step": 3008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770526391361996e-05,
+      "loss": 1.582,
+      "step": 3009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977030558377461e-05,
+      "loss": 1.4121,
+      "step": 3010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977008467123813e-05,
+      "loss": 1.2812,
+      "step": 3011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976986365375492e-05,
+      "loss": 1.2549,
+      "step": 3012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976964253132736e-05,
+      "loss": 1.0615,
+      "step": 3013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976942130395783e-05,
+      "loss": 1.2041,
+      "step": 3014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769199971648696e-05,
+      "loss": 1.7578,
+      "step": 3015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768978534402343e-05,
+      "loss": 1.5625,
+      "step": 3016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768756992221146e-05,
+      "loss": 1.4551,
+      "step": 3017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976853534510749e-05,
+      "loss": 1.4268,
+      "step": 3018
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976831359306375e-05,
+      "loss": 1.7129,
+      "step": 3019
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976809173609231e-05,
+      "loss": 1.333,
+      "step": 3020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976786977419555e-05,
+      "loss": 1.1055,
+      "step": 3021
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767647707375858e-05,
+      "loss": 1.7305,
+      "step": 3022
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976742553563562e-05,
+      "loss": 1.3594,
+      "step": 3023
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976720325897722e-05,
+      "loss": 0.9531,
+      "step": 3024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766980877403047e-05,
+      "loss": 1.2725,
+      "step": 3025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766758390915494e-05,
+      "loss": 1.5254,
+      "step": 3026
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976653579951694e-05,
+      "loss": 1.4648,
+      "step": 3027
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766313103209785e-05,
+      "loss": 1.126,
+      "step": 3028
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976609030199642e-05,
+      "loss": 1.125,
+      "step": 3029
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976586739587923e-05,
+      "loss": 1.0459,
+      "step": 3030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765644384860624e-05,
+      "loss": 1.0469,
+      "step": 3031
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765421268942986e-05,
+      "loss": 1.3086,
+      "step": 3032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765198048128717e-05,
+      "loss": 1.5312,
+      "step": 3033
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764974722420215e-05,
+      "loss": 1.3555,
+      "step": 3034
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764751291819872e-05,
+      "loss": 1.2549,
+      "step": 3035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764527756330102e-05,
+      "loss": 1.7852,
+      "step": 3036
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764304115953292e-05,
+      "loss": 1.1162,
+      "step": 3037
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764080370691855e-05,
+      "loss": 0.9248,
+      "step": 3038
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763856520548182e-05,
+      "loss": 1.4395,
+      "step": 3039
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763632565524694e-05,
+      "loss": 1.291,
+      "step": 3040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976340850562378e-05,
+      "loss": 1.1592,
+      "step": 3041
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976318434084786e-05,
+      "loss": 1.6289,
+      "step": 3042
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762960071199334e-05,
+      "loss": 1.1777,
+      "step": 3043
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762735696680617e-05,
+      "loss": 1.3076,
+      "step": 3044
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976251121729411e-05,
+      "loss": 1.2393,
+      "step": 3045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762286633042237e-05,
+      "loss": 1.2559,
+      "step": 3046
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97620619439274e-05,
+      "loss": 1.2705,
+      "step": 3047
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761837149952014e-05,
+      "loss": 1.4941,
+      "step": 3048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761612251118493e-05,
+      "loss": 1.5469,
+      "step": 3049
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976138724742926e-05,
+      "loss": 1.335,
+      "step": 3050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761162138886724e-05,
+      "loss": 1.1846,
+      "step": 3051
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760936925493312e-05,
+      "loss": 1.3066,
+      "step": 3052
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976071160725143e-05,
+      "loss": 1.166,
+      "step": 3053
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976048618416351e-05,
+      "loss": 1.1533,
+      "step": 3054
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760260656231968e-05,
+      "loss": 1.0518,
+      "step": 3055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760035023459228e-05,
+      "loss": 1.498,
+      "step": 3056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759809285847712e-05,
+      "loss": 1.6084,
+      "step": 3057
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759583443399847e-05,
+      "loss": 1.2549,
+      "step": 3058
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759357496118058e-05,
+      "loss": 1.3057,
+      "step": 3059
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759131444004774e-05,
+      "loss": 1.4922,
+      "step": 3060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975890528706242e-05,
+      "loss": 1.1992,
+      "step": 3061
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758679025293427e-05,
+      "loss": 1.4512,
+      "step": 3062
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758452658700222e-05,
+      "loss": 1.1914,
+      "step": 3063
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975822618728524e-05,
+      "loss": 1.1992,
+      "step": 3064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757999611050917e-05,
+      "loss": 1.3174,
+      "step": 3065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757772929999678e-05,
+      "loss": 1.1279,
+      "step": 3066
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757546144133967e-05,
+      "loss": 1.4961,
+      "step": 3067
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757319253456212e-05,
+      "loss": 1.5576,
+      "step": 3068
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757092257968857e-05,
+      "loss": 0.9365,
+      "step": 3069
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756865157674338e-05,
+      "loss": 1.0649,
+      "step": 3070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756637952575093e-05,
+      "loss": 1.5566,
+      "step": 3071
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975641064267356e-05,
+      "loss": 1.1875,
+      "step": 3072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756183227972188e-05,
+      "loss": 1.582,
+      "step": 3073
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755955708473416e-05,
+      "loss": 1.043,
+      "step": 3074
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755728084179687e-05,
+      "loss": 0.9463,
+      "step": 3075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755500355093448e-05,
+      "loss": 1.3691,
+      "step": 3076
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975527252121714e-05,
+      "loss": 1.4297,
+      "step": 3077
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755044582553218e-05,
+      "loss": 1.2842,
+      "step": 3078
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754816539104124e-05,
+      "loss": 1.0479,
+      "step": 3079
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975458839087231e-05,
+      "loss": 1.916,
+      "step": 3080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754360137860232e-05,
+      "loss": 1.6406,
+      "step": 3081
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754131780070333e-05,
+      "loss": 1.3379,
+      "step": 3082
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975390331750507e-05,
+      "loss": 1.502,
+      "step": 3083
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97536747501669e-05,
+      "loss": 1.2949,
+      "step": 3084
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753446078058273e-05,
+      "loss": 1.375,
+      "step": 3085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753217301181645e-05,
+      "loss": 1.416,
+      "step": 3086
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752988419539477e-05,
+      "loss": 1.457,
+      "step": 3087
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975275943313423e-05,
+      "loss": 1.3701,
+      "step": 3088
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752530341968356e-05,
+      "loss": 1.6543,
+      "step": 3089
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752301146044323e-05,
+      "loss": 1.3301,
+      "step": 3090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752071845364585e-05,
+      "loss": 1.5225,
+      "step": 3091
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751842439931614e-05,
+      "loss": 1.2998,
+      "step": 3092
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751612929747868e-05,
+      "loss": 1.4365,
+      "step": 3093
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751383314815816e-05,
+      "loss": 1.4062,
+      "step": 3094
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751153595137922e-05,
+      "loss": 1.5156,
+      "step": 3095
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750923770716655e-05,
+      "loss": 1.3145,
+      "step": 3096
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750693841554482e-05,
+      "loss": 1.1221,
+      "step": 3097
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750463807653873e-05,
+      "loss": 0.9414,
+      "step": 3098
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97502336690173e-05,
+      "loss": 1.5332,
+      "step": 3099
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975000342564724e-05,
+      "loss": 1.3604,
+      "step": 3100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749773077546156e-05,
+      "loss": 1.2949,
+      "step": 3101
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974954262471653e-05,
+      "loss": 1.0938,
+      "step": 3102
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974931206716083e-05,
+      "loss": 1.4434,
+      "step": 3103
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749081404881544e-05,
+      "loss": 1.3086,
+      "step": 3104
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748850637881136e-05,
+      "loss": 1.0947,
+      "step": 3105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748619766162095e-05,
+      "loss": 1.4707,
+      "step": 3106
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97483887897269e-05,
+      "loss": 1.1709,
+      "step": 3107
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974815770857803e-05,
+      "loss": 0.9077,
+      "step": 3108
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747926522717966e-05,
+      "loss": 1.0469,
+      "step": 3109
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747695232149193e-05,
+      "loss": 1.4561,
+      "step": 3110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747463836874192e-05,
+      "loss": 1.5254,
+      "step": 3111
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747232336895453e-05,
+      "loss": 1.0195,
+      "step": 3112
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747000732215462e-05,
+      "loss": 1.4375,
+      "step": 3113
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746769022836707e-05,
+      "loss": 1.2666,
+      "step": 3114
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746537208761674e-05,
+      "loss": 1.7305,
+      "step": 3115
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746305289992854e-05,
+      "loss": 1.1904,
+      "step": 3116
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974607326653274e-05,
+      "loss": 1.0781,
+      "step": 3117
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745841138383827e-05,
+      "loss": 1.1836,
+      "step": 3118
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97456089055486e-05,
+      "loss": 1.4863,
+      "step": 3119
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974537656802956e-05,
+      "loss": 1.3086,
+      "step": 3120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97451441258292e-05,
+      "loss": 1.3242,
+      "step": 3121
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744911578950023e-05,
+      "loss": 1.0693,
+      "step": 3122
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744678927394514e-05,
+      "loss": 1.293,
+      "step": 3123
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744446171165185e-05,
+      "loss": 1.2432,
+      "step": 3124
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744213310264532e-05,
+      "loss": 1.6328,
+      "step": 3125
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974398034469505e-05,
+      "loss": 1.6426,
+      "step": 3126
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974374727445925e-05,
+      "loss": 0.918,
+      "step": 3127
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743514099559636e-05,
+      "loss": 1.3896,
+      "step": 3128
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743280819998705e-05,
+      "loss": 1.1699,
+      "step": 3129
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974304743577897e-05,
+      "loss": 1.2061,
+      "step": 3130
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742813946902932e-05,
+      "loss": 1.2559,
+      "step": 3131
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97425803533731e-05,
+      "loss": 1.2129,
+      "step": 3132
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742346655191987e-05,
+      "loss": 1.0801,
+      "step": 3133
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742112852362104e-05,
+      "loss": 1.3828,
+      "step": 3134
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741878944885955e-05,
+      "loss": 1.0381,
+      "step": 3135
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974164493276606e-05,
+      "loss": 1.1016,
+      "step": 3136
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741410816004932e-05,
+      "loss": 1.3076,
+      "step": 3137
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741176594605082e-05,
+      "loss": 1.207,
+      "step": 3138
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974094226856903e-05,
+      "loss": 1.3096,
+      "step": 3139
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740707837899284e-05,
+      "loss": 1.3184,
+      "step": 3140
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740473302598375e-05,
+      "loss": 1.3032,
+      "step": 3141
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740238662668815e-05,
+      "loss": 1.3828,
+      "step": 3142
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740003918113122e-05,
+      "loss": 1.3184,
+      "step": 3143
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739769068933824e-05,
+      "loss": 1.5547,
+      "step": 3144
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739534115133444e-05,
+      "loss": 1.3398,
+      "step": 3145
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739299056714497e-05,
+      "loss": 1.3125,
+      "step": 3146
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739063893679517e-05,
+      "loss": 1.3438,
+      "step": 3147
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738828626031024e-05,
+      "loss": 1.0684,
+      "step": 3148
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973859325377155e-05,
+      "loss": 1.1904,
+      "step": 3149
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738357776903623e-05,
+      "loss": 1.2295,
+      "step": 3150
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738122195429766e-05,
+      "loss": 1.1045,
+      "step": 3151
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737886509352517e-05,
+      "loss": 1.4316,
+      "step": 3152
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737650718674403e-05,
+      "loss": 1.2754,
+      "step": 3153
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973741482339796e-05,
+      "loss": 1.7793,
+      "step": 3154
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737178823525717e-05,
+      "loss": 1.1553,
+      "step": 3155
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736942719060217e-05,
+      "loss": 1.2734,
+      "step": 3156
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973670651000399e-05,
+      "loss": 1.1406,
+      "step": 3157
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736470196359572e-05,
+      "loss": 1.8125,
+      "step": 3158
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736233778129507e-05,
+      "loss": 1.1377,
+      "step": 3159
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973599725531633e-05,
+      "loss": 1.3906,
+      "step": 3160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735760627922585e-05,
+      "loss": 1.0732,
+      "step": 3161
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973552389595081e-05,
+      "loss": 1.2139,
+      "step": 3162
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973528705940355e-05,
+      "loss": 1.334,
+      "step": 3163
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735050118283354e-05,
+      "loss": 1.5059,
+      "step": 3164
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734813072592755e-05,
+      "loss": 1.6484,
+      "step": 3165
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734575922334312e-05,
+      "loss": 1.4336,
+      "step": 3166
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734338667510564e-05,
+      "loss": 0.8945,
+      "step": 3167
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973410130812406e-05,
+      "loss": 1.1689,
+      "step": 3168
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973386384417736e-05,
+      "loss": 1.1465,
+      "step": 3169
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733626275673e-05,
+      "loss": 1.6074,
+      "step": 3170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733388602613537e-05,
+      "loss": 1.4727,
+      "step": 3171
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733150825001528e-05,
+      "loss": 1.8691,
+      "step": 3172
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732912942839528e-05,
+      "loss": 1.4512,
+      "step": 3173
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732674956130083e-05,
+      "loss": 1.5811,
+      "step": 3174
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973243686487576e-05,
+      "loss": 1.4141,
+      "step": 3175
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973219866907911e-05,
+      "loss": 1.4648,
+      "step": 3176
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731960368742692e-05,
+      "loss": 1.1152,
+      "step": 3177
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731721963869067e-05,
+      "loss": 1.4043,
+      "step": 3178
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731483454460795e-05,
+      "loss": 1.1289,
+      "step": 3179
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731244840520442e-05,
+      "loss": 1.4287,
+      "step": 3180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731006122050567e-05,
+      "loss": 1.2266,
+      "step": 3181
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730767299053738e-05,
+      "loss": 1.2773,
+      "step": 3182
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730528371532513e-05,
+      "loss": 1.4248,
+      "step": 3183
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730289339489467e-05,
+      "loss": 1.291,
+      "step": 3184
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973005020292716e-05,
+      "loss": 1.208,
+      "step": 3185
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729810961848164e-05,
+      "loss": 0.915,
+      "step": 3186
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729571616255052e-05,
+      "loss": 1.2832,
+      "step": 3187
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729332166150393e-05,
+      "loss": 1.5371,
+      "step": 3188
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729092611536756e-05,
+      "loss": 0.8516,
+      "step": 3189
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972885295241672e-05,
+      "loss": 1.3525,
+      "step": 3190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728613188792854e-05,
+      "loss": 1.377,
+      "step": 3191
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972837332066773e-05,
+      "loss": 1.3926,
+      "step": 3192
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728133348043936e-05,
+      "loss": 0.8091,
+      "step": 3193
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727893270924043e-05,
+      "loss": 1.4434,
+      "step": 3194
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972765308931063e-05,
+      "loss": 1.2168,
+      "step": 3195
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727412803206273e-05,
+      "loss": 0.9619,
+      "step": 3196
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727172412613564e-05,
+      "loss": 1.2666,
+      "step": 3197
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726931917535075e-05,
+      "loss": 1.5801,
+      "step": 3198
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726691317973394e-05,
+      "loss": 1.2227,
+      "step": 3199
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726450613931105e-05,
+      "loss": 1.542,
+      "step": 3200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972620980541079e-05,
+      "loss": 1.6641,
+      "step": 3201
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972596889241504e-05,
+      "loss": 1.502,
+      "step": 3202
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725727874946444e-05,
+      "loss": 1.5273,
+      "step": 3203
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725486753007582e-05,
+      "loss": 1.416,
+      "step": 3204
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972524552660106e-05,
+      "loss": 1.2959,
+      "step": 3205
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725004195729452e-05,
+      "loss": 1.3398,
+      "step": 3206
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972476276039536e-05,
+      "loss": 1.2803,
+      "step": 3207
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724521220601374e-05,
+      "loss": 1.4043,
+      "step": 3208
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972427957635009e-05,
+      "loss": 1.3105,
+      "step": 3209
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724037827644103e-05,
+      "loss": 1.3213,
+      "step": 3210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972379597448601e-05,
+      "loss": 0.9531,
+      "step": 3211
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972355401687841e-05,
+      "loss": 0.7251,
+      "step": 3212
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723311954823905e-05,
+      "loss": 0.9268,
+      "step": 3213
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723069788325086e-05,
+      "loss": 1.0791,
+      "step": 3214
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972282751738456e-05,
+      "loss": 1.3301,
+      "step": 3215
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972258514200493e-05,
+      "loss": 1.1582,
+      "step": 3216
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722342662188797e-05,
+      "loss": 1.4697,
+      "step": 3217
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972210007793877e-05,
+      "loss": 1.3057,
+      "step": 3218
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972185738925745e-05,
+      "loss": 1.582,
+      "step": 3219
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721614596147446e-05,
+      "loss": 1.748,
+      "step": 3220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721371698611363e-05,
+      "loss": 0.9746,
+      "step": 3221
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972112869665182e-05,
+      "loss": 1.4678,
+      "step": 3222
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720885590271415e-05,
+      "loss": 1.2793,
+      "step": 3223
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720642379472763e-05,
+      "loss": 1.4805,
+      "step": 3224
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972039906425848e-05,
+      "loss": 1.2656,
+      "step": 3225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720155644631177e-05,
+      "loss": 1.2871,
+      "step": 3226
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719912120593467e-05,
+      "loss": 1.3086,
+      "step": 3227
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971966849214797e-05,
+      "loss": 1.4385,
+      "step": 3228
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719424759297304e-05,
+      "loss": 1.0195,
+      "step": 3229
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971918092204408e-05,
+      "loss": 1.4551,
+      "step": 3230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718936980390923e-05,
+      "loss": 1.1533,
+      "step": 3231
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971869293434045e-05,
+      "loss": 1.2451,
+      "step": 3232
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718448783895286e-05,
+      "loss": 1.7578,
+      "step": 3233
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971820452905805e-05,
+      "loss": 1.3213,
+      "step": 3234
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717960169831367e-05,
+      "loss": 1.4346,
+      "step": 3235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717715706217864e-05,
+      "loss": 1.2568,
+      "step": 3236
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717471138220164e-05,
+      "loss": 0.959,
+      "step": 3237
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717226465840897e-05,
+      "loss": 1.502,
+      "step": 3238
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716981689082688e-05,
+      "loss": 1.127,
+      "step": 3239
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716736807948166e-05,
+      "loss": 1.6211,
+      "step": 3240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716491822439968e-05,
+      "loss": 1.3154,
+      "step": 3241
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716246732560715e-05,
+      "loss": 1.2725,
+      "step": 3242
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716001538313048e-05,
+      "loss": 1.2148,
+      "step": 3243
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97157562396996e-05,
+      "loss": 0.9502,
+      "step": 3244
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715510836723e-05,
+      "loss": 1.3496,
+      "step": 3245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971526532938589e-05,
+      "loss": 0.9834,
+      "step": 3246
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715019717690907e-05,
+      "loss": 1.0215,
+      "step": 3247
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714774001640684e-05,
+      "loss": 1.6152,
+      "step": 3248
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714528181237863e-05,
+      "loss": 1.5518,
+      "step": 3249
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971428225648509e-05,
+      "loss": 1.0449,
+      "step": 3250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714036227385002e-05,
+      "loss": 1.083,
+      "step": 3251
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713790093940237e-05,
+      "loss": 1.0996,
+      "step": 3252
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713543856153447e-05,
+      "loss": 1.3145,
+      "step": 3253
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713297514027272e-05,
+      "loss": 1.4629,
+      "step": 3254
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971305106756436e-05,
+      "loss": 0.915,
+      "step": 3255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712804516767357e-05,
+      "loss": 1.2695,
+      "step": 3256
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712557861638913e-05,
+      "loss": 1.3965,
+      "step": 3257
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712311102181682e-05,
+      "loss": 1.2207,
+      "step": 3258
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97120642383983e-05,
+      "loss": 1.4492,
+      "step": 3259
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711817270291435e-05,
+      "loss": 1.1475,
+      "step": 3260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711570197863733e-05,
+      "loss": 1.4238,
+      "step": 3261
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711323021117844e-05,
+      "loss": 1.0703,
+      "step": 3262
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971107574005643e-05,
+      "loss": 1.0508,
+      "step": 3263
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710828354682142e-05,
+      "loss": 1.4707,
+      "step": 3264
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971058086499764e-05,
+      "loss": 1.1631,
+      "step": 3265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710333271005586e-05,
+      "loss": 1.3066,
+      "step": 3266
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710085572708634e-05,
+      "loss": 1.3828,
+      "step": 3267
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970983777010944e-05,
+      "loss": 1.1396,
+      "step": 3268
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970958986321068e-05,
+      "loss": 0.918,
+      "step": 3269
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709341852015005e-05,
+      "loss": 0.9766,
+      "step": 3270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709093736525083e-05,
+      "loss": 1.4082,
+      "step": 3271
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708845516743583e-05,
+      "loss": 1.1475,
+      "step": 3272
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708597192673163e-05,
+      "loss": 1.1875,
+      "step": 3273
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708348764316497e-05,
+      "loss": 1.1592,
+      "step": 3274
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970810023167625e-05,
+      "loss": 0.9404,
+      "step": 3275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970785159475509e-05,
+      "loss": 1.3584,
+      "step": 3276
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707602853555697e-05,
+      "loss": 1.6973,
+      "step": 3277
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707354008080736e-05,
+      "loss": 1.4658,
+      "step": 3278
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970710505833288e-05,
+      "loss": 1.4258,
+      "step": 3279
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706856004314802e-05,
+      "loss": 1.5312,
+      "step": 3280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970660684602918e-05,
+      "loss": 1.0449,
+      "step": 3281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970635758347869e-05,
+      "loss": 1.6895,
+      "step": 3282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970610821666601e-05,
+      "loss": 1.085,
+      "step": 3283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705858745593812e-05,
+      "loss": 1.7812,
+      "step": 3284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705609170264787e-05,
+      "loss": 0.8643,
+      "step": 3285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970535949068161e-05,
+      "loss": 1.0908,
+      "step": 3286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970510970684696e-05,
+      "loss": 1.207,
+      "step": 3287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704859818763527e-05,
+      "loss": 0.873,
+      "step": 3288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970460982643399e-05,
+      "loss": 1.2715,
+      "step": 3289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704359729861035e-05,
+      "loss": 0.9961,
+      "step": 3290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970410952904735e-05,
+      "loss": 1.4336,
+      "step": 3291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970385922399562e-05,
+      "loss": 1.2666,
+      "step": 3292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703608814708538e-05,
+      "loss": 1.498,
+      "step": 3293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970335830118879e-05,
+      "loss": 0.9414,
+      "step": 3294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703107683439067e-05,
+      "loss": 1.4199,
+      "step": 3295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702856961462064e-05,
+      "loss": 1.0186,
+      "step": 3296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702606135260472e-05,
+      "loss": 1.5195,
+      "step": 3297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702355204836985e-05,
+      "loss": 0.9248,
+      "step": 3298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.97021041701943e-05,
+      "loss": 1.5,
+      "step": 3299
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701853031335113e-05,
+      "loss": 1.0957,
+      "step": 3300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701601788262116e-05,
+      "loss": 1.2949,
+      "step": 3301
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970135044097802e-05,
+      "loss": 1.3643,
+      "step": 3302
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701098989485517e-05,
+      "loss": 1.1104,
+      "step": 3303
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700847433787305e-05,
+      "loss": 1.4043,
+      "step": 3304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970059577388609e-05,
+      "loss": 1.2178,
+      "step": 3305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970034400978458e-05,
+      "loss": 1.0654,
+      "step": 3306
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700092141485473e-05,
+      "loss": 1.1494,
+      "step": 3307
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699840168991472e-05,
+      "loss": 1.2266,
+      "step": 3308
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699588092305294e-05,
+      "loss": 1.0957,
+      "step": 3309
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699335911429635e-05,
+      "loss": 1.2354,
+      "step": 3310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969908362636721e-05,
+      "loss": 1.1279,
+      "step": 3311
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969883123712073e-05,
+      "loss": 1.2529,
+      "step": 3312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698578743692906e-05,
+      "loss": 1.2832,
+      "step": 3313
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698326146086446e-05,
+      "loss": 1.7734,
+      "step": 3314
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698073444304065e-05,
+      "loss": 1.3848,
+      "step": 3315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697820638348482e-05,
+      "loss": 0.7705,
+      "step": 3316
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697567728222403e-05,
+      "loss": 1.502,
+      "step": 3317
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697314713928554e-05,
+      "loss": 1.543,
+      "step": 3318
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697061595469652e-05,
+      "loss": 1.4756,
+      "step": 3319
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696808372848405e-05,
+      "loss": 1.3652,
+      "step": 3320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696555046067548e-05,
+      "loss": 1.0381,
+      "step": 3321
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696301615129793e-05,
+      "loss": 1.5127,
+      "step": 3322
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696048080037865e-05,
+      "loss": 1.2134,
+      "step": 3323
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695794440794488e-05,
+      "loss": 1.2227,
+      "step": 3324
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695540697402382e-05,
+      "loss": 1.5391,
+      "step": 3325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695286849864283e-05,
+      "loss": 0.9912,
+      "step": 3326
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695032898182903e-05,
+      "loss": 1.166,
+      "step": 3327
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969477884236098e-05,
+      "loss": 1.3721,
+      "step": 3328
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969452468240124e-05,
+      "loss": 1.5156,
+      "step": 3329
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694270418306418e-05,
+      "loss": 1.5273,
+      "step": 3330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969401605007924e-05,
+      "loss": 1.1768,
+      "step": 3331
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693761577722434e-05,
+      "loss": 1.627,
+      "step": 3332
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693507001238745e-05,
+      "loss": 1.5273,
+      "step": 3333
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693252320630897e-05,
+      "loss": 1.3613,
+      "step": 3334
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969299753590163e-05,
+      "loss": 1.2861,
+      "step": 3335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692742647053683e-05,
+      "loss": 1.333,
+      "step": 3336
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692487654089788e-05,
+      "loss": 1.2041,
+      "step": 3337
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692232557012693e-05,
+      "loss": 1.3242,
+      "step": 3338
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691977355825132e-05,
+      "loss": 1.2656,
+      "step": 3339
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691722050529845e-05,
+      "loss": 1.1865,
+      "step": 3340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691466641129575e-05,
+      "loss": 1.4883,
+      "step": 3341
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969121112762707e-05,
+      "loss": 1.541,
+      "step": 3342
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690955510025072e-05,
+      "loss": 1.6133,
+      "step": 3343
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690699788326325e-05,
+      "loss": 1.1055,
+      "step": 3344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969044396253358e-05,
+      "loss": 1.5723,
+      "step": 3345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969018803264958e-05,
+      "loss": 1.1191,
+      "step": 3346
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689931998677077e-05,
+      "loss": 1.4561,
+      "step": 3347
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968967586061882e-05,
+      "loss": 0.9287,
+      "step": 3348
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968941961847756e-05,
+      "loss": 1.6094,
+      "step": 3349
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968916327225605e-05,
+      "loss": 1.5469,
+      "step": 3350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968890682195705e-05,
+      "loss": 1.7344,
+      "step": 3351
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688650267583302e-05,
+      "loss": 1.5996,
+      "step": 3352
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968839360913757e-05,
+      "loss": 1.1738,
+      "step": 3353
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968813684662261e-05,
+      "loss": 1.1235,
+      "step": 3354
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687879980041183e-05,
+      "loss": 1.418,
+      "step": 3355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687623009396042e-05,
+      "loss": 1.7422,
+      "step": 3356
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968736593468995e-05,
+      "loss": 1.0107,
+      "step": 3357
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968710875592567e-05,
+      "loss": 1.7461,
+      "step": 3358
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686851473105962e-05,
+      "loss": 1.3652,
+      "step": 3359
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686594086233588e-05,
+      "loss": 1.4844,
+      "step": 3360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686336595311317e-05,
+      "loss": 1.2373,
+      "step": 3361
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686079000341914e-05,
+      "loss": 1.3223,
+      "step": 3362
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685821301328144e-05,
+      "loss": 1.2754,
+      "step": 3363
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685563498272778e-05,
+      "loss": 1.377,
+      "step": 3364
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968530559117858e-05,
+      "loss": 1.1113,
+      "step": 3365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968504758004833e-05,
+      "loss": 1.1836,
+      "step": 3366
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684789464884792e-05,
+      "loss": 1.5645,
+      "step": 3367
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684531245690735e-05,
+      "loss": 1.3633,
+      "step": 3368
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968427292246894e-05,
+      "loss": 1.0273,
+      "step": 3369
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684014495222178e-05,
+      "loss": 1.4131,
+      "step": 3370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968375596395323e-05,
+      "loss": 1.2695,
+      "step": 3371
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683497328664868e-05,
+      "loss": 1.2627,
+      "step": 3372
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683238589359872e-05,
+      "loss": 1.0957,
+      "step": 3373
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682979746041016e-05,
+      "loss": 1.4336,
+      "step": 3374
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968272079871109e-05,
+      "loss": 1.4648,
+      "step": 3375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682461747372873e-05,
+      "loss": 1.627,
+      "step": 3376
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968220259202914e-05,
+      "loss": 1.4297,
+      "step": 3377
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681943332682683e-05,
+      "loss": 1.127,
+      "step": 3378
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681683969336284e-05,
+      "loss": 1.2832,
+      "step": 3379
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968142450199273e-05,
+      "loss": 1.0674,
+      "step": 3380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681164930654807e-05,
+      "loss": 1.2432,
+      "step": 3381
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680905255325302e-05,
+      "loss": 1.3438,
+      "step": 3382
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680645476007006e-05,
+      "loss": 1.0586,
+      "step": 3383
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680385592702707e-05,
+      "loss": 1.2139,
+      "step": 3384
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96801256054152e-05,
+      "loss": 1.334,
+      "step": 3385
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967986551414728e-05,
+      "loss": 1.2656,
+      "step": 3386
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679605318901732e-05,
+      "loss": 1.2969,
+      "step": 3387
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967934501968136e-05,
+      "loss": 1.3145,
+      "step": 3388
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679084616488955e-05,
+      "loss": 1.0342,
+      "step": 3389
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678824109327316e-05,
+      "loss": 0.9893,
+      "step": 3390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967856349819924e-05,
+      "loss": 1.2227,
+      "step": 3391
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678302783107526e-05,
+      "loss": 1.2061,
+      "step": 3392
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678041964054976e-05,
+      "loss": 1.7832,
+      "step": 3393
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967778104104439e-05,
+      "loss": 1.4219,
+      "step": 3394
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677520014078576e-05,
+      "loss": 1.3936,
+      "step": 3395
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967725888316033e-05,
+      "loss": 0.8818,
+      "step": 3396
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967699764829246e-05,
+      "loss": 1.3662,
+      "step": 3397
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676736309477778e-05,
+      "loss": 1.1826,
+      "step": 3398
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676474866719083e-05,
+      "loss": 1.0713,
+      "step": 3399
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676213320019183e-05,
+      "loss": 1.0518,
+      "step": 3400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675951669380894e-05,
+      "loss": 1.1748,
+      "step": 3401
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675689914807023e-05,
+      "loss": 1.0527,
+      "step": 3402
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675428056300384e-05,
+      "loss": 1.1484,
+      "step": 3403
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675166093863785e-05,
+      "loss": 1.2402,
+      "step": 3404
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674904027500044e-05,
+      "loss": 1.2598,
+      "step": 3405
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674641857211974e-05,
+      "loss": 1.416,
+      "step": 3406
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967437958300239e-05,
+      "loss": 1.5,
+      "step": 3407
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674117204874114e-05,
+      "loss": 1.082,
+      "step": 3408
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967385472282996e-05,
+      "loss": 1.8652,
+      "step": 3409
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967359213687275e-05,
+      "loss": 1.7031,
+      "step": 3410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673329447005303e-05,
+      "loss": 1.4883,
+      "step": 3411
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967306665323044e-05,
+      "loss": 1.4688,
+      "step": 3412
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672803755550987e-05,
+      "loss": 0.833,
+      "step": 3413
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672540753969767e-05,
+      "loss": 1.7832,
+      "step": 3414
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672277648489605e-05,
+      "loss": 1.1074,
+      "step": 3415
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672014439113324e-05,
+      "loss": 1.2363,
+      "step": 3416
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671751125843756e-05,
+      "loss": 1.2734,
+      "step": 3417
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671487708683723e-05,
+      "loss": 1.2959,
+      "step": 3418
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967122418763606e-05,
+      "loss": 1.2871,
+      "step": 3419
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670960562703596e-05,
+      "loss": 1.334,
+      "step": 3420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670696833889165e-05,
+      "loss": 1.3008,
+      "step": 3421
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670433001195595e-05,
+      "loss": 1.1611,
+      "step": 3422
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670169064625722e-05,
+      "loss": 1.7051,
+      "step": 3423
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669905024182387e-05,
+      "loss": 1.1006,
+      "step": 3424
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669640879868418e-05,
+      "loss": 1.502,
+      "step": 3425
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669376631686658e-05,
+      "loss": 1.2383,
+      "step": 3426
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669112279639942e-05,
+      "loss": 1.335,
+      "step": 3427
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668847823731107e-05,
+      "loss": 1.3271,
+      "step": 3428
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668583263963002e-05,
+      "loss": 1.0469,
+      "step": 3429
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668318600338462e-05,
+      "loss": 1.5742,
+      "step": 3430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966805383286033e-05,
+      "loss": 1.2715,
+      "step": 3431
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667788961531458e-05,
+      "loss": 1.0879,
+      "step": 3432
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966752398635468e-05,
+      "loss": 1.3008,
+      "step": 3433
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966725890733285e-05,
+      "loss": 1.3301,
+      "step": 3434
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666993724468812e-05,
+      "loss": 1.5938,
+      "step": 3435
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666728437765414e-05,
+      "loss": 1.1484,
+      "step": 3436
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666463047225506e-05,
+      "loss": 1.4629,
+      "step": 3437
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666197552851943e-05,
+      "loss": 1.3574,
+      "step": 3438
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665931954647574e-05,
+      "loss": 1.2646,
+      "step": 3439
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966566625261525e-05,
+      "loss": 0.8984,
+      "step": 3440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665400446757824e-05,
+      "loss": 1.6504,
+      "step": 3441
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966513453707816e-05,
+      "loss": 1.4365,
+      "step": 3442
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664868523579103e-05,
+      "loss": 1.1699,
+      "step": 3443
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664602406263514e-05,
+      "loss": 1.3496,
+      "step": 3444
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966433618513426e-05,
+      "loss": 1.4688,
+      "step": 3445
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664069860194188e-05,
+      "loss": 0.9238,
+      "step": 3446
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663803431446164e-05,
+      "loss": 1.2754,
+      "step": 3447
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663536898893052e-05,
+      "loss": 1.1045,
+      "step": 3448
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663270262537716e-05,
+      "loss": 1.5332,
+      "step": 3449
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663003522383013e-05,
+      "loss": 1.7754,
+      "step": 3450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662736678431814e-05,
+      "loss": 1.0518,
+      "step": 3451
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662469730686987e-05,
+      "loss": 1.1875,
+      "step": 3452
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662202679151395e-05,
+      "loss": 1.3896,
+      "step": 3453
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661935523827904e-05,
+      "loss": 1.6641,
+      "step": 3454
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661668264719393e-05,
+      "loss": 1.334,
+      "step": 3455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661400901828723e-05,
+      "loss": 0.957,
+      "step": 3456
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661133435158773e-05,
+      "loss": 1.0762,
+      "step": 3457
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660865864712413e-05,
+      "loss": 0.9404,
+      "step": 3458
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660598190492516e-05,
+      "loss": 0.9199,
+      "step": 3459
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660330412501963e-05,
+      "loss": 1.2715,
+      "step": 3460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660062530743626e-05,
+      "loss": 1.3945,
+      "step": 3461
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659794545220378e-05,
+      "loss": 1.9082,
+      "step": 3462
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659526455935106e-05,
+      "loss": 1.2178,
+      "step": 3463
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659258262890683e-05,
+      "loss": 1.6035,
+      "step": 3464
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658989966089997e-05,
+      "loss": 1.2227,
+      "step": 3465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658721565535923e-05,
+      "loss": 1.2046,
+      "step": 3466
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658453061231347e-05,
+      "loss": 1.3936,
+      "step": 3467
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658184453179155e-05,
+      "loss": 1.2295,
+      "step": 3468
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657915741382228e-05,
+      "loss": 1.4121,
+      "step": 3469
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657646925843456e-05,
+      "loss": 1.1455,
+      "step": 3470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657378006565727e-05,
+      "loss": 1.1172,
+      "step": 3471
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965710898355192e-05,
+      "loss": 1.0898,
+      "step": 3472
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965683985680494e-05,
+      "loss": 1.1924,
+      "step": 3473
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965657062632767e-05,
+      "loss": 1.1533,
+      "step": 3474
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656301292123005e-05,
+      "loss": 1.2207,
+      "step": 3475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965603185419383e-05,
+      "loss": 0.9902,
+      "step": 3476
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655762312543047e-05,
+      "loss": 1.4355,
+      "step": 3477
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655492667173548e-05,
+      "loss": 1.293,
+      "step": 3478
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655222918088235e-05,
+      "loss": 1.2402,
+      "step": 3479
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654953065289993e-05,
+      "loss": 0.9375,
+      "step": 3480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654683108781735e-05,
+      "loss": 1.2002,
+      "step": 3481
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654413048566353e-05,
+      "loss": 1.4902,
+      "step": 3482
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965414288464675e-05,
+      "loss": 1.5078,
+      "step": 3483
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653872617025828e-05,
+      "loss": 1.457,
+      "step": 3484
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965360224570649e-05,
+      "loss": 1.0303,
+      "step": 3485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653331770691638e-05,
+      "loss": 1.0928,
+      "step": 3486
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653061191984178e-05,
+      "loss": 1.2227,
+      "step": 3487
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965279050958702e-05,
+      "loss": 1.3848,
+      "step": 3488
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965251972350307e-05,
+      "loss": 1.7949,
+      "step": 3489
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965224883373523e-05,
+      "loss": 1.4531,
+      "step": 3490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965197784028642e-05,
+      "loss": 1.625,
+      "step": 3491
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965170674315955e-05,
+      "loss": 1.2969,
+      "step": 3492
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651435542357522e-05,
+      "loss": 1.2705,
+      "step": 3493
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965116423788326e-05,
+      "loss": 1.248,
+      "step": 3494
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965089282973967e-05,
+      "loss": 1.3809,
+      "step": 3495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650621317929675e-05,
+      "loss": 1.3574,
+      "step": 3496
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650349702456185e-05,
+      "loss": 1.2969,
+      "step": 3497
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650077983322124e-05,
+      "loss": 1.3828,
+      "step": 3498
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649806160530405e-05,
+      "loss": 1.2988,
+      "step": 3499
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964953423408395e-05,
+      "loss": 1.1973,
+      "step": 3500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964926220398568e-05,
+      "loss": 1.5781,
+      "step": 3501
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648990070238517e-05,
+      "loss": 1.3789,
+      "step": 3502
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648717832845383e-05,
+      "loss": 1.1553,
+      "step": 3503
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648445491809206e-05,
+      "loss": 1.1611,
+      "step": 3504
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648173047132907e-05,
+      "loss": 1.0186,
+      "step": 3505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647900498819412e-05,
+      "loss": 1.0078,
+      "step": 3506
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647627846871654e-05,
+      "loss": 1.0723,
+      "step": 3507
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964735509129256e-05,
+      "loss": 1.373,
+      "step": 3508
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647082232085057e-05,
+      "loss": 0.9048,
+      "step": 3509
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646809269252073e-05,
+      "loss": 1.0469,
+      "step": 3510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964653620279655e-05,
+      "loss": 1.1484,
+      "step": 3511
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646263032721412e-05,
+      "loss": 1.457,
+      "step": 3512
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96459897590296e-05,
+      "loss": 1.3496,
+      "step": 3513
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964571638172405e-05,
+      "loss": 1.3643,
+      "step": 3514
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645442900807686e-05,
+      "loss": 1.2549,
+      "step": 3515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964516931628346e-05,
+      "loss": 1.3223,
+      "step": 3516
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644895628154303e-05,
+      "loss": 1.1436,
+      "step": 3517
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644621836423157e-05,
+      "loss": 1.04,
+      "step": 3518
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644347941092964e-05,
+      "loss": 1.4424,
+      "step": 3519
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644073942166667e-05,
+      "loss": 1.0322,
+      "step": 3520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643799839647207e-05,
+      "loss": 1.2646,
+      "step": 3521
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643525633537525e-05,
+      "loss": 1.4297,
+      "step": 3522
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643251323840572e-05,
+      "loss": 1.0176,
+      "step": 3523
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642976910559296e-05,
+      "loss": 1.4062,
+      "step": 3524
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642702393696638e-05,
+      "loss": 1.2207,
+      "step": 3525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964242777325555e-05,
+      "loss": 1.4492,
+      "step": 3526
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642153049238982e-05,
+      "loss": 1.0742,
+      "step": 3527
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641878221649886e-05,
+      "loss": 1.1426,
+      "step": 3528
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964160329049121e-05,
+      "loss": 1.2305,
+      "step": 3529
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641328255765916e-05,
+      "loss": 1.6553,
+      "step": 3530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964105311747695e-05,
+      "loss": 1.4541,
+      "step": 3531
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964077787562727e-05,
+      "loss": 1.4629,
+      "step": 3532
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640502530219828e-05,
+      "loss": 1.4512,
+      "step": 3533
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964022708125759e-05,
+      "loss": 1.248,
+      "step": 3534
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639951528743513e-05,
+      "loss": 1.0986,
+      "step": 3535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639675872680552e-05,
+      "loss": 1.1572,
+      "step": 3536
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963940011307167e-05,
+      "loss": 1.1865,
+      "step": 3537
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963912424991983e-05,
+      "loss": 0.9185,
+      "step": 3538
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638848283228e-05,
+      "loss": 1.6016,
+      "step": 3539
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638572212999135e-05,
+      "loss": 1.25,
+      "step": 3540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638296039236206e-05,
+      "loss": 1.1953,
+      "step": 3541
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963801976194218e-05,
+      "loss": 1.2676,
+      "step": 3542
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963774338112002e-05,
+      "loss": 1.1592,
+      "step": 3543
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96374668967727e-05,
+      "loss": 1.0566,
+      "step": 3544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963719030890319e-05,
+      "loss": 1.3223,
+      "step": 3545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636913617514456e-05,
+      "loss": 1.5469,
+      "step": 3546
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636636822609473e-05,
+      "loss": 1.6035,
+      "step": 3547
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636359924191216e-05,
+      "loss": 1.5605,
+      "step": 3548
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636082922262653e-05,
+      "loss": 0.8516,
+      "step": 3549
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963580581682677e-05,
+      "loss": 1.5449,
+      "step": 3550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635528607886532e-05,
+      "loss": 1.4199,
+      "step": 3551
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635251295444927e-05,
+      "loss": 1.2754,
+      "step": 3552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634973879504928e-05,
+      "loss": 1.3311,
+      "step": 3553
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634696360069517e-05,
+      "loss": 0.9648,
+      "step": 3554
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963441873714167e-05,
+      "loss": 1.2705,
+      "step": 3555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634141010724373e-05,
+      "loss": 1.0312,
+      "step": 3556
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633863180820617e-05,
+      "loss": 1.4756,
+      "step": 3557
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963358524743337e-05,
+      "loss": 0.9121,
+      "step": 3558
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963330721056563e-05,
+      "loss": 1.1797,
+      "step": 3559
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633029070220382e-05,
+      "loss": 1.2754,
+      "step": 3560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632750826400606e-05,
+      "loss": 1.5703,
+      "step": 3561
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96324724791093e-05,
+      "loss": 0.9775,
+      "step": 3562
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963219402834945e-05,
+      "loss": 1.1377,
+      "step": 3563
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631915474124045e-05,
+      "loss": 1.2559,
+      "step": 3564
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631636816436083e-05,
+      "loss": 1.292,
+      "step": 3565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631358055288546e-05,
+      "loss": 1.2012,
+      "step": 3566
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963107919068444e-05,
+      "loss": 1.1631,
+      "step": 3567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963080022262676e-05,
+      "loss": 1.623,
+      "step": 3568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630521151118496e-05,
+      "loss": 1.1611,
+      "step": 3569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963024197616265e-05,
+      "loss": 1.3965,
+      "step": 3570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629962697762218e-05,
+      "loss": 1.2949,
+      "step": 3571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96296833159202e-05,
+      "loss": 1.248,
+      "step": 3572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96294038306396e-05,
+      "loss": 1.5771,
+      "step": 3573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629124241923415e-05,
+      "loss": 1.3262,
+      "step": 3574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628844549774657e-05,
+      "loss": 1.2324,
+      "step": 3575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628564754196322e-05,
+      "loss": 1.1123,
+      "step": 3576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962828485519142e-05,
+      "loss": 1.0273,
+      "step": 3577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628004852762952e-05,
+      "loss": 1.1836,
+      "step": 3578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627724746913935e-05,
+      "loss": 1.4688,
+      "step": 3579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962744453764737e-05,
+      "loss": 1.4297,
+      "step": 3580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962716422496627e-05,
+      "loss": 1.2129,
+      "step": 3581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626883808873646e-05,
+      "loss": 1.1523,
+      "step": 3582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626603289372507e-05,
+      "loss": 0.8848,
+      "step": 3583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626322666465874e-05,
+      "loss": 1.2559,
+      "step": 3584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962604194015675e-05,
+      "loss": 1.333,
+      "step": 3585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962576111044816e-05,
+      "loss": 1.4785,
+      "step": 3586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625480177343118e-05,
+      "loss": 1.1982,
+      "step": 3587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962519914084464e-05,
+      "loss": 0.9941,
+      "step": 3588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624918000955743e-05,
+      "loss": 1.2236,
+      "step": 3589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624636757679455e-05,
+      "loss": 1.4297,
+      "step": 3590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962435541101879e-05,
+      "loss": 1.4316,
+      "step": 3591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624073960976765e-05,
+      "loss": 1.8105,
+      "step": 3592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9623792407556416e-05,
+      "loss": 1.3252,
+      "step": 3593
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962351075076076e-05,
+      "loss": 1.252,
+      "step": 3594
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623228990592826e-05,
+      "loss": 1.3262,
+      "step": 3595
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622947127055637e-05,
+      "loss": 1.2686,
+      "step": 3596
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622665160152218e-05,
+      "loss": 1.4531,
+      "step": 3597
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962238308988561e-05,
+      "loss": 1.4043,
+      "step": 3598
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962210091625883e-05,
+      "loss": 1.4707,
+      "step": 3599
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621818639274913e-05,
+      "loss": 1.5391,
+      "step": 3600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621536258936888e-05,
+      "loss": 1.0957,
+      "step": 3601
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96212537752478e-05,
+      "loss": 1.0981,
+      "step": 3602
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620971188210668e-05,
+      "loss": 1.0693,
+      "step": 3603
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620688497828538e-05,
+      "loss": 1.3066,
+      "step": 3604
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620405704104448e-05,
+      "loss": 1.2324,
+      "step": 3605
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620122807041427e-05,
+      "loss": 1.4033,
+      "step": 3606
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619839806642518e-05,
+      "loss": 1.2754,
+      "step": 3607
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961955670291076e-05,
+      "loss": 1.251,
+      "step": 3608
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619273495849198e-05,
+      "loss": 1.5781,
+      "step": 3609
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618990185460865e-05,
+      "loss": 1.0449,
+      "step": 3610
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618706771748814e-05,
+      "loss": 1.3594,
+      "step": 3611
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618423254716083e-05,
+      "loss": 1.2432,
+      "step": 3612
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618139634365723e-05,
+      "loss": 1.2656,
+      "step": 3613
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617855910700776e-05,
+      "loss": 0.9102,
+      "step": 3614
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961757208372429e-05,
+      "loss": 1.6543,
+      "step": 3615
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617288153439317e-05,
+      "loss": 1.3291,
+      "step": 3616
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617004119848904e-05,
+      "loss": 1.5742,
+      "step": 3617
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616719982956097e-05,
+      "loss": 1.1133,
+      "step": 3618
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961643574276396e-05,
+      "loss": 1.1914,
+      "step": 3619
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961615139927554e-05,
+      "loss": 1.4404,
+      "step": 3620
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961586695249389e-05,
+      "loss": 1.1777,
+      "step": 3621
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615582402422067e-05,
+      "loss": 1.2041,
+      "step": 3622
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615297749063128e-05,
+      "loss": 1.5,
+      "step": 3623
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615012992420126e-05,
+      "loss": 1.793,
+      "step": 3624
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961472813249613e-05,
+      "loss": 1.4941,
+      "step": 3625
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614443169294186e-05,
+      "loss": 1.4248,
+      "step": 3626
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614158102817366e-05,
+      "loss": 1.1367,
+      "step": 3627
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961387293306873e-05,
+      "loss": 0.958,
+      "step": 3628
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613587660051338e-05,
+      "loss": 1.4102,
+      "step": 3629
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613302283768254e-05,
+      "loss": 1.1074,
+      "step": 3630
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613016804222547e-05,
+      "loss": 1.4844,
+      "step": 3631
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612731221417286e-05,
+      "loss": 1.1758,
+      "step": 3632
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961244553535553e-05,
+      "loss": 1.6016,
+      "step": 3633
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612159746040355e-05,
+      "loss": 0.9912,
+      "step": 3634
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611873853474826e-05,
+      "loss": 1.1367,
+      "step": 3635
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611587857662016e-05,
+      "loss": 1.1445,
+      "step": 3636
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611301758605e-05,
+      "loss": 1.3604,
+      "step": 3637
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611015556306845e-05,
+      "loss": 1.5176,
+      "step": 3638
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961072925077063e-05,
+      "loss": 1.7441,
+      "step": 3639
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961044284199943e-05,
+      "loss": 1.0996,
+      "step": 3640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961015632999632e-05,
+      "loss": 1.7305,
+      "step": 3641
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609869714764377e-05,
+      "loss": 1.043,
+      "step": 3642
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609582996306682e-05,
+      "loss": 1.6895,
+      "step": 3643
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609296174626313e-05,
+      "loss": 0.8149,
+      "step": 3644
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609009249726353e-05,
+      "loss": 1.0723,
+      "step": 3645
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608722221609878e-05,
+      "loss": 1.582,
+      "step": 3646
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960843509027998e-05,
+      "loss": 1.3975,
+      "step": 3647
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960814785573974e-05,
+      "loss": 1.1797,
+      "step": 3648
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607860517992238e-05,
+      "loss": 1.0981,
+      "step": 3649
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960757307704057e-05,
+      "loss": 0.8721,
+      "step": 3650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607285532887817e-05,
+      "loss": 1.1377,
+      "step": 3651
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960699788553707e-05,
+      "loss": 1.498,
+      "step": 3652
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606710134991414e-05,
+      "loss": 1.3018,
+      "step": 3653
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606422281253945e-05,
+      "loss": 1.7109,
+      "step": 3654
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606134324327755e-05,
+      "loss": 1.2832,
+      "step": 3655
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605846264215936e-05,
+      "loss": 1.4883,
+      "step": 3656
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605558100921583e-05,
+      "loss": 1.4473,
+      "step": 3657
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960526983444779e-05,
+      "loss": 1.0103,
+      "step": 3658
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604981464797655e-05,
+      "loss": 1.2285,
+      "step": 3659
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604692991974274e-05,
+      "loss": 1.168,
+      "step": 3660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604404415980746e-05,
+      "loss": 1.0225,
+      "step": 3661
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604115736820173e-05,
+      "loss": 1.0332,
+      "step": 3662
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603826954495653e-05,
+      "loss": 1.3281,
+      "step": 3663
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603538069010285e-05,
+      "loss": 1.0176,
+      "step": 3664
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960324908036718e-05,
+      "loss": 1.2061,
+      "step": 3665
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960295998856944e-05,
+      "loss": 1.1094,
+      "step": 3666
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602670793620167e-05,
+      "loss": 1.3877,
+      "step": 3667
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602381495522465e-05,
+      "loss": 1.4531,
+      "step": 3668
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960209209427945e-05,
+      "loss": 1.6113,
+      "step": 3669
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601802589894227e-05,
+      "loss": 1.3076,
+      "step": 3670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96015129823699e-05,
+      "loss": 1.416,
+      "step": 3671
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960122327170959e-05,
+      "loss": 0.9746,
+      "step": 3672
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600933457916404e-05,
+      "loss": 1.2285,
+      "step": 3673
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600643540993453e-05,
+      "loss": 1.0508,
+      "step": 3674
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600353520943854e-05,
+      "loss": 1.3477,
+      "step": 3675
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960006339777072e-05,
+      "loss": 1.2344,
+      "step": 3676
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959977317147717e-05,
+      "loss": 1.1797,
+      "step": 3677
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599482842066314e-05,
+      "loss": 1.4609,
+      "step": 3678
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599192409541286e-05,
+      "loss": 1.459,
+      "step": 3679
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598901873905192e-05,
+      "loss": 1.0576,
+      "step": 3680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598611235161157e-05,
+      "loss": 1.1094,
+      "step": 3681
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598320493312307e-05,
+      "loss": 1.4072,
+      "step": 3682
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959802964836176e-05,
+      "loss": 1.126,
+      "step": 3683
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959773870031264e-05,
+      "loss": 1.3369,
+      "step": 3684
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597447649168072e-05,
+      "loss": 1.2334,
+      "step": 3685
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597156494931186e-05,
+      "loss": 1.3643,
+      "step": 3686
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959686523760511e-05,
+      "loss": 1.0859,
+      "step": 3687
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959657387719297e-05,
+      "loss": 1.3564,
+      "step": 3688
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596282413697893e-05,
+      "loss": 0.998,
+      "step": 3689
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595990847123013e-05,
+      "loss": 1.2676,
+      "step": 3690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595699177471465e-05,
+      "loss": 1.5098,
+      "step": 3691
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595407404746377e-05,
+      "loss": 1.0449,
+      "step": 3692
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595115528950882e-05,
+      "loss": 1.2178,
+      "step": 3693
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594823550088123e-05,
+      "loss": 1.0596,
+      "step": 3694
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594531468161227e-05,
+      "loss": 1.3945,
+      "step": 3695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594239283173336e-05,
+      "loss": 1.041,
+      "step": 3696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959394699512759e-05,
+      "loss": 1.252,
+      "step": 3697
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959365460402713e-05,
+      "loss": 1.1348,
+      "step": 3698
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959336210987509e-05,
+      "loss": 1.8555,
+      "step": 3699
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593069512674615e-05,
+      "loss": 1.4707,
+      "step": 3700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959277681242885e-05,
+      "loss": 1.6699,
+      "step": 3701
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592484009140934e-05,
+      "loss": 1.165,
+      "step": 3702
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959219110281402e-05,
+      "loss": 1.1553,
+      "step": 3703
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959189809345125e-05,
+      "loss": 1.1523,
+      "step": 3704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959160498105577e-05,
+      "loss": 1.5723,
+      "step": 3705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959131176563073e-05,
+      "loss": 1.3467,
+      "step": 3706
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9591018447179283e-05,
+      "loss": 1.6621,
+      "step": 3707
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959072502570457e-05,
+      "loss": 1.752,
+      "step": 3708
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590431501209754e-05,
+      "loss": 1.5938,
+      "step": 3709
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959013787369798e-05,
+      "loss": 1.0977,
+      "step": 3710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958984414317241e-05,
+      "loss": 1.293,
+      "step": 3711
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958955030963619e-05,
+      "loss": 1.1602,
+      "step": 3712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589256373092482e-05,
+      "loss": 1.374,
+      "step": 3713
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588962333544442e-05,
+      "loss": 1.0537,
+      "step": 3714
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588668190995227e-05,
+      "loss": 0.9014,
+      "step": 3715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588373945448e-05,
+      "loss": 1.5703,
+      "step": 3716
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958807959690592e-05,
+      "loss": 1.2393,
+      "step": 3717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587785145372146e-05,
+      "loss": 1.1221,
+      "step": 3718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587490590849846e-05,
+      "loss": 1.626,
+      "step": 3719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587195933342177e-05,
+      "loss": 1.623,
+      "step": 3720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586901172852313e-05,
+      "loss": 1.2158,
+      "step": 3721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586606309383415e-05,
+      "loss": 1.5205,
+      "step": 3722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586311342938652e-05,
+      "loss": 1.1133,
+      "step": 3723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586016273521188e-05,
+      "loss": 0.9473,
+      "step": 3724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585721101134198e-05,
+      "loss": 1.2236,
+      "step": 3725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958542582578085e-05,
+      "loss": 1.2285,
+      "step": 3726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585130447464316e-05,
+      "loss": 1.4121,
+      "step": 3727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584834966187768e-05,
+      "loss": 1.4629,
+      "step": 3728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958453938195438e-05,
+      "loss": 1.3037,
+      "step": 3729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584243694767332e-05,
+      "loss": 0.7178,
+      "step": 3730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583947904629794e-05,
+      "loss": 1.1982,
+      "step": 3731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583652011544946e-05,
+      "loss": 1.0879,
+      "step": 3732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583356015515966e-05,
+      "loss": 1.4238,
+      "step": 3733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583059916546034e-05,
+      "loss": 0.9336,
+      "step": 3734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958276371463833e-05,
+      "loss": 1.0576,
+      "step": 3735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582467409796035e-05,
+      "loss": 1.0244,
+      "step": 3736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582171002022333e-05,
+      "loss": 0.9507,
+      "step": 3737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581874491320406e-05,
+      "loss": 1.3486,
+      "step": 3738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581577877693446e-05,
+      "loss": 1.3213,
+      "step": 3739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581281161144626e-05,
+      "loss": 1.0684,
+      "step": 3740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580984341677144e-05,
+      "loss": 1.2266,
+      "step": 3741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580687419294185e-05,
+      "loss": 1.0098,
+      "step": 3742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958039039399894e-05,
+      "loss": 1.2422,
+      "step": 3743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95800932657946e-05,
+      "loss": 0.8926,
+      "step": 3744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957979603468435e-05,
+      "loss": 1.5879,
+      "step": 3745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579498700671386e-05,
+      "loss": 1.0869,
+      "step": 3746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957920126375891e-05,
+      "loss": 1.4375,
+      "step": 3747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578903723950103e-05,
+      "loss": 1.7266,
+      "step": 3748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578606081248175e-05,
+      "loss": 1.25,
+      "step": 3749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578308335656315e-05,
+      "loss": 1.7773,
+      "step": 3750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578010487177717e-05,
+      "loss": 1.0835,
+      "step": 3751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957771253581559e-05,
+      "loss": 1.3584,
+      "step": 3752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957741448157313e-05,
+      "loss": 1.8633,
+      "step": 3753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957711632445354e-05,
+      "loss": 1.0342,
+      "step": 3754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576818064460025e-05,
+      "loss": 1.2656,
+      "step": 3755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576519701595784e-05,
+      "loss": 0.9873,
+      "step": 3756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576221235864024e-05,
+      "loss": 1.4883,
+      "step": 3757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575922667267945e-05,
+      "loss": 1.2812,
+      "step": 3758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575623995810767e-05,
+      "loss": 1.418,
+      "step": 3759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575325221495687e-05,
+      "loss": 0.9316,
+      "step": 3760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957502634432592e-05,
+      "loss": 1.4395,
+      "step": 3761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574727364304674e-05,
+      "loss": 1.0186,
+      "step": 3762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957442828143516e-05,
+      "loss": 1.4531,
+      "step": 3763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574129095720597e-05,
+      "loss": 1.0361,
+      "step": 3764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957382980716419e-05,
+      "loss": 1.3906,
+      "step": 3765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573530415769158e-05,
+      "loss": 1.2168,
+      "step": 3766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573230921538714e-05,
+      "loss": 1.4141,
+      "step": 3767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957293132447608e-05,
+      "loss": 1.5303,
+      "step": 3768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957263162458447e-05,
+      "loss": 1.2812,
+      "step": 3769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572331821867108e-05,
+      "loss": 0.9639,
+      "step": 3770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572031916327208e-05,
+      "loss": 1.4219,
+      "step": 3771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571731907968e-05,
+      "loss": 1.0869,
+      "step": 3772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571431796792698e-05,
+      "loss": 1.2061,
+      "step": 3773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571131582804527e-05,
+      "loss": 1.6504,
+      "step": 3774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570831266006717e-05,
+      "loss": 0.8213,
+      "step": 3775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570530846402488e-05,
+      "loss": 1.6523,
+      "step": 3776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570230323995073e-05,
+      "loss": 1.0518,
+      "step": 3777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569929698787694e-05,
+      "loss": 1.3232,
+      "step": 3778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569628970783586e-05,
+      "loss": 1.4521,
+      "step": 3779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569328139985973e-05,
+      "loss": 1.5195,
+      "step": 3780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569027206398093e-05,
+      "loss": 1.3066,
+      "step": 3781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568726170023174e-05,
+      "loss": 1.2676,
+      "step": 3782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956842503086445e-05,
+      "loss": 0.9346,
+      "step": 3783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956812378892516e-05,
+      "loss": 1.2432,
+      "step": 3784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956782244420853e-05,
+      "loss": 0.9575,
+      "step": 3785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956752099671781e-05,
+      "loss": 1.1455,
+      "step": 3786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567219446456226e-05,
+      "loss": 1.4531,
+      "step": 3787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566917793427028e-05,
+      "loss": 1.3555,
+      "step": 3788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566616037633447e-05,
+      "loss": 1.3467,
+      "step": 3789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956631417907873e-05,
+      "loss": 1.2949,
+      "step": 3790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566012217766118e-05,
+      "loss": 0.9219,
+      "step": 3791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565710153698853e-05,
+      "loss": 1.2549,
+      "step": 3792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565407986880184e-05,
+      "loss": 1.3066,
+      "step": 3793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565105717313352e-05,
+      "loss": 0.8828,
+      "step": 3794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564803345001603e-05,
+      "loss": 1.1016,
+      "step": 3795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564500869948193e-05,
+      "loss": 1.0439,
+      "step": 3796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956419829215636e-05,
+      "loss": 1.2334,
+      "step": 3797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956389561162936e-05,
+      "loss": 1.127,
+      "step": 3798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563592828370448e-05,
+      "loss": 1.75,
+      "step": 3799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956328994238287e-05,
+      "loss": 1.1797,
+      "step": 3800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562986953669883e-05,
+      "loss": 1.04,
+      "step": 3801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956268386223474e-05,
+      "loss": 1.3062,
+      "step": 3802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562380668080696e-05,
+      "loss": 1.4346,
+      "step": 3803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956207737121101e-05,
+      "loss": 1.2598,
+      "step": 3804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956177397162894e-05,
+      "loss": 1.3242,
+      "step": 3805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561470469337744e-05,
+      "loss": 1.2441,
+      "step": 3806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561166864340682e-05,
+      "loss": 1.1152,
+      "step": 3807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560863156641013e-05,
+      "loss": 1.2793,
+      "step": 3808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560559346242e-05,
+      "loss": 1.1211,
+      "step": 3809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560255433146916e-05,
+      "loss": 1.5625,
+      "step": 3810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955995141735901e-05,
+      "loss": 1.2998,
+      "step": 3811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955964729888156e-05,
+      "loss": 1.4219,
+      "step": 3812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559343077717826e-05,
+      "loss": 0.9902,
+      "step": 3813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559038753871077e-05,
+      "loss": 1.2236,
+      "step": 3814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558734327344582e-05,
+      "loss": 1.3359,
+      "step": 3815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558429798141614e-05,
+      "loss": 1.0273,
+      "step": 3816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955812516626544e-05,
+      "loss": 1.4043,
+      "step": 3817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557820431719333e-05,
+      "loss": 1.543,
+      "step": 3818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557515594506572e-05,
+      "loss": 1.1035,
+      "step": 3819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557210654630423e-05,
+      "loss": 1.5918,
+      "step": 3820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955690561209417e-05,
+      "loss": 0.96,
+      "step": 3821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955660046690108e-05,
+      "loss": 1.2207,
+      "step": 3822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556295219054437e-05,
+      "loss": 1.0449,
+      "step": 3823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555989868557522e-05,
+      "loss": 1.249,
+      "step": 3824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555684415413607e-05,
+      "loss": 1.4219,
+      "step": 3825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555378859625978e-05,
+      "loss": 1.7598,
+      "step": 3826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955507320119792e-05,
+      "loss": 1.7402,
+      "step": 3827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554767440132713e-05,
+      "loss": 1.5156,
+      "step": 3828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955446157643364e-05,
+      "loss": 1.3906,
+      "step": 3829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955415561010399e-05,
+      "loss": 1.2051,
+      "step": 3830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955384954114705e-05,
+      "loss": 1.2969,
+      "step": 3831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95535433695661e-05,
+      "loss": 1.416,
+      "step": 3832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553237095364433e-05,
+      "loss": 1.2461,
+      "step": 3833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552930718545346e-05,
+      "loss": 0.9258,
+      "step": 3834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552624239112118e-05,
+      "loss": 0.6448,
+      "step": 3835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955231765706805e-05,
+      "loss": 0.8369,
+      "step": 3836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552010972416435e-05,
+      "loss": 0.8584,
+      "step": 3837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955170418516056e-05,
+      "loss": 1.0078,
+      "step": 3838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551397295303728e-05,
+      "loss": 1.2998,
+      "step": 3839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551090302849233e-05,
+      "loss": 1.3691,
+      "step": 3840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955078320780037e-05,
+      "loss": 0.6519,
+      "step": 3841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955047601016044e-05,
+      "loss": 1.6055,
+      "step": 3842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955016870993274e-05,
+      "loss": 1.1992,
+      "step": 3843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549861307120578e-05,
+      "loss": 1.3296,
+      "step": 3844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954955380172725e-05,
+      "loss": 0.8955,
+      "step": 3845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954924619375606e-05,
+      "loss": 1.4941,
+      "step": 3846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548938483210317e-05,
+      "loss": 1.3218,
+      "step": 3847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548630670093318e-05,
+      "loss": 1.4512,
+      "step": 3848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548322754408374e-05,
+      "loss": 1.3789,
+      "step": 3849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954801473615879e-05,
+      "loss": 1.1533,
+      "step": 3850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954770661534788e-05,
+      "loss": 1.3496,
+      "step": 3851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954739839197895e-05,
+      "loss": 1.4551,
+      "step": 3852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954709006605531e-05,
+      "loss": 0.8115,
+      "step": 3853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546781637580272e-05,
+      "loss": 1.1406,
+      "step": 3854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546473106557154e-05,
+      "loss": 1.0088,
+      "step": 3855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546164472989263e-05,
+      "loss": 1.2246,
+      "step": 3856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954585573687992e-05,
+      "loss": 1.6367,
+      "step": 3857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545546898232437e-05,
+      "loss": 1.1992,
+      "step": 3858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545237957050135e-05,
+      "loss": 1.4746,
+      "step": 3859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544928913336328e-05,
+      "loss": 1.6934,
+      "step": 3860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544619767094343e-05,
+      "loss": 1.4219,
+      "step": 3861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544310518327493e-05,
+      "loss": 1.4795,
+      "step": 3862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544001167039103e-05,
+      "loss": 1.291,
+      "step": 3863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95436917132325e-05,
+      "loss": 1.0322,
+      "step": 3864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543382156910997e-05,
+      "loss": 0.9297,
+      "step": 3865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954307249807793e-05,
+      "loss": 0.9561,
+      "step": 3866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954276273673662e-05,
+      "loss": 1.1973,
+      "step": 3867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95424528728904e-05,
+      "loss": 1.1914,
+      "step": 3868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954214290654259e-05,
+      "loss": 1.2036,
+      "step": 3869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541832837696527e-05,
+      "loss": 1.3594,
+      "step": 3870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541522666355537e-05,
+      "loss": 0.9492,
+      "step": 3871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541212392522953e-05,
+      "loss": 1.1328,
+      "step": 3872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540902016202107e-05,
+      "loss": 1.9727,
+      "step": 3873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540591537396336e-05,
+      "loss": 1.0723,
+      "step": 3874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540280956108973e-05,
+      "loss": 1.2354,
+      "step": 3875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539970272343353e-05,
+      "loss": 1.1689,
+      "step": 3876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539659486102813e-05,
+      "loss": 1.4785,
+      "step": 3877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539348597390693e-05,
+      "loss": 1.2314,
+      "step": 3878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539037606210335e-05,
+      "loss": 1.4492,
+      "step": 3879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538726512565075e-05,
+      "loss": 1.1338,
+      "step": 3880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953841531645826e-05,
+      "loss": 1.0488,
+      "step": 3881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953810401789322e-05,
+      "loss": 1.2822,
+      "step": 3882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953779261687332e-05,
+      "loss": 0.9519,
+      "step": 3883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537481113401884e-05,
+      "loss": 1.5312,
+      "step": 3884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953716950748227e-05,
+      "loss": 1.4043,
+      "step": 3885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536857799117823e-05,
+      "loss": 1.2109,
+      "step": 3886
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953654598831189e-05,
+      "loss": 1.374,
+      "step": 3887
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953623407506782e-05,
+      "loss": 1.2861,
+      "step": 3888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535922059388967e-05,
+      "loss": 1.5625,
+      "step": 3889
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535609941278676e-05,
+      "loss": 1.0967,
+      "step": 3890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953529772074031e-05,
+      "loss": 0.9746,
+      "step": 3891
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953498539777721e-05,
+      "loss": 1.5469,
+      "step": 3892
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534672972392737e-05,
+      "loss": 0.9609,
+      "step": 3893
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953436044459025e-05,
+      "loss": 1.0059,
+      "step": 3894
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534047814373103e-05,
+      "loss": 1.0605,
+      "step": 3895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533735081744654e-05,
+      "loss": 1.167,
+      "step": 3896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953342224670826e-05,
+      "loss": 1.1396,
+      "step": 3897
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953310930926729e-05,
+      "loss": 1.457,
+      "step": 3898
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532796269425093e-05,
+      "loss": 1.1689,
+      "step": 3899
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953248312718504e-05,
+      "loss": 1.0811,
+      "step": 3900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532169882550493e-05,
+      "loss": 1.293,
+      "step": 3901
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953185653552482e-05,
+      "loss": 1.1113,
+      "step": 3902
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531543086111377e-05,
+      "loss": 1.2793,
+      "step": 3903
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953122953431354e-05,
+      "loss": 1.0908,
+      "step": 3904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9530915880134677e-05,
+      "loss": 1.1982,
+      "step": 3905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953060212357815e-05,
+      "loss": 1.3896,
+      "step": 3906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9530288264647335e-05,
+      "loss": 1.542,
+      "step": 3907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529974303345605e-05,
+      "loss": 1.0117,
+      "step": 3908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529660239676325e-05,
+      "loss": 1.123,
+      "step": 3909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952934607364288e-05,
+      "loss": 1.582,
+      "step": 3910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952903180524863e-05,
+      "loss": 1.4473,
+      "step": 3911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528717434496963e-05,
+      "loss": 1.0254,
+      "step": 3912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528402961391248e-05,
+      "loss": 1.3896,
+      "step": 3913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952808838593487e-05,
+      "loss": 1.2715,
+      "step": 3914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95277737081312e-05,
+      "loss": 1.4053,
+      "step": 3915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527458927983626e-05,
+      "loss": 0.8652,
+      "step": 3916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527144045495524e-05,
+      "loss": 1.2002,
+      "step": 3917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952682906067028e-05,
+      "loss": 1.9336,
+      "step": 3918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526513973511276e-05,
+      "loss": 0.9678,
+      "step": 3919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526198784021896e-05,
+      "loss": 1.1113,
+      "step": 3920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525883492205522e-05,
+      "loss": 1.4688,
+      "step": 3921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952556809806555e-05,
+      "loss": 1.4746,
+      "step": 3922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525252601605357e-05,
+      "loss": 1.6172,
+      "step": 3923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524937002828343e-05,
+      "loss": 1.377,
+      "step": 3924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524621301737886e-05,
+      "loss": 1.1719,
+      "step": 3925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524305498337387e-05,
+      "loss": 1.0913,
+      "step": 3926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523989592630235e-05,
+      "loss": 1.1982,
+      "step": 3927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523673584619823e-05,
+      "loss": 1.2676,
+      "step": 3928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523357474309545e-05,
+      "loss": 1.0029,
+      "step": 3929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95230412617028e-05,
+      "loss": 1.1626,
+      "step": 3930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522724946802977e-05,
+      "loss": 1.3906,
+      "step": 3931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522408529613483e-05,
+      "loss": 0.9678,
+      "step": 3932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952209201013771e-05,
+      "loss": 0.98,
+      "step": 3933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952177538837906e-05,
+      "loss": 1.5684,
+      "step": 3934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521458664340932e-05,
+      "loss": 1.0107,
+      "step": 3935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521141838026735e-05,
+      "loss": 1.2812,
+      "step": 3936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520824909439865e-05,
+      "loss": 1.6406,
+      "step": 3937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520507878583728e-05,
+      "loss": 1.585,
+      "step": 3938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520190745461733e-05,
+      "loss": 1.0264,
+      "step": 3939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519873510077283e-05,
+      "loss": 1.209,
+      "step": 3940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519556172433785e-05,
+      "loss": 1.4971,
+      "step": 3941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519238732534657e-05,
+      "loss": 1.4355,
+      "step": 3942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518921190383294e-05,
+      "loss": 1.3027,
+      "step": 3943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518603545983114e-05,
+      "loss": 1.6719,
+      "step": 3944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951828579933753e-05,
+      "loss": 1.2588,
+      "step": 3945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517967950449956e-05,
+      "loss": 1.1992,
+      "step": 3946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95176499993238e-05,
+      "loss": 1.3486,
+      "step": 3947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517331945962487e-05,
+      "loss": 1.2207,
+      "step": 3948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517013790369427e-05,
+      "loss": 1.0889,
+      "step": 3949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516695532548037e-05,
+      "loss": 0.8926,
+      "step": 3950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516377172501738e-05,
+      "loss": 1.0869,
+      "step": 3951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516058710233948e-05,
+      "loss": 1.0146,
+      "step": 3952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951574014574809e-05,
+      "loss": 0.958,
+      "step": 3953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515421479047588e-05,
+      "loss": 1.3496,
+      "step": 3954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515102710135856e-05,
+      "loss": 1.3467,
+      "step": 3955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514783839016328e-05,
+      "loss": 1.4629,
+      "step": 3956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951446486569242e-05,
+      "loss": 1.4277,
+      "step": 3957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514145790167566e-05,
+      "loss": 1.6094,
+      "step": 3958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513826612445194e-05,
+      "loss": 1.0684,
+      "step": 3959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513507332528724e-05,
+      "loss": 0.9204,
+      "step": 3960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513187950421595e-05,
+      "loss": 1.0566,
+      "step": 3961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512868466127232e-05,
+      "loss": 1.7129,
+      "step": 3962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512548879649067e-05,
+      "loss": 1.6523,
+      "step": 3963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512229190990536e-05,
+      "loss": 1.0283,
+      "step": 3964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511909400155068e-05,
+      "loss": 1.1221,
+      "step": 3965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511589507146105e-05,
+      "loss": 1.1953,
+      "step": 3966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951126951196708e-05,
+      "loss": 1.4746,
+      "step": 3967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510949414621425e-05,
+      "loss": 1.3604,
+      "step": 3968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510629215112584e-05,
+      "loss": 1.2812,
+      "step": 3969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510308913443997e-05,
+      "loss": 1.0293,
+      "step": 3970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509988509619102e-05,
+      "loss": 1.2646,
+      "step": 3971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509668003641344e-05,
+      "loss": 1.3027,
+      "step": 3972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509347395514162e-05,
+      "loss": 1.0,
+      "step": 3973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509026685241e-05,
+      "loss": 0.8057,
+      "step": 3974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950870587282531e-05,
+      "loss": 1.1499,
+      "step": 3975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508384958270526e-05,
+      "loss": 1.6191,
+      "step": 3976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508063941580106e-05,
+      "loss": 0.9482,
+      "step": 3977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507742822757486e-05,
+      "loss": 1.1045,
+      "step": 3978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950742160180613e-05,
+      "loss": 0.8828,
+      "step": 3979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950710027872948e-05,
+      "loss": 1.291,
+      "step": 3980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950677885353099e-05,
+      "loss": 1.5781,
+      "step": 3981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950645732621411e-05,
+      "loss": 1.2822,
+      "step": 3982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506135696782294e-05,
+      "loss": 1.3027,
+      "step": 3983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505813965239e-05,
+      "loss": 1.0508,
+      "step": 3984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505492131587686e-05,
+      "loss": 0.9297,
+      "step": 3985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95051701958318e-05,
+      "loss": 1.0176,
+      "step": 3986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504848157974807e-05,
+      "loss": 1.1904,
+      "step": 3987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504526018020166e-05,
+      "loss": 1.5059,
+      "step": 3988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504203775971332e-05,
+      "loss": 1.0117,
+      "step": 3989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503881431831772e-05,
+      "loss": 1.0991,
+      "step": 3990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503558985604948e-05,
+      "loss": 1.3496,
+      "step": 3991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950323643729432e-05,
+      "loss": 0.9082,
+      "step": 3992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502913786903357e-05,
+      "loss": 1.3545,
+      "step": 3993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502591034435525e-05,
+      "loss": 1.2109,
+      "step": 3994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502268179894288e-05,
+      "loss": 1.3936,
+      "step": 3995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950194522328311e-05,
+      "loss": 1.3848,
+      "step": 3996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501622164605468e-05,
+      "loss": 1.3184,
+      "step": 3997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501299003864828e-05,
+      "loss": 1.2412,
+      "step": 3998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500975741064663e-05,
+      "loss": 1.2832,
+      "step": 3999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500652376208443e-05,
+      "loss": 0.9229,
+      "step": 4000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500328909299647e-05,
+      "loss": 1.4414,
+      "step": 4001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950000534034174e-05,
+      "loss": 0.9951,
+      "step": 4002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949968166933821e-05,
+      "loss": 0.9863,
+      "step": 4003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499357896292522e-05,
+      "loss": 1.5078,
+      "step": 4004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949903402120816e-05,
+      "loss": 1.2686,
+      "step": 4005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498710044088605e-05,
+      "loss": 1.2402,
+      "step": 4006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949838596493733e-05,
+      "loss": 1.8945,
+      "step": 4007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949806178375782e-05,
+      "loss": 1.4844,
+      "step": 4008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949773750055356e-05,
+      "loss": 1.1572,
+      "step": 4009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497413115328028e-05,
+      "loss": 1.4355,
+      "step": 4010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497088628084713e-05,
+      "loss": 1.3086,
+      "step": 4011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496764038827097e-05,
+      "loss": 1.3438,
+      "step": 4012
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949643934755867e-05,
+      "loss": 1.2178,
+      "step": 4013
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496114554282917e-05,
+      "loss": 1.291,
+      "step": 4014
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949578965900333e-05,
+      "loss": 1.29,
+      "step": 4015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495464661723394e-05,
+      "loss": 1.416,
+      "step": 4016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94951395624466e-05,
+      "loss": 1.3398,
+      "step": 4017
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494814361176448e-05,
+      "loss": 0.7334,
+      "step": 4018
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494489057916424e-05,
+      "loss": 1.1953,
+      "step": 4019
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494163652670025e-05,
+      "loss": 1.4287,
+      "step": 4020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949383814544074e-05,
+      "loss": 1.2129,
+      "step": 4021
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949351253623208e-05,
+      "loss": 1.1846,
+      "step": 4022
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493186825047526e-05,
+      "loss": 1.457,
+      "step": 4023
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949286101189059e-05,
+      "loss": 1.167,
+      "step": 4024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492535096764765e-05,
+      "loss": 1.1768,
+      "step": 4025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492209079673553e-05,
+      "loss": 1.2188,
+      "step": 4026
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949188296062045e-05,
+      "loss": 1.457,
+      "step": 4027
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949155673960897e-05,
+      "loss": 1.0986,
+      "step": 4028
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491230416642614e-05,
+      "loss": 1.3955,
+      "step": 4029
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490903991724882e-05,
+      "loss": 1.3848,
+      "step": 4030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490577464859288e-05,
+      "loss": 1.1904,
+      "step": 4031
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949025083604933e-05,
+      "loss": 1.1279,
+      "step": 4032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489924105298525e-05,
+      "loss": 1.2285,
+      "step": 4033
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489597272610377e-05,
+      "loss": 1.7188,
+      "step": 4034
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94892703379884e-05,
+      "loss": 1.3984,
+      "step": 4035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488943301436105e-05,
+      "loss": 1.0562,
+      "step": 4036
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488616162957005e-05,
+      "loss": 1.0078,
+      "step": 4037
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488288922554613e-05,
+      "loss": 1.1064,
+      "step": 4038
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487961580232447e-05,
+      "loss": 1.3438,
+      "step": 4039
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948763413599402e-05,
+      "loss": 0.9395,
+      "step": 4040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948730658984285e-05,
+      "loss": 1.2744,
+      "step": 4041
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486978941782454e-05,
+      "loss": 1.3613,
+      "step": 4042
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486651191816354e-05,
+      "loss": 1.2607,
+      "step": 4043
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948632333994807e-05,
+      "loss": 1.4707,
+      "step": 4044
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485995386181124e-05,
+      "loss": 1.6484,
+      "step": 4045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485667330519038e-05,
+      "loss": 1.4297,
+      "step": 4046
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948533917296534e-05,
+      "loss": 1.5176,
+      "step": 4047
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485010913523544e-05,
+      "loss": 1.6562,
+      "step": 4048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484682552197184e-05,
+      "loss": 1.0752,
+      "step": 4049
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484354088989793e-05,
+      "loss": 0.998,
+      "step": 4050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484025523904884e-05,
+      "loss": 1.2227,
+      "step": 4051
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483696856946e-05,
+      "loss": 1.168,
+      "step": 4052
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483368088116664e-05,
+      "loss": 1.4258,
+      "step": 4053
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948303921742041e-05,
+      "loss": 1.3154,
+      "step": 4054
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948271024486077e-05,
+      "loss": 1.1748,
+      "step": 4055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482381170441277e-05,
+      "loss": 1.6855,
+      "step": 4056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948205199416547e-05,
+      "loss": 1.2119,
+      "step": 4057
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481722716036883e-05,
+      "loss": 1.3887,
+      "step": 4058
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481393336059048e-05,
+      "loss": 1.7344,
+      "step": 4059
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481063854235507e-05,
+      "loss": 1.1602,
+      "step": 4060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480734270569795e-05,
+      "loss": 0.9619,
+      "step": 4061
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480404585065465e-05,
+      "loss": 1.4883,
+      "step": 4062
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480074797726044e-05,
+      "loss": 1.3906,
+      "step": 4063
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947974490855508e-05,
+      "loss": 1.207,
+      "step": 4064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479414917556118e-05,
+      "loss": 1.1797,
+      "step": 4065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479084824732702e-05,
+      "loss": 1.3525,
+      "step": 4066
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478754630088373e-05,
+      "loss": 1.3252,
+      "step": 4067
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478424333626686e-05,
+      "loss": 1.3496,
+      "step": 4068
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478093935351182e-05,
+      "loss": 1.2832,
+      "step": 4069
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477763435265417e-05,
+      "loss": 1.0137,
+      "step": 4070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477432833372933e-05,
+      "loss": 1.5723,
+      "step": 4071
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477102129677284e-05,
+      "loss": 1.3555,
+      "step": 4072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476771324182027e-05,
+      "loss": 1.1836,
+      "step": 4073
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947644041689071e-05,
+      "loss": 1.0918,
+      "step": 4074
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947610940780689e-05,
+      "loss": 1.2109,
+      "step": 4075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947577829693412e-05,
+      "loss": 1.3838,
+      "step": 4076
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947544708427596e-05,
+      "loss": 1.1855,
+      "step": 4077
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475115769835967e-05,
+      "loss": 1.3486,
+      "step": 4078
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474784353617697e-05,
+      "loss": 1.0664,
+      "step": 4079
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947445283562471e-05,
+      "loss": 1.2363,
+      "step": 4080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474121215860574e-05,
+      "loss": 1.5859,
+      "step": 4081
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473789494328844e-05,
+      "loss": 1.291,
+      "step": 4082
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473457671033085e-05,
+      "loss": 1.333,
+      "step": 4083
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947312574597686e-05,
+      "loss": 1.5742,
+      "step": 4084
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472793719163744e-05,
+      "loss": 0.9531,
+      "step": 4085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947246159059729e-05,
+      "loss": 1.1836,
+      "step": 4086
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472129360281072e-05,
+      "loss": 1.415,
+      "step": 4087
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471797028218656e-05,
+      "loss": 1.2998,
+      "step": 4088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471464594413617e-05,
+      "loss": 1.2461,
+      "step": 4089
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947113205886952e-05,
+      "loss": 1.3901,
+      "step": 4090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470799421589942e-05,
+      "loss": 1.3213,
+      "step": 4091
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947046668257845e-05,
+      "loss": 1.1104,
+      "step": 4092
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470133841838628e-05,
+      "loss": 1.5,
+      "step": 4093
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946980089937404e-05,
+      "loss": 1.7656,
+      "step": 4094
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469467855188272e-05,
+      "loss": 1.1035,
+      "step": 4095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946913470928489e-05,
+      "loss": 1.1748,
+      "step": 4096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468801461667486e-05,
+      "loss": 1.7969,
+      "step": 4097
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468468112339633e-05,
+      "loss": 0.9648,
+      "step": 4098
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946813466130491e-05,
+      "loss": 1.6377,
+      "step": 4099
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467801108566896e-05,
+      "loss": 1.0811,
+      "step": 4100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467467454129186e-05,
+      "loss": 1.8008,
+      "step": 4101
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467133697995354e-05,
+      "loss": 0.8691,
+      "step": 4102
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466799840168987e-05,
+      "loss": 1.1641,
+      "step": 4103
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466465880653672e-05,
+      "loss": 1.2012,
+      "step": 4104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466131819452997e-05,
+      "loss": 1.3809,
+      "step": 4105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465797656570546e-05,
+      "loss": 0.9795,
+      "step": 4106
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465463392009915e-05,
+      "loss": 1.127,
+      "step": 4107
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946512902577469e-05,
+      "loss": 1.4531,
+      "step": 4108
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464794557868466e-05,
+      "loss": 1.1924,
+      "step": 4109
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464459988294835e-05,
+      "loss": 1.3535,
+      "step": 4110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946412531705739e-05,
+      "loss": 1.3789,
+      "step": 4111
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463790544159716e-05,
+      "loss": 1.3369,
+      "step": 4112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463455669605428e-05,
+      "loss": 1.4082,
+      "step": 4113
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463120693398113e-05,
+      "loss": 1.1729,
+      "step": 4114
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462785615541365e-05,
+      "loss": 1.2939,
+      "step": 4115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462450436038794e-05,
+      "loss": 1.0781,
+      "step": 4116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462115154893993e-05,
+      "loss": 1.3906,
+      "step": 4117
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461779772110562e-05,
+      "loss": 1.3525,
+      "step": 4118
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461444287692108e-05,
+      "loss": 1.3691,
+      "step": 4119
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461108701642236e-05,
+      "loss": 1.2012,
+      "step": 4120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460773013964546e-05,
+      "loss": 1.0693,
+      "step": 4121
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460437224662644e-05,
+      "loss": 1.4121,
+      "step": 4122
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946010133374014e-05,
+      "loss": 1.1523,
+      "step": 4123
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459765341200643e-05,
+      "loss": 1.4355,
+      "step": 4124
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459429247047755e-05,
+      "loss": 1.6953,
+      "step": 4125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459093051285096e-05,
+      "loss": 1.1816,
+      "step": 4126
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945875675391627e-05,
+      "loss": 1.2539,
+      "step": 4127
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458420354944892e-05,
+      "loss": 1.042,
+      "step": 4128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458083854374578e-05,
+      "loss": 1.3105,
+      "step": 4129
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457747252208936e-05,
+      "loss": 0.9707,
+      "step": 4130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945741054845159e-05,
+      "loss": 1.4727,
+      "step": 4131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457073743106145e-05,
+      "loss": 1.1274,
+      "step": 4132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945673683617623e-05,
+      "loss": 1.4121,
+      "step": 4133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945639982766546e-05,
+      "loss": 1.0664,
+      "step": 4134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9456062717577454e-05,
+      "loss": 1.2539,
+      "step": 4135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455725505915834e-05,
+      "loss": 1.2949,
+      "step": 4136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455388192684224e-05,
+      "loss": 1.0654,
+      "step": 4137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455050777886245e-05,
+      "loss": 1.1719,
+      "step": 4138
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454713261525523e-05,
+      "loss": 0.8843,
+      "step": 4139
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454375643605682e-05,
+      "loss": 1.2568,
+      "step": 4140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454037924130345e-05,
+      "loss": 1.3105,
+      "step": 4141
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945370010310315e-05,
+      "loss": 1.7832,
+      "step": 4142
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453362180527717e-05,
+      "loss": 1.1621,
+      "step": 4143
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453024156407677e-05,
+      "loss": 1.2656,
+      "step": 4144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452686030746665e-05,
+      "loss": 1.3896,
+      "step": 4145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452347803548307e-05,
+      "loss": 1.3809,
+      "step": 4146
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452009474816242e-05,
+      "loss": 1.291,
+      "step": 4147
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94516710445541e-05,
+      "loss": 1.293,
+      "step": 4148
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945133251276552e-05,
+      "loss": 1.4199,
+      "step": 4149
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945099387945413e-05,
+      "loss": 1.4121,
+      "step": 4150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945065514462358e-05,
+      "loss": 1.1807,
+      "step": 4151
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450316308277502e-05,
+      "loss": 1.5039,
+      "step": 4152
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944997737041953e-05,
+      "loss": 1.2207,
+      "step": 4153
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944963833105332e-05,
+      "loss": 1.2012,
+      "step": 4154
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94492991901825e-05,
+      "loss": 0.9824,
+      "step": 4155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448959947810716e-05,
+      "loss": 1.3955,
+      "step": 4156
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448620603941613e-05,
+      "loss": 1.252,
+      "step": 4157
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944828115857884e-05,
+      "loss": 1.4805,
+      "step": 4158
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944794161172604e-05,
+      "loss": 1.2646,
+      "step": 4159
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447601963386857e-05,
+      "loss": 1.2031,
+      "step": 4160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447262213564943e-05,
+      "loss": 1.0293,
+      "step": 4161
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944692236226395e-05,
+      "loss": 0.8242,
+      "step": 4162
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446582409487523e-05,
+      "loss": 1.2637,
+      "step": 4163
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446242355239315e-05,
+      "loss": 1.0029,
+      "step": 4164
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944590219952298e-05,
+      "loss": 1.1182,
+      "step": 4165
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445561942342174e-05,
+      "loss": 1.2871,
+      "step": 4166
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944522158370055e-05,
+      "loss": 1.2754,
+      "step": 4167
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444881123601764e-05,
+      "loss": 1.4697,
+      "step": 4168
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944454056204947e-05,
+      "loss": 1.2275,
+      "step": 4169
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944419989904733e-05,
+      "loss": 1.0332,
+      "step": 4170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443859134599004e-05,
+      "loss": 1.2949,
+      "step": 4171
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944351826870815e-05,
+      "loss": 1.2432,
+      "step": 4172
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443177301378427e-05,
+      "loss": 1.1025,
+      "step": 4173
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94428362326135e-05,
+      "loss": 0.8159,
+      "step": 4174
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944249506241704e-05,
+      "loss": 1.5723,
+      "step": 4175
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94421537907927e-05,
+      "loss": 1.4062,
+      "step": 4176
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944181241774415e-05,
+      "loss": 1.3232,
+      "step": 4177
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944147094327506e-05,
+      "loss": 1.0088,
+      "step": 4178
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441129367389094e-05,
+      "loss": 1.5449,
+      "step": 4179
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440787690089926e-05,
+      "loss": 1.2812,
+      "step": 4180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944044591138122e-05,
+      "loss": 1.168,
+      "step": 4181
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944010403126665e-05,
+      "loss": 1.457,
+      "step": 4182
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943976204974989e-05,
+      "loss": 1.8535,
+      "step": 4183
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943941996683461e-05,
+      "loss": 1.2295,
+      "step": 4184
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943907778252449e-05,
+      "loss": 0.7812,
+      "step": 4185
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94387354968232e-05,
+      "loss": 1.1123,
+      "step": 4186
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943839310973442e-05,
+      "loss": 1.4043,
+      "step": 4187
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438050621261824e-05,
+      "loss": 1.2588,
+      "step": 4188
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437708031409097e-05,
+      "loss": 1.5098,
+      "step": 4189
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437365340179913e-05,
+      "loss": 1.2354,
+      "step": 4190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437022547577954e-05,
+      "loss": 1.0918,
+      "step": 4191
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436679653606908e-05,
+      "loss": 1.167,
+      "step": 4192
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436336658270453e-05,
+      "loss": 1.0889,
+      "step": 4193
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435993561572272e-05,
+      "loss": 1.1104,
+      "step": 4194
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435650363516055e-05,
+      "loss": 1.4795,
+      "step": 4195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435307064105486e-05,
+      "loss": 1.1191,
+      "step": 4196
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434963663344252e-05,
+      "loss": 1.3213,
+      "step": 4197
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434620161236045e-05,
+      "loss": 1.168,
+      "step": 4198
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943427655778455e-05,
+      "loss": 1.5312,
+      "step": 4199
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433932852993462e-05,
+      "loss": 1.6992,
+      "step": 4200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433589046866465e-05,
+      "loss": 1.2744,
+      "step": 4201
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433245139407266e-05,
+      "loss": 0.9702,
+      "step": 4202
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943290113061955e-05,
+      "loss": 1.0674,
+      "step": 4203
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432557020507006e-05,
+      "loss": 1.1768,
+      "step": 4204
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432212809073345e-05,
+      "loss": 1.3057,
+      "step": 4205
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9431868496322254e-05,
+      "loss": 1.3145,
+      "step": 4206
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943152408225744e-05,
+      "loss": 1.0615,
+      "step": 4207
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943117956688259e-05,
+      "loss": 1.373,
+      "step": 4208
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430834950201418e-05,
+      "loss": 1.1006,
+      "step": 4209
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943049023221762e-05,
+      "loss": 0.9922,
+      "step": 4210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430145412934893e-05,
+      "loss": 1.2881,
+      "step": 4211
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942980049235695e-05,
+      "loss": 1.4912,
+      "step": 4212
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942945547048749e-05,
+      "loss": 1.208,
+      "step": 4213
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429110347330225e-05,
+      "loss": 1.0723,
+      "step": 4214
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428765122888858e-05,
+      "loss": 1.4824,
+      "step": 4215
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428419797167095e-05,
+      "loss": 1.4492,
+      "step": 4216
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942807437016865e-05,
+      "loss": 1.0713,
+      "step": 4217
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9427728841897234e-05,
+      "loss": 1.1953,
+      "step": 4218
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9427383212356554e-05,
+      "loss": 0.876,
+      "step": 4219
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9427037481550326e-05,
+      "loss": 2.0859,
+      "step": 4220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426691649482262e-05,
+      "loss": 1.4473,
+      "step": 4221
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426345716156078e-05,
+      "loss": 1.1475,
+      "step": 4222
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942599968157549e-05,
+      "loss": 0.9873,
+      "step": 4223
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425653545744213e-05,
+      "loss": 1.4072,
+      "step": 4224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942530730866597e-05,
+      "loss": 1.0186,
+      "step": 4225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942496097034447e-05,
+      "loss": 1.4287,
+      "step": 4226
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424614530783444e-05,
+      "loss": 1.4141,
+      "step": 4227
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942426798998661e-05,
+      "loss": 1.1104,
+      "step": 4228
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942392134795769e-05,
+      "loss": 1.373,
+      "step": 4229
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423574604700403e-05,
+      "loss": 1.3213,
+      "step": 4230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942322776021848e-05,
+      "loss": 1.3828,
+      "step": 4231
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422880814515646e-05,
+      "loss": 1.1826,
+      "step": 4232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422533767595625e-05,
+      "loss": 1.1387,
+      "step": 4233
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422186619462144e-05,
+      "loss": 1.3838,
+      "step": 4234
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421839370118935e-05,
+      "loss": 1.1689,
+      "step": 4235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942149201956973e-05,
+      "loss": 1.1797,
+      "step": 4236
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421144567818257e-05,
+      "loss": 1.209,
+      "step": 4237
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9420797014868244e-05,
+      "loss": 1.0107,
+      "step": 4238
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942044936072343e-05,
+      "loss": 1.1113,
+      "step": 4239
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942010160538755e-05,
+      "loss": 1.6602,
+      "step": 4240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419753748864334e-05,
+      "loss": 1.1865,
+      "step": 4241
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419405791157527e-05,
+      "loss": 1.166,
+      "step": 4242
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419057732270856e-05,
+      "loss": 1.6699,
+      "step": 4243
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418709572208068e-05,
+      "loss": 1.1592,
+      "step": 4244
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94183613109729e-05,
+      "loss": 1.2695,
+      "step": 4245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418012948569093e-05,
+      "loss": 1.377,
+      "step": 4246
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941766448500039e-05,
+      "loss": 1.1899,
+      "step": 4247
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417315920270532e-05,
+      "loss": 1.2607,
+      "step": 4248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416967254383263e-05,
+      "loss": 1.707,
+      "step": 4249
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416618487342333e-05,
+      "loss": 1.1445,
+      "step": 4250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941626961915148e-05,
+      "loss": 1.1309,
+      "step": 4251
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941592064981446e-05,
+      "loss": 1.4395,
+      "step": 4252
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415571579335016e-05,
+      "loss": 1.1074,
+      "step": 4253
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94152224077169e-05,
+      "loss": 1.3281,
+      "step": 4254
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414873134963864e-05,
+      "loss": 1.2676,
+      "step": 4255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414523761079654e-05,
+      "loss": 0.8643,
+      "step": 4256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414174286068027e-05,
+      "loss": 1.3643,
+      "step": 4257
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941382470993274e-05,
+      "loss": 1.0332,
+      "step": 4258
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413475032677543e-05,
+      "loss": 1.3564,
+      "step": 4259
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941312525430619e-05,
+      "loss": 1.1611,
+      "step": 4260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412775374822447e-05,
+      "loss": 1.2207,
+      "step": 4261
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412425394230065e-05,
+      "loss": 1.2305,
+      "step": 4262
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412075312532804e-05,
+      "loss": 1.0488,
+      "step": 4263
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411725129734427e-05,
+      "loss": 1.6934,
+      "step": 4264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411374845838696e-05,
+      "loss": 1.4473,
+      "step": 4265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941102446084937e-05,
+      "loss": 1.6982,
+      "step": 4266
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410673974770215e-05,
+      "loss": 1.0547,
+      "step": 4267
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410323387604996e-05,
+      "loss": 1.1201,
+      "step": 4268
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940997269935748e-05,
+      "loss": 1.3984,
+      "step": 4269
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940962191003143e-05,
+      "loss": 1.1348,
+      "step": 4270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940927101963062e-05,
+      "loss": 0.874,
+      "step": 4271
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408920028158812e-05,
+      "loss": 1.1729,
+      "step": 4272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940856893561978e-05,
+      "loss": 1.1719,
+      "step": 4273
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94082177420173e-05,
+      "loss": 0.9199,
+      "step": 4274
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407866447355136e-05,
+      "loss": 1.0479,
+      "step": 4275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940751505163706e-05,
+      "loss": 1.332,
+      "step": 4276
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940716355486686e-05,
+      "loss": 1.1006,
+      "step": 4277
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94068119570483e-05,
+      "loss": 1.1309,
+      "step": 4278
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940646025818516e-05,
+      "loss": 1.3594,
+      "step": 4279
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940610845828122e-05,
+      "loss": 1.0127,
+      "step": 4280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405756557340255e-05,
+      "loss": 1.5039,
+      "step": 4281
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405404555366046e-05,
+      "loss": 1.1875,
+      "step": 4282
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405052452362374e-05,
+      "loss": 1.4355,
+      "step": 4283
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404700248333025e-05,
+      "loss": 1.2656,
+      "step": 4284
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404347943281777e-05,
+      "loss": 1.1924,
+      "step": 4285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403995537212418e-05,
+      "loss": 1.5195,
+      "step": 4286
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940364303012873e-05,
+      "loss": 1.0474,
+      "step": 4287
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403290422034502e-05,
+      "loss": 1.1289,
+      "step": 4288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940293771293352e-05,
+      "loss": 1.5879,
+      "step": 4289
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402584902829575e-05,
+      "loss": 1.0405,
+      "step": 4290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402231991726456e-05,
+      "loss": 1.7227,
+      "step": 4291
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401878979627954e-05,
+      "loss": 1.084,
+      "step": 4292
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401525866537862e-05,
+      "loss": 1.2256,
+      "step": 4293
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940117265245997e-05,
+      "loss": 1.1406,
+      "step": 4294
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940081933739807e-05,
+      "loss": 1.0146,
+      "step": 4295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940046592135596e-05,
+      "loss": 0.9585,
+      "step": 4296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400112404337442e-05,
+      "loss": 1.6855,
+      "step": 4297
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399758786346305e-05,
+      "loss": 1.418,
+      "step": 4298
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399405067386353e-05,
+      "loss": 1.1104,
+      "step": 4299
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939905124746138e-05,
+      "loss": 1.041,
+      "step": 4300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398697326575192e-05,
+      "loss": 1.0645,
+      "step": 4301
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398343304731586e-05,
+      "loss": 1.1973,
+      "step": 4302
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939798918193437e-05,
+      "loss": 1.418,
+      "step": 4303
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397634958187345e-05,
+      "loss": 1.4121,
+      "step": 4304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939728063349431e-05,
+      "loss": 1.3027,
+      "step": 4305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396926207859085e-05,
+      "loss": 1.3711,
+      "step": 4306
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396571681285466e-05,
+      "loss": 1.0342,
+      "step": 4307
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396217053777264e-05,
+      "loss": 1.1797,
+      "step": 4308
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395862325338292e-05,
+      "loss": 1.5312,
+      "step": 4309
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395507495972353e-05,
+      "loss": 1.2822,
+      "step": 4310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395152565683265e-05,
+      "loss": 1.7617,
+      "step": 4311
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394797534474836e-05,
+      "loss": 1.2285,
+      "step": 4312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394442402350882e-05,
+      "loss": 0.9902,
+      "step": 4313
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394087169315217e-05,
+      "loss": 1.0435,
+      "step": 4314
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939373183537166e-05,
+      "loss": 1.1328,
+      "step": 4315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939337640052402e-05,
+      "loss": 0.9189,
+      "step": 4316
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393020864776123e-05,
+      "loss": 1.3662,
+      "step": 4317
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392665228131786e-05,
+      "loss": 1.0371,
+      "step": 4318
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392309490594824e-05,
+      "loss": 1.1387,
+      "step": 4319
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391953652169065e-05,
+      "loss": 1.6406,
+      "step": 4320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391597712858328e-05,
+      "loss": 1.5508,
+      "step": 4321
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391241672666438e-05,
+      "loss": 0.9873,
+      "step": 4322
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939088553159722e-05,
+      "loss": 1.6602,
+      "step": 4323
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390529289654493e-05,
+      "loss": 1.2803,
+      "step": 4324
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390172946842095e-05,
+      "loss": 1.3701,
+      "step": 4325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938981650316384e-05,
+      "loss": 1.5293,
+      "step": 4326
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389459958623567e-05,
+      "loss": 1.2861,
+      "step": 4327
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389103313225105e-05,
+      "loss": 1.0264,
+      "step": 4328
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938874656697228e-05,
+      "loss": 1.4463,
+      "step": 4329
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938838971986893e-05,
+      "loss": 1.208,
+      "step": 4330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388032771918884e-05,
+      "loss": 1.3711,
+      "step": 4331
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387675723125975e-05,
+      "loss": 1.5537,
+      "step": 4332
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387318573494044e-05,
+      "loss": 1.1523,
+      "step": 4333
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938696132302692e-05,
+      "loss": 1.498,
+      "step": 4334
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386603971728445e-05,
+      "loss": 1.6172,
+      "step": 4335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938624651960246e-05,
+      "loss": 1.6162,
+      "step": 4336
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93858889666528e-05,
+      "loss": 1.6738,
+      "step": 4337
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938553131288331e-05,
+      "loss": 1.2246,
+      "step": 4338
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938517355829782e-05,
+      "loss": 1.0967,
+      "step": 4339
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384815702900193e-05,
+      "loss": 1.1895,
+      "step": 4340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384457746694254e-05,
+      "loss": 1.2168,
+      "step": 4341
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938409968968386e-05,
+      "loss": 0.8589,
+      "step": 4342
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383741531872853e-05,
+      "loss": 1.1035,
+      "step": 4343
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938338327326508e-05,
+      "loss": 1.2891,
+      "step": 4344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383024913864385e-05,
+      "loss": 1.1211,
+      "step": 4345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9382666453674625e-05,
+      "loss": 1.209,
+      "step": 4346
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938230789269965e-05,
+      "loss": 1.4297,
+      "step": 4347
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381949230943305e-05,
+      "loss": 1.0615,
+      "step": 4348
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381590468409447e-05,
+      "loss": 1.3047,
+      "step": 4349
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381231605101928e-05,
+      "loss": 1.625,
+      "step": 4350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938087264102461e-05,
+      "loss": 1.2871,
+      "step": 4351
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380513576181336e-05,
+      "loss": 1.2148,
+      "step": 4352
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380154410575974e-05,
+      "loss": 1.5312,
+      "step": 4353
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379795144212374e-05,
+      "loss": 1.0088,
+      "step": 4354
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93794357770944e-05,
+      "loss": 0.9209,
+      "step": 4355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937907630922591e-05,
+      "loss": 1.0479,
+      "step": 4356
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937871674061077e-05,
+      "loss": 1.5439,
+      "step": 4357
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937835707125284e-05,
+      "loss": 1.8828,
+      "step": 4358
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937799730115598e-05,
+      "loss": 1.3467,
+      "step": 4359
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377637430324054e-05,
+      "loss": 1.0586,
+      "step": 4360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377277458760934e-05,
+      "loss": 1.0908,
+      "step": 4361
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937691738647048e-05,
+      "loss": 1.0752,
+      "step": 4362
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937655721345656e-05,
+      "loss": 1.2197,
+      "step": 4363
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376196939723053e-05,
+      "loss": 1.3359,
+      "step": 4364
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375836565273814e-05,
+      "loss": 1.2178,
+      "step": 4365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375476090112727e-05,
+      "loss": 1.1777,
+      "step": 4366
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375115514243658e-05,
+      "loss": 1.2383,
+      "step": 4367
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937475483767048e-05,
+      "loss": 1.5254,
+      "step": 4368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374394060397067e-05,
+      "loss": 1.3662,
+      "step": 4369
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374033182427297e-05,
+      "loss": 1.4619,
+      "step": 4370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373672203765045e-05,
+      "loss": 1.374,
+      "step": 4371
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373311124414184e-05,
+      "loss": 1.2539,
+      "step": 4372
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93729499443786e-05,
+      "loss": 1.5117,
+      "step": 4373
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937258866366217e-05,
+      "loss": 1.4619,
+      "step": 4374
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372227282268774e-05,
+      "loss": 1.377,
+      "step": 4375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371865800202296e-05,
+      "loss": 1.5,
+      "step": 4376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371504217466612e-05,
+      "loss": 1.167,
+      "step": 4377
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371142534065613e-05,
+      "loss": 0.9062,
+      "step": 4378
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370780750003185e-05,
+      "loss": 1.0244,
+      "step": 4379
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937041886528321e-05,
+      "loss": 1.4697,
+      "step": 4380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370056879909575e-05,
+      "loss": 1.2402,
+      "step": 4381
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936969479388617e-05,
+      "loss": 1.4238,
+      "step": 4382
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369332607216885e-05,
+      "loss": 1.2695,
+      "step": 4383
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936897031990561e-05,
+      "loss": 1.4521,
+      "step": 4384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368607931956232e-05,
+      "loss": 1.7383,
+      "step": 4385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368245443372652e-05,
+      "loss": 1.2998,
+      "step": 4386
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367882854158763e-05,
+      "loss": 1.2793,
+      "step": 4387
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936752016431845e-05,
+      "loss": 1.2988,
+      "step": 4388
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936715737385562e-05,
+      "loss": 1.2178,
+      "step": 4389
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936679448277416e-05,
+      "loss": 1.1094,
+      "step": 4390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366431491077976e-05,
+      "loss": 1.1113,
+      "step": 4391
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936606839877097e-05,
+      "loss": 1.3516,
+      "step": 4392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936570520585703e-05,
+      "loss": 1.4004,
+      "step": 4393
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936534191234006e-05,
+      "loss": 1.2217,
+      "step": 4394
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364978518223974e-05,
+      "loss": 1.2178,
+      "step": 4395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364615023512664e-05,
+      "loss": 1.3359,
+      "step": 4396
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936425142821004e-05,
+      "loss": 0.9404,
+      "step": 4397
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363887732320008e-05,
+      "loss": 1.3711,
+      "step": 4398
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363523935846467e-05,
+      "loss": 1.0654,
+      "step": 4399
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363160038793335e-05,
+      "loss": 1.2129,
+      "step": 4400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362796041164513e-05,
+      "loss": 1.4043,
+      "step": 4401
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362431942963915e-05,
+      "loss": 1.3359,
+      "step": 4402
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936206774419545e-05,
+      "loss": 1.4053,
+      "step": 4403
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361703444863035e-05,
+      "loss": 0.9893,
+      "step": 4404
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936133904497058e-05,
+      "loss": 1.2695,
+      "step": 4405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360974544521993e-05,
+      "loss": 0.9258,
+      "step": 4406
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360609943521198e-05,
+      "loss": 1.5449,
+      "step": 4407
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360245241972107e-05,
+      "loss": 1.1895,
+      "step": 4408
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935988043987864e-05,
+      "loss": 1.2783,
+      "step": 4409
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359515537244715e-05,
+      "loss": 1.2676,
+      "step": 4410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935915053407425e-05,
+      "loss": 1.3311,
+      "step": 4411
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358785430371166e-05,
+      "loss": 1.2549,
+      "step": 4412
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358420226139388e-05,
+      "loss": 1.418,
+      "step": 4413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358054921382834e-05,
+      "loss": 0.9023,
+      "step": 4414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357689516105434e-05,
+      "loss": 1.4111,
+      "step": 4415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357324010311106e-05,
+      "loss": 1.2031,
+      "step": 4416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356958404003782e-05,
+      "loss": 0.9648,
+      "step": 4417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356592697187383e-05,
+      "loss": 0.9805,
+      "step": 4418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935622688986585e-05,
+      "loss": 1.2324,
+      "step": 4419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355860982043097e-05,
+      "loss": 0.8667,
+      "step": 4420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935549497372306e-05,
+      "loss": 1.2686,
+      "step": 4421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355128864909673e-05,
+      "loss": 1.3555,
+      "step": 4422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354762655606868e-05,
+      "loss": 1.2324,
+      "step": 4423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354396345818576e-05,
+      "loss": 1.1367,
+      "step": 4424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354029935548735e-05,
+      "loss": 1.0249,
+      "step": 4425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935366342480128e-05,
+      "loss": 1.377,
+      "step": 4426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353296813580146e-05,
+      "loss": 1.3906,
+      "step": 4427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352930101889272e-05,
+      "loss": 0.8408,
+      "step": 4428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93525632897326e-05,
+      "loss": 1.207,
+      "step": 4429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352196377114064e-05,
+      "loss": 1.0537,
+      "step": 4430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351829364037613e-05,
+      "loss": 1.3926,
+      "step": 4431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351462250507184e-05,
+      "loss": 1.0801,
+      "step": 4432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935109503652672e-05,
+      "loss": 1.3164,
+      "step": 4433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935072772210017e-05,
+      "loss": 1.4658,
+      "step": 4434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935036030723147e-05,
+      "loss": 1.0957,
+      "step": 4435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934999279192458e-05,
+      "loss": 1.1973,
+      "step": 4436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934962517618344e-05,
+      "loss": 1.5566,
+      "step": 4437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349257460012e-05,
+      "loss": 1.0693,
+      "step": 4438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348889643414206e-05,
+      "loss": 1.5117,
+      "step": 4439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348521726394012e-05,
+      "loss": 1.1582,
+      "step": 4440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348153708955376e-05,
+      "loss": 1.1953,
+      "step": 4441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347785591102244e-05,
+      "loss": 1.0176,
+      "step": 4442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934741737283857e-05,
+      "loss": 1.2266,
+      "step": 4443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347049054168313e-05,
+      "loss": 1.6904,
+      "step": 4444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934668063509543e-05,
+      "loss": 1.0,
+      "step": 4445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346312115623867e-05,
+      "loss": 1.1973,
+      "step": 4446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93459434957576e-05,
+      "loss": 1.1758,
+      "step": 4447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345574775500577e-05,
+      "loss": 1.2129,
+      "step": 4448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934520595485676e-05,
+      "loss": 0.6626,
+      "step": 4449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344837033830116e-05,
+      "loss": 1.0557,
+      "step": 4450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344468012424605e-05,
+      "loss": 1.625,
+      "step": 4451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344098890644188e-05,
+      "loss": 1.3691,
+      "step": 4452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343729668492834e-05,
+      "loss": 1.3613,
+      "step": 4453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343360345974506e-05,
+      "loss": 1.5586,
+      "step": 4454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934299092309317e-05,
+      "loss": 1.1816,
+      "step": 4455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342621399852796e-05,
+      "loss": 1.0798,
+      "step": 4456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342251776257356e-05,
+      "loss": 0.9033,
+      "step": 4457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934188205231082e-05,
+      "loss": 1.2559,
+      "step": 4458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341512228017153e-05,
+      "loss": 1.3887,
+      "step": 4459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934114230338034e-05,
+      "loss": 1.4189,
+      "step": 4460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934077227840434e-05,
+      "loss": 1.0225,
+      "step": 4461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340402153093136e-05,
+      "loss": 0.9297,
+      "step": 4462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340031927450702e-05,
+      "loss": 1.1494,
+      "step": 4463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339661601481014e-05,
+      "loss": 1.2812,
+      "step": 4464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339291175188052e-05,
+      "loss": 0.8999,
+      "step": 4465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338920648575798e-05,
+      "loss": 1.1768,
+      "step": 4466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338550021648222e-05,
+      "loss": 1.5684,
+      "step": 4467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338179294409313e-05,
+      "loss": 1.0527,
+      "step": 4468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933780846686305e-05,
+      "loss": 1.3984,
+      "step": 4469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933743753901342e-05,
+      "loss": 1.0654,
+      "step": 4470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337066510864404e-05,
+      "loss": 1.1045,
+      "step": 4471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336695382419987e-05,
+      "loss": 0.9619,
+      "step": 4472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933632415368416e-05,
+      "loss": 1.3379,
+      "step": 4473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335952824660908e-05,
+      "loss": 1.2969,
+      "step": 4474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335581395354217e-05,
+      "loss": 1.5176,
+      "step": 4475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335209865768078e-05,
+      "loss": 1.5312,
+      "step": 4476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334838235906486e-05,
+      "loss": 0.9893,
+      "step": 4477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334466505773425e-05,
+      "loss": 1.1611,
+      "step": 4478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334094675372896e-05,
+      "loss": 1.2344,
+      "step": 4479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933372274470889e-05,
+      "loss": 1.2842,
+      "step": 4480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93333507137854e-05,
+      "loss": 1.1377,
+      "step": 4481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332978582606428e-05,
+      "loss": 1.1953,
+      "step": 4482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332606351175965e-05,
+      "loss": 1.2031,
+      "step": 4483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933223401949801e-05,
+      "loss": 1.6816,
+      "step": 4484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331861587576568e-05,
+      "loss": 1.2812,
+      "step": 4485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331489055415632e-05,
+      "loss": 1.4824,
+      "step": 4486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933111642301921e-05,
+      "loss": 1.2109,
+      "step": 4487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330743690391302e-05,
+      "loss": 1.0117,
+      "step": 4488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933037085753591e-05,
+      "loss": 0.8457,
+      "step": 4489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329997924457046e-05,
+      "loss": 1.4404,
+      "step": 4490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329624891158706e-05,
+      "loss": 1.1699,
+      "step": 4491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329251757644904e-05,
+      "loss": 1.2891,
+      "step": 4492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328878523919643e-05,
+      "loss": 1.1299,
+      "step": 4493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328505189986935e-05,
+      "loss": 1.1377,
+      "step": 4494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328131755850793e-05,
+      "loss": 1.4102,
+      "step": 4495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327758221515223e-05,
+      "loss": 1.3174,
+      "step": 4496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327384586984237e-05,
+      "loss": 1.2646,
+      "step": 4497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327010852261855e-05,
+      "loss": 1.041,
+      "step": 4498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932663701735209e-05,
+      "loss": 1.4082,
+      "step": 4499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932626308225895e-05,
+      "loss": 1.04,
+      "step": 4500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932588904698646e-05,
+      "loss": 1.4092,
+      "step": 4501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325514911538634e-05,
+      "loss": 1.2568,
+      "step": 4502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325140675919493e-05,
+      "loss": 1.0762,
+      "step": 4503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324766340133055e-05,
+      "loss": 1.2432,
+      "step": 4504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324391904183337e-05,
+      "loss": 1.2129,
+      "step": 4505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932401736807437e-05,
+      "loss": 0.916,
+      "step": 4506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323642731810172e-05,
+      "loss": 1.1377,
+      "step": 4507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323267995394768e-05,
+      "loss": 1.4707,
+      "step": 4508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322893158832185e-05,
+      "loss": 1.3862,
+      "step": 4509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322518222126445e-05,
+      "loss": 0.9092,
+      "step": 4510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322143185281576e-05,
+      "loss": 1.1514,
+      "step": 4511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932176804830161e-05,
+      "loss": 1.085,
+      "step": 4512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321392811190574e-05,
+      "loss": 0.8286,
+      "step": 4513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93210174739525e-05,
+      "loss": 1.0557,
+      "step": 4514
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932064203659142e-05,
+      "loss": 1.0088,
+      "step": 4515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320266499111368e-05,
+      "loss": 1.4326,
+      "step": 4516
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319890861516374e-05,
+      "loss": 1.1133,
+      "step": 4517
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931951512381048e-05,
+      "loss": 1.251,
+      "step": 4518
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931913928599771e-05,
+      "loss": 1.2148,
+      "step": 4519
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931876334808211e-05,
+      "loss": 0.8623,
+      "step": 4520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318387310067716e-05,
+      "loss": 0.8516,
+      "step": 4521
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931801117195857e-05,
+      "loss": 1.1792,
+      "step": 4522
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317634933758712e-05,
+      "loss": 1.2715,
+      "step": 4523
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317258595472175e-05,
+      "loss": 1.3145,
+      "step": 4524
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931688215710301e-05,
+      "loss": 1.5332,
+      "step": 4525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931650561865526e-05,
+      "loss": 1.6045,
+      "step": 4526
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931612898013297e-05,
+      "loss": 1.2344,
+      "step": 4527
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931575224154018e-05,
+      "loss": 1.2065,
+      "step": 4528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315375402880945e-05,
+      "loss": 0.98,
+      "step": 4529
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9314998464159305e-05,
+      "loss": 1.4102,
+      "step": 4530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9314621425379313e-05,
+      "loss": 1.1514,
+      "step": 4531
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931424428654502e-05,
+      "loss": 1.459,
+      "step": 4532
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931386704766047e-05,
+      "loss": 1.188,
+      "step": 4533
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931348970872973e-05,
+      "loss": 0.8486,
+      "step": 4534
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931311226975684e-05,
+      "loss": 1.082,
+      "step": 4535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312734730745856e-05,
+      "loss": 1.5996,
+      "step": 4536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312357091700838e-05,
+      "loss": 0.9609,
+      "step": 4537
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311979352625837e-05,
+      "loss": 1.1631,
+      "step": 4538
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311601513524915e-05,
+      "loss": 1.2598,
+      "step": 4539
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931122357440213e-05,
+      "loss": 0.9424,
+      "step": 4540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931084553526154e-05,
+      "loss": 1.5371,
+      "step": 4541
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931046739610721e-05,
+      "loss": 1.3184,
+      "step": 4542
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310089156943194e-05,
+      "loss": 1.1387,
+      "step": 4543
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9309710817773563e-05,
+      "loss": 1.5703,
+      "step": 4544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9309332378602375e-05,
+      "loss": 1.2061,
+      "step": 4545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.93089538394337e-05,
+      "loss": 1.4375,
+      "step": 4546
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308575200271598e-05,
+      "loss": 1.2539,
+      "step": 4547
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308196461120144e-05,
+      "loss": 1.2715,
+      "step": 4548
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307817621983396e-05,
+      "loss": 1.1279,
+      "step": 4549
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307438682865434e-05,
+      "loss": 0.877,
+      "step": 4550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307059643770324e-05,
+      "loss": 1.1221,
+      "step": 4551
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306680504702137e-05,
+      "loss": 1.1621,
+      "step": 4552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306301265664946e-05,
+      "loss": 1.4219,
+      "step": 4553
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930592192666282e-05,
+      "loss": 0.9951,
+      "step": 4554
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9305542487699848e-05,
+      "loss": 1.5879,
+      "step": 4555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930516294878009e-05,
+      "loss": 1.6602,
+      "step": 4556
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304783309907626e-05,
+      "loss": 1.3027,
+      "step": 4557
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304403571086543e-05,
+      "loss": 0.9912,
+      "step": 4558
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930402373232091e-05,
+      "loss": 1.0752,
+      "step": 4559
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930364379361481e-05,
+      "loss": 1.1816,
+      "step": 4560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930326375497233e-05,
+      "loss": 1.0547,
+      "step": 4561
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302883616397546e-05,
+      "loss": 1.333,
+      "step": 4562
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930250337789454e-05,
+      "loss": 1.2803,
+      "step": 4563
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.93021230394674e-05,
+      "loss": 1.3047,
+      "step": 4564
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930174260112021e-05,
+      "loss": 0.7886,
+      "step": 4565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301362062857062e-05,
+      "loss": 1.5918,
+      "step": 4566
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300981424682033e-05,
+      "loss": 1.6133,
+      "step": 4567
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930060068659922e-05,
+      "loss": 1.3613,
+      "step": 4568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300219848612712e-05,
+      "loss": 1.2148,
+      "step": 4569
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299838910726594e-05,
+      "loss": 1.3301,
+      "step": 4570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299457872944965e-05,
+      "loss": 1.3945,
+      "step": 4571
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299076735271914e-05,
+      "loss": 1.2119,
+      "step": 4572
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298695497711534e-05,
+      "loss": 1.4277,
+      "step": 4573
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298314160267923e-05,
+      "loss": 1.6094,
+      "step": 4574
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929793272294518e-05,
+      "loss": 1.0518,
+      "step": 4575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297551185747396e-05,
+      "loss": 1.0674,
+      "step": 4576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929716954867867e-05,
+      "loss": 1.4434,
+      "step": 4577
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296787811743104e-05,
+      "loss": 0.8379,
+      "step": 4578
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296405974944796e-05,
+      "loss": 1.3584,
+      "step": 4579
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296024038287853e-05,
+      "loss": 1.3955,
+      "step": 4580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929564200177637e-05,
+      "loss": 1.1504,
+      "step": 4581
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295259865414456e-05,
+      "loss": 1.3418,
+      "step": 4582
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929487762920622e-05,
+      "loss": 0.9629,
+      "step": 4583
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294495293155754e-05,
+      "loss": 1.2441,
+      "step": 4584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294112857267177e-05,
+      "loss": 0.9404,
+      "step": 4585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929373032154459e-05,
+      "loss": 1.041,
+      "step": 4586
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9293347685992108e-05,
+      "loss": 1.2979,
+      "step": 4587
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292964950613836e-05,
+      "loss": 1.2441,
+      "step": 4588
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929258211541389e-05,
+      "loss": 1.4326,
+      "step": 4589
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929219918039638e-05,
+      "loss": 1.0986,
+      "step": 4590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291816145565415e-05,
+      "loss": 1.0752,
+      "step": 4591
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929143301092512e-05,
+      "loss": 1.041,
+      "step": 4592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92910497764796e-05,
+      "loss": 1.4219,
+      "step": 4593
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290666442232976e-05,
+      "loss": 1.3613,
+      "step": 4594
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290283008189366e-05,
+      "loss": 1.1035,
+      "step": 4595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289899474352888e-05,
+      "loss": 1.3623,
+      "step": 4596
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928951584072766e-05,
+      "loss": 1.3516,
+      "step": 4597
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289132107317806e-05,
+      "loss": 1.2676,
+      "step": 4598
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288748274127446e-05,
+      "loss": 1.0557,
+      "step": 4599
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288364341160705e-05,
+      "loss": 1.2129,
+      "step": 4600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287980308421702e-05,
+      "loss": 1.3535,
+      "step": 4601
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287596175914568e-05,
+      "loss": 0.7007,
+      "step": 4602
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287211943643425e-05,
+      "loss": 1.3193,
+      "step": 4603
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286827611612405e-05,
+      "loss": 1.5586,
+      "step": 4604
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286443179825633e-05,
+      "loss": 1.252,
+      "step": 4605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928605864828724e-05,
+      "loss": 1.4844,
+      "step": 4606
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285674017001353e-05,
+      "loss": 1.2344,
+      "step": 4607
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285289285972106e-05,
+      "loss": 0.7546,
+      "step": 4608
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284904455203634e-05,
+      "loss": 1.2607,
+      "step": 4609
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928451952470007e-05,
+      "loss": 1.1357,
+      "step": 4610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284134494465543e-05,
+      "loss": 1.1777,
+      "step": 4611
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283749364504193e-05,
+      "loss": 1.209,
+      "step": 4612
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283364134820158e-05,
+      "loss": 1.1904,
+      "step": 4613
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282978805417576e-05,
+      "loss": 1.4072,
+      "step": 4614
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282593376300585e-05,
+      "loss": 1.0674,
+      "step": 4615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282207847473322e-05,
+      "loss": 1.4473,
+      "step": 4616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281822218939937e-05,
+      "loss": 1.2925,
+      "step": 4617
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928143649070456e-05,
+      "loss": 1.3652,
+      "step": 4618
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281050662771345e-05,
+      "loss": 1.0,
+      "step": 4619
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280664735144433e-05,
+      "loss": 1.2305,
+      "step": 4620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280278707827967e-05,
+      "loss": 1.2861,
+      "step": 4621
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279892580826097e-05,
+      "loss": 1.3369,
+      "step": 4622
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927950635414297e-05,
+      "loss": 1.4395,
+      "step": 4623
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279120027782733e-05,
+      "loss": 0.9023,
+      "step": 4624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927873360174954e-05,
+      "loss": 1.1396,
+      "step": 4625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278347076047533e-05,
+      "loss": 1.3164,
+      "step": 4626
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927796045068087e-05,
+      "loss": 1.3008,
+      "step": 4627
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277573725653705e-05,
+      "loss": 1.2686,
+      "step": 4628
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927718690097019e-05,
+      "loss": 1.4414,
+      "step": 4629
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927679997663448e-05,
+      "loss": 1.0308,
+      "step": 4630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276412952650732e-05,
+      "loss": 1.5254,
+      "step": 4631
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276025829023104e-05,
+      "loss": 1.5645,
+      "step": 4632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9275638605755754e-05,
+      "loss": 1.146,
+      "step": 4633
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927525128285284e-05,
+      "loss": 1.1133,
+      "step": 4634
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274863860318523e-05,
+      "loss": 1.1211,
+      "step": 4635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274476338156963e-05,
+      "loss": 0.6616,
+      "step": 4636
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927408871637233e-05,
+      "loss": 1.2773,
+      "step": 4637
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273700994968778e-05,
+      "loss": 0.8535,
+      "step": 4638
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273313173950477e-05,
+      "loss": 1.377,
+      "step": 4639
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927292525332159e-05,
+      "loss": 1.0918,
+      "step": 4640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927253723308629e-05,
+      "loss": 1.6855,
+      "step": 4641
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9272149113248737e-05,
+      "loss": 1.2549,
+      "step": 4642
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9271760893813104e-05,
+      "loss": 0.9805,
+      "step": 4643
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927137257478356e-05,
+      "loss": 1.3223,
+      "step": 4644
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270984156164278e-05,
+      "loss": 1.0957,
+      "step": 4645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927059563795943e-05,
+      "loss": 0.7803,
+      "step": 4646
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270207020173187e-05,
+      "loss": 1.4395,
+      "step": 4647
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269818302809727e-05,
+      "loss": 1.4404,
+      "step": 4648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926942948587322e-05,
+      "loss": 1.3198,
+      "step": 4649
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926904056936785e-05,
+      "loss": 1.1309,
+      "step": 4650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926865155329779e-05,
+      "loss": 1.0625,
+      "step": 4651
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926826243766722e-05,
+      "loss": 1.2627,
+      "step": 4652
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267873222480317e-05,
+      "loss": 1.1387,
+      "step": 4653
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926748390774126e-05,
+      "loss": 1.2656,
+      "step": 4654
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926709449345424e-05,
+      "loss": 1.75,
+      "step": 4655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266704979623433e-05,
+      "loss": 1.083,
+      "step": 4656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266315366253025e-05,
+      "loss": 1.5098,
+      "step": 4657
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265925653347198e-05,
+      "loss": 1.2129,
+      "step": 4658
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265535840910145e-05,
+      "loss": 1.1484,
+      "step": 4659
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265145928946047e-05,
+      "loss": 0.9561,
+      "step": 4660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264755917459098e-05,
+      "loss": 1.4395,
+      "step": 4661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264365806453478e-05,
+      "loss": 1.4199,
+      "step": 4662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9263975595933386e-05,
+      "loss": 1.0059,
+      "step": 4663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926358528590301e-05,
+      "loss": 1.0684,
+      "step": 4664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926319487636654e-05,
+      "loss": 1.9004,
+      "step": 4665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926280436732818e-05,
+      "loss": 1.1787,
+      "step": 4666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9262413758792117e-05,
+      "loss": 1.1953,
+      "step": 4667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926202305076254e-05,
+      "loss": 1.2646,
+      "step": 4668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261632243243657e-05,
+      "loss": 0.9961,
+      "step": 4669
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261241336239666e-05,
+      "loss": 1.1025,
+      "step": 4670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260850329754758e-05,
+      "loss": 0.8232,
+      "step": 4671
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926045922379314e-05,
+      "loss": 1.3291,
+      "step": 4672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260068018359007e-05,
+      "loss": 1.2725,
+      "step": 4673
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259676713456565e-05,
+      "loss": 1.2891,
+      "step": 4674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259285309090016e-05,
+      "loss": 1.4258,
+      "step": 4675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258893805263568e-05,
+      "loss": 1.3809,
+      "step": 4676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925850220198142e-05,
+      "loss": 1.2617,
+      "step": 4677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258110499247787e-05,
+      "loss": 1.2109,
+      "step": 4678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257718697066865e-05,
+      "loss": 1.6025,
+      "step": 4679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925732679544287e-05,
+      "loss": 1.0317,
+      "step": 4680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256934794380015e-05,
+      "loss": 1.4434,
+      "step": 4681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256542693882505e-05,
+      "loss": 1.2969,
+      "step": 4682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925615049395455e-05,
+      "loss": 1.2422,
+      "step": 4683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255758194600373e-05,
+      "loss": 0.9453,
+      "step": 4684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255365795824174e-05,
+      "loss": 1.1094,
+      "step": 4685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925497329763018e-05,
+      "loss": 1.4512,
+      "step": 4686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92545807000226e-05,
+      "loss": 1.2422,
+      "step": 4687
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9254188003005652e-05,
+      "loss": 1.3389,
+      "step": 4688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253795206583555e-05,
+      "loss": 1.3008,
+      "step": 4689
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253402310760532e-05,
+      "loss": 1.4121,
+      "step": 4690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253009315540798e-05,
+      "loss": 0.9175,
+      "step": 4691
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252616220928578e-05,
+      "loss": 0.8584,
+      "step": 4692
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925222302692809e-05,
+      "loss": 1.375,
+      "step": 4693
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251829733543563e-05,
+      "loss": 1.0967,
+      "step": 4694
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251436340779215e-05,
+      "loss": 1.4209,
+      "step": 4695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925104284863928e-05,
+      "loss": 1.2793,
+      "step": 4696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925064925712798e-05,
+      "loss": 1.0869,
+      "step": 4697
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9250255566249546e-05,
+      "loss": 1.2451,
+      "step": 4698
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92498617760082e-05,
+      "loss": 1.2891,
+      "step": 4699
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924946788640818e-05,
+      "loss": 1.666,
+      "step": 4700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924907389745371e-05,
+      "loss": 1.3555,
+      "step": 4701
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248679809149026e-05,
+      "loss": 1.5781,
+      "step": 4702
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924828562149836e-05,
+      "loss": 1.3926,
+      "step": 4703
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924789133450595e-05,
+      "loss": 0.8066,
+      "step": 4704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247496948176024e-05,
+      "loss": 1.2031,
+      "step": 4705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247102462512823e-05,
+      "loss": 1.3301,
+      "step": 4706
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9246707877520586e-05,
+      "loss": 1.2334,
+      "step": 4707
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9246313193203544e-05,
+      "loss": 0.9124,
+      "step": 4708
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924591840956595e-05,
+      "loss": 1.0811,
+      "step": 4709
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245523526612032e-05,
+      "loss": 0.9395,
+      "step": 4710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924512854434604e-05,
+      "loss": 1.7773,
+      "step": 4711
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244733462772205e-05,
+      "loss": 1.3447,
+      "step": 4712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244338281894783e-05,
+      "loss": 1.3213,
+      "step": 4713
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924394300171801e-05,
+      "loss": 1.3643,
+      "step": 4714
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9243547622246143e-05,
+      "loss": 0.9502,
+      "step": 4715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9243152143483425e-05,
+      "loss": 1.2461,
+      "step": 4716
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242756565434094e-05,
+      "loss": 1.998,
+      "step": 4717
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924236088810241e-05,
+      "loss": 0.8047,
+      "step": 4718
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241965111492623e-05,
+      "loss": 1.0049,
+      "step": 4719
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241569235608978e-05,
+      "loss": 1.0098,
+      "step": 4720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924117326045573e-05,
+      "loss": 1.3408,
+      "step": 4721
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240777186037137e-05,
+      "loss": 1.5039,
+      "step": 4722
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240381012357446e-05,
+      "loss": 0.9316,
+      "step": 4723
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239984739420916e-05,
+      "loss": 1.5098,
+      "step": 4724
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239588367231804e-05,
+      "loss": 1.0205,
+      "step": 4725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923919189579437e-05,
+      "loss": 1.3418,
+      "step": 4726
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238795325112867e-05,
+      "loss": 1.1475,
+      "step": 4727
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238398655191562e-05,
+      "loss": 1.0049,
+      "step": 4728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923800188603471e-05,
+      "loss": 1.7236,
+      "step": 4729
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237605017646573e-05,
+      "loss": 1.0693,
+      "step": 4730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923720805003142e-05,
+      "loss": 0.8672,
+      "step": 4731
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236810983193506e-05,
+      "loss": 1.001,
+      "step": 4732
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236413817137106e-05,
+      "loss": 1.0391,
+      "step": 4733
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923601655186648e-05,
+      "loss": 1.1816,
+      "step": 4734
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235619187385895e-05,
+      "loss": 1.4453,
+      "step": 4735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235221723699623e-05,
+      "loss": 1.4219,
+      "step": 4736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234824160811926e-05,
+      "loss": 1.1914,
+      "step": 4737
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234426498727085e-05,
+      "loss": 0.791,
+      "step": 4738
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234028737449367e-05,
+      "loss": 1.4912,
+      "step": 4739
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923363087698304e-05,
+      "loss": 1.3223,
+      "step": 4740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9233232917332384e-05,
+      "loss": 1.3184,
+      "step": 4741
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923283485850167e-05,
+      "loss": 1.1064,
+      "step": 4742
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923243670049518e-05,
+      "loss": 1.0488,
+      "step": 4743
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232038443317182e-05,
+      "loss": 0.9033,
+      "step": 4744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231640086971957e-05,
+      "loss": 1.3633,
+      "step": 4745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923124163146379e-05,
+      "loss": 1.1699,
+      "step": 4746
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9230843076796954e-05,
+      "loss": 1.6816,
+      "step": 4747
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923044442297573e-05,
+      "loss": 1.0947,
+      "step": 4748
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9230045670004403e-05,
+      "loss": 1.3467,
+      "step": 4749
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229646817887254e-05,
+      "loss": 1.3418,
+      "step": 4750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229247866628574e-05,
+      "loss": 1.3008,
+      "step": 4751
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922884881623264e-05,
+      "loss": 1.4824,
+      "step": 4752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9228449666703744e-05,
+      "loss": 1.1074,
+      "step": 4753
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922805041804617e-05,
+      "loss": 1.2021,
+      "step": 4754
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9227651070264207e-05,
+      "loss": 1.3213,
+      "step": 4755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9227251623362145e-05,
+      "loss": 1.0508,
+      "step": 4756
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226852077344282e-05,
+      "loss": 1.2012,
+      "step": 4757
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226452432214894e-05,
+      "loss": 1.332,
+      "step": 4758
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226052687978287e-05,
+      "loss": 1.3145,
+      "step": 4759
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922565284463875e-05,
+      "loss": 1.1934,
+      "step": 4760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922525290220058e-05,
+      "loss": 1.0825,
+      "step": 4761
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224852860668072e-05,
+      "loss": 1.1855,
+      "step": 4762
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224452720045525e-05,
+      "loss": 0.8706,
+      "step": 4763
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922405248033723e-05,
+      "loss": 1.0288,
+      "step": 4764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223652141547495e-05,
+      "loss": 1.4609,
+      "step": 4765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223251703680616e-05,
+      "loss": 1.4043,
+      "step": 4766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222851166740893e-05,
+      "loss": 1.1973,
+      "step": 4767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222450530732634e-05,
+      "loss": 1.252,
+      "step": 4768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222049795660135e-05,
+      "loss": 1.3496,
+      "step": 4769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922164896152771e-05,
+      "loss": 0.8838,
+      "step": 4770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221248028339655e-05,
+      "loss": 1.0156,
+      "step": 4771
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220846996100283e-05,
+      "loss": 1.1426,
+      "step": 4772
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220445864813896e-05,
+      "loss": 1.2568,
+      "step": 4773
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220044634484807e-05,
+      "loss": 1.1836,
+      "step": 4774
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921964330511733e-05,
+      "loss": 0.9517,
+      "step": 4775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219241876715763e-05,
+      "loss": 1.3389,
+      "step": 4776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9218840349284433e-05,
+      "loss": 0.7363,
+      "step": 4777
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9218438722827644e-05,
+      "loss": 1.6953,
+      "step": 4778
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921803699734971e-05,
+      "loss": 1.3174,
+      "step": 4779
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217635172854955e-05,
+      "loss": 1.21,
+      "step": 4780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217233249347682e-05,
+      "loss": 1.4443,
+      "step": 4781
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9216831226832218e-05,
+      "loss": 1.0752,
+      "step": 4782
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921642910531288e-05,
+      "loss": 1.1191,
+      "step": 4783
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9216026884793982e-05,
+      "loss": 1.6055,
+      "step": 4784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921562456527985e-05,
+      "loss": 1.3594,
+      "step": 4785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215222146774806e-05,
+      "loss": 1.2549,
+      "step": 4786
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921481962928317e-05,
+      "loss": 1.2744,
+      "step": 4787
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214417012809266e-05,
+      "loss": 1.3613,
+      "step": 4788
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921401429735742e-05,
+      "loss": 1.6367,
+      "step": 4789
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213611482931954e-05,
+      "loss": 1.2344,
+      "step": 4790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213208569537198e-05,
+      "loss": 1.46,
+      "step": 4791
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212805557177484e-05,
+      "loss": 0.8584,
+      "step": 4792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921240244585713e-05,
+      "loss": 1.2881,
+      "step": 4793
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211999235580475e-05,
+      "loss": 0.8955,
+      "step": 4794
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211595926351854e-05,
+      "loss": 1.2793,
+      "step": 4795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921119251817559e-05,
+      "loss": 1.4463,
+      "step": 4796
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9210789011056015e-05,
+      "loss": 0.9932,
+      "step": 4797
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921038540499747e-05,
+      "loss": 1.1162,
+      "step": 4798
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920998170000429e-05,
+      "loss": 1.4033,
+      "step": 4799
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920957789608081e-05,
+      "loss": 1.2725,
+      "step": 4800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209173993231368e-05,
+      "loss": 1.4375,
+      "step": 4801
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208769991460298e-05,
+      "loss": 1.2871,
+      "step": 4802
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920836589077194e-05,
+      "loss": 1.2705,
+      "step": 4803
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9207961691170648e-05,
+      "loss": 1.168,
+      "step": 4804
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920755739266075e-05,
+      "loss": 0.876,
+      "step": 4805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920715299524659e-05,
+      "loss": 1.2783,
+      "step": 4806
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206748498932514e-05,
+      "loss": 1.1904,
+      "step": 4807
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206343903722874e-05,
+      "loss": 1.375,
+      "step": 4808
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9205939209622e-05,
+      "loss": 1.4214,
+      "step": 4809
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9205534416634254e-05,
+      "loss": 0.8076,
+      "step": 4810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920512952476398e-05,
+      "loss": 1.4141,
+      "step": 4811
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204724534015525e-05,
+      "loss": 1.3164,
+      "step": 4812
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920431944439324e-05,
+      "loss": 1.1431,
+      "step": 4813
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203914255901477e-05,
+      "loss": 1.3457,
+      "step": 4814
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203508968544586e-05,
+      "loss": 0.877,
+      "step": 4815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203103582326922e-05,
+      "loss": 1.6367,
+      "step": 4816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920269809725284e-05,
+      "loss": 1.0488,
+      "step": 4817
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202292513326698e-05,
+      "loss": 1.2871,
+      "step": 4818
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920188683055285e-05,
+      "loss": 1.2012,
+      "step": 4819
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201481048935652e-05,
+      "loss": 1.3018,
+      "step": 4820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201075168479468e-05,
+      "loss": 1.2891,
+      "step": 4821
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920066918918865e-05,
+      "loss": 1.3145,
+      "step": 4822
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200263111067566e-05,
+      "loss": 1.3047,
+      "step": 4823
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199856934120574e-05,
+      "loss": 0.9609,
+      "step": 4824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199450658352033e-05,
+      "loss": 1.2441,
+      "step": 4825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919904428376632e-05,
+      "loss": 1.416,
+      "step": 4826
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919863781036779e-05,
+      "loss": 1.1279,
+      "step": 4827
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9198231238160813e-05,
+      "loss": 1.2754,
+      "step": 4828
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919782456714975e-05,
+      "loss": 1.1357,
+      "step": 4829
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197417797338978e-05,
+      "loss": 1.0635,
+      "step": 4830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197010928732863e-05,
+      "loss": 1.0273,
+      "step": 4831
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196603961335774e-05,
+      "loss": 1.2949,
+      "step": 4832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196196895152083e-05,
+      "loss": 0.9385,
+      "step": 4833
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919578973018616e-05,
+      "loss": 1.2676,
+      "step": 4834
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195382466442386e-05,
+      "loss": 1.3887,
+      "step": 4835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194975103925132e-05,
+      "loss": 1.2158,
+      "step": 4836
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919456764263877e-05,
+      "loss": 1.1045,
+      "step": 4837
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919416008258768e-05,
+      "loss": 1.2754,
+      "step": 4838
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9193752423776243e-05,
+      "loss": 1.4521,
+      "step": 4839
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9193344666208833e-05,
+      "loss": 0.9658,
+      "step": 4840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192936809889833e-05,
+      "loss": 1.2441,
+      "step": 4841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919252885482362e-05,
+      "loss": 1.5146,
+      "step": 4842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919212080101458e-05,
+      "loss": 1.2153,
+      "step": 4843
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191712648467093e-05,
+      "loss": 1.4971,
+      "step": 4844
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191304397185548e-05,
+      "loss": 1.2041,
+      "step": 4845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190896047174326e-05,
+      "loss": 1.2969,
+      "step": 4846
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190487598437813e-05,
+      "loss": 1.2637,
+      "step": 4847
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190079050980403e-05,
+      "loss": 1.2783,
+      "step": 4848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189670404806476e-05,
+      "loss": 1.0049,
+      "step": 4849
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189261659920428e-05,
+      "loss": 1.0576,
+      "step": 4850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188852816326645e-05,
+      "loss": 1.167,
+      "step": 4851
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188443874029515e-05,
+      "loss": 1.0869,
+      "step": 4852
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188034833033444e-05,
+      "loss": 1.0132,
+      "step": 4853
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187625693342818e-05,
+      "loss": 0.9336,
+      "step": 4854
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187216454962028e-05,
+      "loss": 1.1543,
+      "step": 4855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186807117895477e-05,
+      "loss": 1.3164,
+      "step": 4856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186397682147554e-05,
+      "loss": 1.1914,
+      "step": 4857
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185988147722664e-05,
+      "loss": 1.4434,
+      "step": 4858
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185578514625206e-05,
+      "loss": 1.1787,
+      "step": 4859
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185168782859573e-05,
+      "loss": 1.2949,
+      "step": 4860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9184758952430177e-05,
+      "loss": 1.1426,
+      "step": 4861
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918434902334141e-05,
+      "loss": 1.2988,
+      "step": 4862
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918393899559768e-05,
+      "loss": 1.2451,
+      "step": 4863
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918352886920339e-05,
+      "loss": 1.2842,
+      "step": 4864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183118644162948e-05,
+      "loss": 1.5293,
+      "step": 4865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918270832048076e-05,
+      "loss": 1.0723,
+      "step": 4866
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918229789816123e-05,
+      "loss": 1.2188,
+      "step": 4867
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181887377208774e-05,
+      "loss": 1.0752,
+      "step": 4868
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918147675762779e-05,
+      "loss": 1.75,
+      "step": 4869
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91810660394227e-05,
+      "loss": 1.0674,
+      "step": 4870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918065522259791e-05,
+      "loss": 1.0361,
+      "step": 4871
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918024430715783e-05,
+      "loss": 1.481,
+      "step": 4872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179833293106884e-05,
+      "loss": 0.9092,
+      "step": 4873
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179422180449478e-05,
+      "loss": 1.1523,
+      "step": 4874
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179010969190034e-05,
+      "loss": 1.1094,
+      "step": 4875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917859965933296e-05,
+      "loss": 1.1104,
+      "step": 4876
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178188250882685e-05,
+      "loss": 1.2744,
+      "step": 4877
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177776743843625e-05,
+      "loss": 1.377,
+      "step": 4878
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177365138220193e-05,
+      "loss": 1.5283,
+      "step": 4879
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917695343401682e-05,
+      "loss": 1.1152,
+      "step": 4880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917654163123792e-05,
+      "loss": 1.3496,
+      "step": 4881
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9176129729887925e-05,
+      "loss": 1.1445,
+      "step": 4882
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175717729971256e-05,
+      "loss": 1.3086,
+      "step": 4883
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175305631492335e-05,
+      "loss": 1.1641,
+      "step": 4884
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917489343445559e-05,
+      "loss": 1.0315,
+      "step": 4885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174481138865457e-05,
+      "loss": 0.9648,
+      "step": 4886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174068744726352e-05,
+      "loss": 1.333,
+      "step": 4887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917365625204271e-05,
+      "loss": 1.3154,
+      "step": 4888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173243660818967e-05,
+      "loss": 1.3105,
+      "step": 4889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172830971059546e-05,
+      "loss": 1.4102,
+      "step": 4890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172418182768885e-05,
+      "loss": 1.4824,
+      "step": 4891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917200529595142e-05,
+      "loss": 1.2207,
+      "step": 4892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171592310611582e-05,
+      "loss": 1.502,
+      "step": 4893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917117922675381e-05,
+      "loss": 1.1494,
+      "step": 4894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170766044382537e-05,
+      "loss": 1.1475,
+      "step": 4895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170352763502204e-05,
+      "loss": 1.1885,
+      "step": 4896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9169939384117252e-05,
+      "loss": 1.2646,
+      "step": 4897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916952590623212e-05,
+      "loss": 1.4014,
+      "step": 4898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916911232985125e-05,
+      "loss": 1.3125,
+      "step": 4899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168698654979078e-05,
+      "loss": 1.459,
+      "step": 4900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168284881620057e-05,
+      "loss": 1.1768,
+      "step": 4901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916787100977863e-05,
+      "loss": 1.2773,
+      "step": 4902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916745703945924e-05,
+      "loss": 1.3086,
+      "step": 4903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9167042970666333e-05,
+      "loss": 1.2646,
+      "step": 4904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166628803404357e-05,
+      "loss": 1.0215,
+      "step": 4905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916621453767776e-05,
+      "loss": 1.085,
+      "step": 4906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165800173490998e-05,
+      "loss": 1.1338,
+      "step": 4907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165385710848515e-05,
+      "loss": 1.2422,
+      "step": 4908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916497114975477e-05,
+      "loss": 1.0068,
+      "step": 4909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916455649021421e-05,
+      "loss": 1.1182,
+      "step": 4910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164141732231286e-05,
+      "loss": 0.9326,
+      "step": 4911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916372687581046e-05,
+      "loss": 1.3057,
+      "step": 4912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163311920956186e-05,
+      "loss": 1.3164,
+      "step": 4913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162896867672926e-05,
+      "loss": 0.9932,
+      "step": 4914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162481715965132e-05,
+      "loss": 1.4707,
+      "step": 4915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162066465837263e-05,
+      "loss": 1.0508,
+      "step": 4916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916165111729378e-05,
+      "loss": 1.5107,
+      "step": 4917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916123567033915e-05,
+      "loss": 0.8066,
+      "step": 4918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160820124977834e-05,
+      "loss": 1.29,
+      "step": 4919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916040448121429e-05,
+      "loss": 1.5586,
+      "step": 4920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159988739052987e-05,
+      "loss": 0.9697,
+      "step": 4921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159572898498387e-05,
+      "loss": 1.2852,
+      "step": 4922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159156959554963e-05,
+      "loss": 1.2188,
+      "step": 4923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915874092222718e-05,
+      "loss": 1.2529,
+      "step": 4924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9158324786519506e-05,
+      "loss": 1.1357,
+      "step": 4925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915790855243641e-05,
+      "loss": 1.1406,
+      "step": 4926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157492219982367e-05,
+      "loss": 0.7388,
+      "step": 4927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157075789161842e-05,
+      "loss": 1.2329,
+      "step": 4928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915665925997932e-05,
+      "loss": 0.8594,
+      "step": 4929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156242632439264e-05,
+      "loss": 1.2656,
+      "step": 4930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155825906546152e-05,
+      "loss": 1.2109,
+      "step": 4931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155409082304464e-05,
+      "loss": 1.4785,
+      "step": 4932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154992159718674e-05,
+      "loss": 1.0742,
+      "step": 4933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154575138793263e-05,
+      "loss": 1.1631,
+      "step": 4934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154158019532707e-05,
+      "loss": 1.6523,
+      "step": 4935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153740801941486e-05,
+      "loss": 1.1914,
+      "step": 4936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153323486024088e-05,
+      "loss": 1.1611,
+      "step": 4937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152906071784993e-05,
+      "loss": 1.0215,
+      "step": 4938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915248855922868e-05,
+      "loss": 1.4062,
+      "step": 4939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152070948359637e-05,
+      "loss": 1.0977,
+      "step": 4940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915165323918235e-05,
+      "loss": 0.9717,
+      "step": 4941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151235431701312e-05,
+      "loss": 0.8066,
+      "step": 4942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150817525920998e-05,
+      "loss": 1.0889,
+      "step": 4943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150399521845905e-05,
+      "loss": 1.4492,
+      "step": 4944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149981419480524e-05,
+      "loss": 1.459,
+      "step": 4945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149563218829345e-05,
+      "loss": 1.29,
+      "step": 4946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149144919896856e-05,
+      "loss": 1.2012,
+      "step": 4947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148726522687555e-05,
+      "loss": 1.2129,
+      "step": 4948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148308027205932e-05,
+      "loss": 1.0933,
+      "step": 4949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147889433456487e-05,
+      "loss": 1.126,
+      "step": 4950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147470741443715e-05,
+      "loss": 1.0488,
+      "step": 4951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147051951172112e-05,
+      "loss": 1.207,
+      "step": 4952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146633062646178e-05,
+      "loss": 1.3828,
+      "step": 4953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914621407587041e-05,
+      "loss": 1.2773,
+      "step": 4954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9145794990849312e-05,
+      "loss": 1.1621,
+      "step": 4955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914537580758738e-05,
+      "loss": 1.2344,
+      "step": 4956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144956526089128e-05,
+      "loss": 0.9541,
+      "step": 4957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144537146359047e-05,
+      "loss": 1.0674,
+      "step": 4958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144117668401652e-05,
+      "loss": 1.1797,
+      "step": 4959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914369809222144e-05,
+      "loss": 1.2578,
+      "step": 4960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143278417822925e-05,
+      "loss": 0.7539,
+      "step": 4961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914285864521061e-05,
+      "loss": 0.9707,
+      "step": 4962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914243877438901e-05,
+      "loss": 1.4766,
+      "step": 4963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142018805362626e-05,
+      "loss": 1.168,
+      "step": 4964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141598738135977e-05,
+      "loss": 0.8994,
+      "step": 4965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141178572713574e-05,
+      "loss": 1.3486,
+      "step": 4966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914075830909993e-05,
+      "loss": 1.4678,
+      "step": 4967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140337947299556e-05,
+      "loss": 1.2988,
+      "step": 4968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913991748731697e-05,
+      "loss": 1.0713,
+      "step": 4969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139496929156685e-05,
+      "loss": 1.0527,
+      "step": 4970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139076272823226e-05,
+      "loss": 1.5996,
+      "step": 4971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138655518321104e-05,
+      "loss": 1.1621,
+      "step": 4972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138234665654844e-05,
+      "loss": 1.3828,
+      "step": 4973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137813714828967e-05,
+      "loss": 1.0137,
+      "step": 4974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137392665847987e-05,
+      "loss": 1.2422,
+      "step": 4975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9136971518716435e-05,
+      "loss": 1.5664,
+      "step": 4976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913655027343883e-05,
+      "loss": 0.9248,
+      "step": 4977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91361289300197e-05,
+      "loss": 0.9453,
+      "step": 4978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135707488463565e-05,
+      "loss": 1.3965,
+      "step": 4979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135285948774962e-05,
+      "loss": 1.1328,
+      "step": 4980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9134864310958413e-05,
+      "loss": 1.1895,
+      "step": 4981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9134442575018443e-05,
+      "loss": 1.4141,
+      "step": 4982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913402074095959e-05,
+      "loss": 1.0659,
+      "step": 4983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913359880878638e-05,
+      "loss": 1.2373,
+      "step": 4984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913317677850335e-05,
+      "loss": 1.21,
+      "step": 4985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132754650115028e-05,
+      "loss": 1.3477,
+      "step": 4986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913233242362595e-05,
+      "loss": 1.5156,
+      "step": 4987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131910099040655e-05,
+      "loss": 0.9258,
+      "step": 4988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131487676363675e-05,
+      "loss": 1.2842,
+      "step": 4989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131065155599547e-05,
+      "loss": 1.1777,
+      "step": 4990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130642536752816e-05,
+      "loss": 1.2188,
+      "step": 4991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130219819828015e-05,
+      "loss": 1.0908,
+      "step": 4992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129797004829686e-05,
+      "loss": 1.1133,
+      "step": 4993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129374091762375e-05,
+      "loss": 1.2031,
+      "step": 4994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128951080630617e-05,
+      "loss": 1.5684,
+      "step": 4995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128527971438965e-05,
+      "loss": 1.0596,
+      "step": 4996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128104764191954e-05,
+      "loss": 1.1523,
+      "step": 4997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9127681458894137e-05,
+      "loss": 1.6016,
+      "step": 4998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912725805555006e-05,
+      "loss": 1.416,
+      "step": 4999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912683455416427e-05,
+      "loss": 1.7031,
+      "step": 5000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9126410954741316e-05,
+      "loss": 1.0869,
+      "step": 5001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912598725728575e-05,
+      "loss": 1.5605,
+      "step": 5002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912556346180212e-05,
+      "loss": 1.3252,
+      "step": 5003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912513956829498e-05,
+      "loss": 1.5684,
+      "step": 5004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124715576768885e-05,
+      "loss": 1.165,
+      "step": 5005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124291487228386e-05,
+      "loss": 0.8809,
+      "step": 5006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912386729967804e-05,
+      "loss": 1.1328,
+      "step": 5007
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123443014122402e-05,
+      "loss": 1.2949,
+      "step": 5008
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123018630566033e-05,
+      "loss": 1.5586,
+      "step": 5009
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9122594149013488e-05,
+      "loss": 1.6416,
+      "step": 5010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912216956946933e-05,
+      "loss": 1.2773,
+      "step": 5011
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121744891938115e-05,
+      "loss": 1.4785,
+      "step": 5012
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121320116424412e-05,
+      "loss": 0.8535,
+      "step": 5013
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120895242932774e-05,
+      "loss": 1.0332,
+      "step": 5014
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120470271467774e-05,
+      "loss": 1.3516,
+      "step": 5015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120045202033973e-05,
+      "loss": 1.2178,
+      "step": 5016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119620034635937e-05,
+      "loss": 1.1221,
+      "step": 5017
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911919476927823e-05,
+      "loss": 0.9727,
+      "step": 5018
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9118769405965425e-05,
+      "loss": 0.7295,
+      "step": 5019
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9118343944702088e-05,
+      "loss": 1.3486,
+      "step": 5020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117918385492793e-05,
+      "loss": 1.2529,
+      "step": 5021
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911749272834211e-05,
+      "loss": 0.8779,
+      "step": 5022
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117066973254603e-05,
+      "loss": 1.3203,
+      "step": 5023
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116641120234858e-05,
+      "loss": 1.1533,
+      "step": 5024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116215169287442e-05,
+      "loss": 1.2656,
+      "step": 5025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911578912041693e-05,
+      "loss": 1.3379,
+      "step": 5026
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115362973627902e-05,
+      "loss": 1.2871,
+      "step": 5027
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114936728924934e-05,
+      "loss": 0.833,
+      "step": 5028
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114510386312603e-05,
+      "loss": 1.4688,
+      "step": 5029
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114083945795492e-05,
+      "loss": 1.1719,
+      "step": 5030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911365740737818e-05,
+      "loss": 1.2812,
+      "step": 5031
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9113230771065246e-05,
+      "loss": 1.0791,
+      "step": 5032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9112804036861273e-05,
+      "loss": 0.8521,
+      "step": 5033
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9112377204770853e-05,
+      "loss": 1.627,
+      "step": 5034
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911195027479856e-05,
+      "loss": 1.6133,
+      "step": 5035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111523246948987e-05,
+      "loss": 0.9229,
+      "step": 5036
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111096121226718e-05,
+      "loss": 1.3672,
+      "step": 5037
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110668897636344e-05,
+      "loss": 1.3164,
+      "step": 5038
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911024157618245e-05,
+      "loss": 1.1826,
+      "step": 5039
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109814156869627e-05,
+      "loss": 1.5469,
+      "step": 5040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109386639702472e-05,
+      "loss": 1.0947,
+      "step": 5041
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910895902468557e-05,
+      "loss": 1.4492,
+      "step": 5042
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108531311823514e-05,
+      "loss": 1.582,
+      "step": 5043
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108103501120903e-05,
+      "loss": 1.1729,
+      "step": 5044
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910767559258233e-05,
+      "loss": 1.0132,
+      "step": 5045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9107247586212395e-05,
+      "loss": 1.6113,
+      "step": 5046
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910681948201569e-05,
+      "loss": 1.4121,
+      "step": 5047
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106391279996813e-05,
+      "loss": 0.998,
+      "step": 5048
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910596298016037e-05,
+      "loss": 1.2275,
+      "step": 5049
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105534582510956e-05,
+      "loss": 1.3105,
+      "step": 5050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910510608705318e-05,
+      "loss": 1.2998,
+      "step": 5051
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104677493791633e-05,
+      "loss": 0.8975,
+      "step": 5052
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910424880273093e-05,
+      "loss": 1.2393,
+      "step": 5053
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910382001387567e-05,
+      "loss": 1.6514,
+      "step": 5054
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103391127230457e-05,
+      "loss": 1.1606,
+      "step": 5055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102962142799905e-05,
+      "loss": 1.4062,
+      "step": 5056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102533060588618e-05,
+      "loss": 1.0078,
+      "step": 5057
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102103880601205e-05,
+      "loss": 1.0459,
+      "step": 5058
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101674602842272e-05,
+      "loss": 1.1973,
+      "step": 5059
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101245227316442e-05,
+      "loss": 1.3613,
+      "step": 5060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100815754028312e-05,
+      "loss": 1.4775,
+      "step": 5061
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100386182982508e-05,
+      "loss": 0.9561,
+      "step": 5062
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9099956514183636e-05,
+      "loss": 1.4287,
+      "step": 5063
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9099526747636314e-05,
+      "loss": 0.9316,
+      "step": 5064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9099096883345162e-05,
+      "loss": 1.1553,
+      "step": 5065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098666921314793e-05,
+      "loss": 1.4014,
+      "step": 5066
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098236861549826e-05,
+      "loss": 0.9629,
+      "step": 5067
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097806704054882e-05,
+      "loss": 1.0,
+      "step": 5068
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097376448834582e-05,
+      "loss": 0.9082,
+      "step": 5069
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096946095893544e-05,
+      "loss": 1.4824,
+      "step": 5070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096515645236394e-05,
+      "loss": 1.126,
+      "step": 5071
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096085096867756e-05,
+      "loss": 0.8711,
+      "step": 5072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095654450792254e-05,
+      "loss": 1.0547,
+      "step": 5073
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095223707014514e-05,
+      "loss": 1.7344,
+      "step": 5074
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094792865539164e-05,
+      "loss": 1.459,
+      "step": 5075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909436192637083e-05,
+      "loss": 1.4023,
+      "step": 5076
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093930889514135e-05,
+      "loss": 1.3652,
+      "step": 5077
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909349975497372e-05,
+      "loss": 1.0791,
+      "step": 5078
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093068522754214e-05,
+      "loss": 1.126,
+      "step": 5079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092637192860243e-05,
+      "loss": 1.3535,
+      "step": 5080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092205765296447e-05,
+      "loss": 1.4092,
+      "step": 5081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091774240067455e-05,
+      "loss": 1.0615,
+      "step": 5082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091342617177905e-05,
+      "loss": 1.2627,
+      "step": 5083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090910896632436e-05,
+      "loss": 1.1855,
+      "step": 5084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909047907843568e-05,
+      "loss": 1.5332,
+      "step": 5085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090047162592274e-05,
+      "loss": 1.0459,
+      "step": 5086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089615149106868e-05,
+      "loss": 1.4531,
+      "step": 5087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908918303798409e-05,
+      "loss": 1.1377,
+      "step": 5088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908875082922859e-05,
+      "loss": 1.0811,
+      "step": 5089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908831852284501e-05,
+      "loss": 1.3682,
+      "step": 5090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908788611883799e-05,
+      "loss": 0.7832,
+      "step": 5091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908745361721217e-05,
+      "loss": 1.4316,
+      "step": 5092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908702101797221e-05,
+      "loss": 1.2422,
+      "step": 5093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086588321122745e-05,
+      "loss": 1.2002,
+      "step": 5094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086155526668428e-05,
+      "loss": 1.1992,
+      "step": 5095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085722634613906e-05,
+      "loss": 0.8594,
+      "step": 5096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908528964496383e-05,
+      "loss": 1.2266,
+      "step": 5097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084856557722854e-05,
+      "loss": 0.9531,
+      "step": 5098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084423372895622e-05,
+      "loss": 1.625,
+      "step": 5099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083990090486793e-05,
+      "loss": 1.3711,
+      "step": 5100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083556710501022e-05,
+      "loss": 0.916,
+      "step": 5101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908312323294296e-05,
+      "loss": 1.2715,
+      "step": 5102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9082689657817265e-05,
+      "loss": 1.1494,
+      "step": 5103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9082255985128597e-05,
+      "loss": 1.3594,
+      "step": 5104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908182221488161e-05,
+      "loss": 1.3037,
+      "step": 5105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908138834708097e-05,
+      "loss": 1.3809,
+      "step": 5106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908095438173133e-05,
+      "loss": 1.3184,
+      "step": 5107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080520318837353e-05,
+      "loss": 1.333,
+      "step": 5108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080086158403705e-05,
+      "loss": 1.4414,
+      "step": 5109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079651900435046e-05,
+      "loss": 1.4629,
+      "step": 5110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079217544936043e-05,
+      "loss": 1.0103,
+      "step": 5111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907878309191136e-05,
+      "loss": 1.251,
+      "step": 5112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9078348541365666e-05,
+      "loss": 1.249,
+      "step": 5113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907791389330363e-05,
+      "loss": 1.2207,
+      "step": 5114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077479147729916e-05,
+      "loss": 1.249,
+      "step": 5115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077044304649197e-05,
+      "loss": 1.0176,
+      "step": 5116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076609364066142e-05,
+      "loss": 1.4668,
+      "step": 5117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076174325985425e-05,
+      "loss": 1.5762,
+      "step": 5118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075739190411717e-05,
+      "loss": 1.0234,
+      "step": 5119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075303957349693e-05,
+      "loss": 1.165,
+      "step": 5120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9074868626804027e-05,
+      "loss": 1.1865,
+      "step": 5121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90744331987794e-05,
+      "loss": 1.1973,
+      "step": 5122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073997673280485e-05,
+      "loss": 1.2959,
+      "step": 5123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073562050311964e-05,
+      "loss": 1.5234,
+      "step": 5124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073126329878506e-05,
+      "loss": 0.7295,
+      "step": 5125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072690511984803e-05,
+      "loss": 1.2861,
+      "step": 5126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072254596635532e-05,
+      "loss": 1.1045,
+      "step": 5127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9071818583835374e-05,
+      "loss": 1.4229,
+      "step": 5128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9071382473589015e-05,
+      "loss": 1.2578,
+      "step": 5129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070946265901138e-05,
+      "loss": 1.209,
+      "step": 5130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907050996077643e-05,
+      "loss": 1.2568,
+      "step": 5131
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070073558219576e-05,
+      "loss": 1.3057,
+      "step": 5132
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069637058235262e-05,
+      "loss": 1.0977,
+      "step": 5133
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069200460828184e-05,
+      "loss": 1.1367,
+      "step": 5134
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068763766003023e-05,
+      "loss": 0.9248,
+      "step": 5135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068326973764475e-05,
+      "loss": 1.0322,
+      "step": 5136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067890084117233e-05,
+      "loss": 1.2891,
+      "step": 5137
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067453097065986e-05,
+      "loss": 1.249,
+      "step": 5138
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067016012615427e-05,
+      "loss": 1.1738,
+      "step": 5139
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066578830770255e-05,
+      "loss": 1.3447,
+      "step": 5140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066141551535166e-05,
+      "loss": 1.1465,
+      "step": 5141
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065704174914855e-05,
+      "loss": 1.248,
+      "step": 5142
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065266700914016e-05,
+      "loss": 0.7715,
+      "step": 5143
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906482912953736e-05,
+      "loss": 1.5293,
+      "step": 5144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064391460789574e-05,
+      "loss": 1.3281,
+      "step": 5145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906395369467537e-05,
+      "loss": 1.3916,
+      "step": 5146
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063515831199444e-05,
+      "loss": 1.3955,
+      "step": 5147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063077870366504e-05,
+      "loss": 1.4116,
+      "step": 5148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062639812181248e-05,
+      "loss": 1.2012,
+      "step": 5149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062201656648382e-05,
+      "loss": 1.4268,
+      "step": 5150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061763403772623e-05,
+      "loss": 0.8604,
+      "step": 5151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061325053558667e-05,
+      "loss": 1.2051,
+      "step": 5152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060886606011227e-05,
+      "loss": 0.9414,
+      "step": 5153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060448061135012e-05,
+      "loss": 1.1426,
+      "step": 5154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060009418934732e-05,
+      "loss": 1.5332,
+      "step": 5155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9059570679415103e-05,
+      "loss": 1.4189,
+      "step": 5156
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905913184258083e-05,
+      "loss": 1.0908,
+      "step": 5157
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058692908436635e-05,
+      "loss": 1.4639,
+      "step": 5158
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058253876987232e-05,
+      "loss": 1.3564,
+      "step": 5159
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057814748237326e-05,
+      "loss": 0.8887,
+      "step": 5160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905737552219165e-05,
+      "loss": 1.0459,
+      "step": 5161
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905693619885491e-05,
+      "loss": 1.4531,
+      "step": 5162
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905649677823183e-05,
+      "loss": 1.0566,
+      "step": 5163
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9056057260327127e-05,
+      "loss": 1.1504,
+      "step": 5164
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905561764514553e-05,
+      "loss": 0.9922,
+      "step": 5165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905517793269175e-05,
+      "loss": 1.4932,
+      "step": 5166
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905473812297052e-05,
+      "loss": 1.0967,
+      "step": 5167
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9054298215986557e-05,
+      "loss": 1.2275,
+      "step": 5168
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053858211744592e-05,
+      "loss": 1.0459,
+      "step": 5169
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053418110249343e-05,
+      "loss": 1.5605,
+      "step": 5170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905297791150555e-05,
+      "loss": 0.9209,
+      "step": 5171
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052537615517933e-05,
+      "loss": 1.2051,
+      "step": 5172
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052097222291222e-05,
+      "loss": 1.2236,
+      "step": 5173
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905165673183015e-05,
+      "loss": 1.0703,
+      "step": 5174
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051216144139446e-05,
+      "loss": 1.4102,
+      "step": 5175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905077545922384e-05,
+      "loss": 1.1396,
+      "step": 5176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050334677088078e-05,
+      "loss": 0.8101,
+      "step": 5177
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049893797736883e-05,
+      "loss": 1.2334,
+      "step": 5178
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049452821174993e-05,
+      "loss": 1.1045,
+      "step": 5179
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049011747407148e-05,
+      "loss": 1.2549,
+      "step": 5180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9048570576438078e-05,
+      "loss": 1.2031,
+      "step": 5181
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9048129308272533e-05,
+      "loss": 1.2349,
+      "step": 5182
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047687942915246e-05,
+      "loss": 1.4043,
+      "step": 5183
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904724648037096e-05,
+      "loss": 1.2061,
+      "step": 5184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9046804920644416e-05,
+      "loss": 1.1025,
+      "step": 5185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904636326374036e-05,
+      "loss": 0.9126,
+      "step": 5186
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045921509663533e-05,
+      "loss": 1.0146,
+      "step": 5187
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904547965841868e-05,
+      "loss": 0.9561,
+      "step": 5188
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045037710010548e-05,
+      "loss": 1.1074,
+      "step": 5189
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044595664443886e-05,
+      "loss": 1.4248,
+      "step": 5190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044153521723436e-05,
+      "loss": 1.3789,
+      "step": 5191
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043711281853956e-05,
+      "loss": 1.2578,
+      "step": 5192
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904326894484019e-05,
+      "loss": 1.2959,
+      "step": 5193
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042826510686895e-05,
+      "loss": 1.25,
+      "step": 5194
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042383979398823e-05,
+      "loss": 1.1426,
+      "step": 5195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904194135098072e-05,
+      "loss": 0.9873,
+      "step": 5196
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041498625437346e-05,
+      "loss": 1.3477,
+      "step": 5197
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904105580277346e-05,
+      "loss": 0.958,
+      "step": 5198
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9040612882993807e-05,
+      "loss": 1.5479,
+      "step": 5199
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904016986610316e-05,
+      "loss": 1.2529,
+      "step": 5200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039726752106264e-05,
+      "loss": 1.3057,
+      "step": 5201
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039283541007888e-05,
+      "loss": 0.9746,
+      "step": 5202
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903884023281279e-05,
+      "loss": 1.4355,
+      "step": 5203
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903839682752573e-05,
+      "loss": 1.1191,
+      "step": 5204
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9037953325151474e-05,
+      "loss": 1.1865,
+      "step": 5205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903750972569478e-05,
+      "loss": 1.3096,
+      "step": 5206
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9037066029160422e-05,
+      "loss": 1.5381,
+      "step": 5207
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9036622235553158e-05,
+      "loss": 1.2773,
+      "step": 5208
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903617834487776e-05,
+      "loss": 1.125,
+      "step": 5209
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035734357139e-05,
+      "loss": 1.3887,
+      "step": 5210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035290272341635e-05,
+      "loss": 1.2676,
+      "step": 5211
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034846090490443e-05,
+      "loss": 1.126,
+      "step": 5212
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034401811590197e-05,
+      "loss": 1.165,
+      "step": 5213
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033957435645664e-05,
+      "loss": 1.3809,
+      "step": 5214
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033512962661618e-05,
+      "loss": 1.1055,
+      "step": 5215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033068392642836e-05,
+      "loss": 1.2324,
+      "step": 5216
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032623725594095e-05,
+      "loss": 1.3066,
+      "step": 5217
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903217896152017e-05,
+      "loss": 1.2783,
+      "step": 5218
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031734100425835e-05,
+      "loss": 1.0288,
+      "step": 5219
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903128914231587e-05,
+      "loss": 1.2324,
+      "step": 5220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030844087195057e-05,
+      "loss": 0.8223,
+      "step": 5221
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030398935068173e-05,
+      "loss": 1.5195,
+      "step": 5222
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9029953685940007e-05,
+      "loss": 1.3457,
+      "step": 5223
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902950833981533e-05,
+      "loss": 1.125,
+      "step": 5224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9029062896698937e-05,
+      "loss": 1.1494,
+      "step": 5225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9028617356595608e-05,
+      "loss": 1.5615,
+      "step": 5226
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9028171719510127e-05,
+      "loss": 0.9194,
+      "step": 5227
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902772598544728e-05,
+      "loss": 1.2754,
+      "step": 5228
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9027280154411865e-05,
+      "loss": 0.8726,
+      "step": 5229
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902683422640866e-05,
+      "loss": 1.2217,
+      "step": 5230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026388201442458e-05,
+      "loss": 1.3936,
+      "step": 5231
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025942079518052e-05,
+      "loss": 1.4248,
+      "step": 5232
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025495860640232e-05,
+      "loss": 1.3848,
+      "step": 5233
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025049544813793e-05,
+      "loss": 1.0049,
+      "step": 5234
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024603132043528e-05,
+      "loss": 1.3438,
+      "step": 5235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024156622334232e-05,
+      "loss": 1.1113,
+      "step": 5236
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90237100156907e-05,
+      "loss": 1.207,
+      "step": 5237
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9023263312117733e-05,
+      "loss": 1.3574,
+      "step": 5238
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022816511620125e-05,
+      "loss": 1.292,
+      "step": 5239
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022369614202677e-05,
+      "loss": 1.0059,
+      "step": 5240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902192261987019e-05,
+      "loss": 0.96,
+      "step": 5241
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902147552862747e-05,
+      "loss": 1.3496,
+      "step": 5242
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021028340479313e-05,
+      "loss": 1.0967,
+      "step": 5243
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9020581055430524e-05,
+      "loss": 1.584,
+      "step": 5244
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9020133673485906e-05,
+      "loss": 1.2388,
+      "step": 5245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019686194650267e-05,
+      "loss": 1.5449,
+      "step": 5246
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019238618928418e-05,
+      "loss": 1.3613,
+      "step": 5247
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901879094632516e-05,
+      "loss": 1.2559,
+      "step": 5248
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90183431768453e-05,
+      "loss": 1.2026,
+      "step": 5249
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901789531049365e-05,
+      "loss": 1.1914,
+      "step": 5250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901744734727503e-05,
+      "loss": 1.3574,
+      "step": 5251
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901699928719424e-05,
+      "loss": 1.3291,
+      "step": 5252
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9016551130256096e-05,
+      "loss": 1.0493,
+      "step": 5253
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9016102876465412e-05,
+      "loss": 1.4111,
+      "step": 5254
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9015654525827007e-05,
+      "loss": 1.3838,
+      "step": 5255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9015206078345693e-05,
+      "loss": 1.4102,
+      "step": 5256
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901475753402629e-05,
+      "loss": 1.1055,
+      "step": 5257
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9014308892873612e-05,
+      "loss": 1.3223,
+      "step": 5258
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9013860154892478e-05,
+      "loss": 1.0586,
+      "step": 5259
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9013411320087717e-05,
+      "loss": 1.3164,
+      "step": 5260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012962388464138e-05,
+      "loss": 1.0098,
+      "step": 5261
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901251336002657e-05,
+      "loss": 0.8867,
+      "step": 5262
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012064234779835e-05,
+      "loss": 1.207,
+      "step": 5263
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011615012728758e-05,
+      "loss": 1.165,
+      "step": 5264
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011165693878165e-05,
+      "loss": 0.9434,
+      "step": 5265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010716278232882e-05,
+      "loss": 1.248,
+      "step": 5266
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010266765797734e-05,
+      "loss": 1.3711,
+      "step": 5267
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9009817156577556e-05,
+      "loss": 1.4023,
+      "step": 5268
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9009367450577168e-05,
+      "loss": 1.1836,
+      "step": 5269
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900891764780141e-05,
+      "loss": 1.1924,
+      "step": 5270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90084677482551e-05,
+      "loss": 1.4053,
+      "step": 5271
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900801775194309e-05,
+      "loss": 1.0801,
+      "step": 5272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9007567658870196e-05,
+      "loss": 1.1191,
+      "step": 5273
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9007117469041265e-05,
+      "loss": 1.3633,
+      "step": 5274
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006667182461128e-05,
+      "loss": 0.9146,
+      "step": 5275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006216799134618e-05,
+      "loss": 1.373,
+      "step": 5276
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9005766319066578e-05,
+      "loss": 1.2031,
+      "step": 5277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900531574226185e-05,
+      "loss": 1.1602,
+      "step": 5278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004865068725264e-05,
+      "loss": 1.3682,
+      "step": 5279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900441429846167e-05,
+      "loss": 1.2393,
+      "step": 5280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003963431475902e-05,
+      "loss": 1.2393,
+      "step": 5281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900351246777281e-05,
+      "loss": 1.0391,
+      "step": 5282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003061407357238e-05,
+      "loss": 0.8799,
+      "step": 5283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002610250234025e-05,
+      "loss": 1.6719,
+      "step": 5284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002158996408025e-05,
+      "loss": 1.0771,
+      "step": 5285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001707645884076e-05,
+      "loss": 1.3867,
+      "step": 5286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001256198667034e-05,
+      "loss": 0.7695,
+      "step": 5287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9000804654761747e-05,
+      "loss": 1.4746,
+      "step": 5288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9000353014173063e-05,
+      "loss": 1.5117,
+      "step": 5289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999901276905835e-05,
+      "loss": 1.8096,
+      "step": 5290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999449442964913e-05,
+      "loss": 1.1094,
+      "step": 5291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998997512355153e-05,
+      "loss": 1.1104,
+      "step": 5292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998545485081408e-05,
+      "loss": 1.2246,
+      "step": 5293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998093361148534e-05,
+      "loss": 1.0283,
+      "step": 5294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997641140561392e-05,
+      "loss": 1.125,
+      "step": 5295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899718882332483e-05,
+      "loss": 1.1367,
+      "step": 5296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996736409443715e-05,
+      "loss": 1.1514,
+      "step": 5297
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89962838989229e-05,
+      "loss": 1.4082,
+      "step": 5298
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899583129176725e-05,
+      "loss": 1.1768,
+      "step": 5299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899537858798163e-05,
+      "loss": 1.3711,
+      "step": 5300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994925787570894e-05,
+      "loss": 0.7803,
+      "step": 5301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994472890539913e-05,
+      "loss": 1.167,
+      "step": 5302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994019896893552e-05,
+      "loss": 1.0483,
+      "step": 5303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993566806636673e-05,
+      "loss": 1.4629,
+      "step": 5304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899311361977414e-05,
+      "loss": 1.2354,
+      "step": 5305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899266033631083e-05,
+      "loss": 1.0234,
+      "step": 5306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992206956251605e-05,
+      "loss": 1.373,
+      "step": 5307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899175347960134e-05,
+      "loss": 1.4414,
+      "step": 5308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89912999063649e-05,
+      "loss": 1.3125,
+      "step": 5309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899084623654716e-05,
+      "loss": 1.2598,
+      "step": 5310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8990392470152998e-05,
+      "loss": 0.9834,
+      "step": 5311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898993860718728e-05,
+      "loss": 0.9072,
+      "step": 5312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8989484647654888e-05,
+      "loss": 1.4316,
+      "step": 5313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8989030591560695e-05,
+      "loss": 1.041,
+      "step": 5314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988576438909577e-05,
+      "loss": 1.165,
+      "step": 5315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988122189706417e-05,
+      "loss": 0.9404,
+      "step": 5316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987667843956087e-05,
+      "loss": 1.0791,
+      "step": 5317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987213401663477e-05,
+      "loss": 0.9043,
+      "step": 5318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898675886283346e-05,
+      "loss": 1.002,
+      "step": 5319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898630422747092e-05,
+      "loss": 1.0898,
+      "step": 5320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8985849495580747e-05,
+      "loss": 1.0859,
+      "step": 5321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8985394667167818e-05,
+      "loss": 0.8682,
+      "step": 5322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898493974223702e-05,
+      "loss": 1.2197,
+      "step": 5323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8984484720793244e-05,
+      "loss": 0.8809,
+      "step": 5324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8984029602841378e-05,
+      "loss": 0.9194,
+      "step": 5325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89835743883863e-05,
+      "loss": 1.7129,
+      "step": 5326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898311907743291e-05,
+      "loss": 0.9053,
+      "step": 5327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89826636699861e-05,
+      "loss": 1.2773,
+      "step": 5328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898220816605075e-05,
+      "loss": 1.3125,
+      "step": 5329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981752565631767e-05,
+      "loss": 1.3252,
+      "step": 5330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898129686873404e-05,
+      "loss": 0.7881,
+      "step": 5331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980841075362458e-05,
+      "loss": 0.9727,
+      "step": 5332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898038518552192e-05,
+      "loss": 1.2148,
+      "step": 5333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897992919921733e-05,
+      "loss": 1.292,
+      "step": 5334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897947311645358e-05,
+      "loss": 1.1914,
+      "step": 5335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979016937235566e-05,
+      "loss": 0.8691,
+      "step": 5336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8978560661568194e-05,
+      "loss": 1.4082,
+      "step": 5337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8978104289456364e-05,
+      "loss": 1.4004,
+      "step": 5338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977647820904974e-05,
+      "loss": 0.9941,
+      "step": 5339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977191255918934e-05,
+      "loss": 1.1758,
+      "step": 5340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976734594503143e-05,
+      "loss": 0.8926,
+      "step": 5341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976277836662506e-05,
+      "loss": 0.937,
+      "step": 5342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8975820982401935e-05,
+      "loss": 0.9951,
+      "step": 5343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897536403172633e-05,
+      "loss": 0.7451,
+      "step": 5344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974906984640602e-05,
+      "loss": 1.1543,
+      "step": 5345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974449841149663e-05,
+      "loss": 1.2207,
+      "step": 5346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973992601258423e-05,
+      "loss": 1.4727,
+      "step": 5347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897353526497179e-05,
+      "loss": 1.2031,
+      "step": 5348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973077832294682e-05,
+      "loss": 1.6602,
+      "step": 5349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897262030323201e-05,
+      "loss": 1.1143,
+      "step": 5350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8972162677788684e-05,
+      "loss": 1.2598,
+      "step": 5351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897170495596963e-05,
+      "loss": 1.2246,
+      "step": 5352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971247137779752e-05,
+      "loss": 0.8682,
+      "step": 5353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970789223223978e-05,
+      "loss": 1.4805,
+      "step": 5354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897033121230722e-05,
+      "loss": 0.9956,
+      "step": 5355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969873105034402e-05,
+      "loss": 1.4424,
+      "step": 5356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969414901410443e-05,
+      "loss": 1.1992,
+      "step": 5357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8968956601440268e-05,
+      "loss": 0.79,
+      "step": 5358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8968498205128792e-05,
+      "loss": 1.4033,
+      "step": 5359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896803971248095e-05,
+      "loss": 1.3574,
+      "step": 5360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896758112350166e-05,
+      "loss": 1.0732,
+      "step": 5361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8967122438195846e-05,
+      "loss": 1.167,
+      "step": 5362
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896666365656844e-05,
+      "loss": 1.2832,
+      "step": 5363
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896620477862437e-05,
+      "loss": 1.3398,
+      "step": 5364
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896574580436856e-05,
+      "loss": 1.2803,
+      "step": 5365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965286733805945e-05,
+      "loss": 1.3203,
+      "step": 5366
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964827566941455e-05,
+      "loss": 1.2598,
+      "step": 5367
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896436830378002e-05,
+      "loss": 1.4082,
+      "step": 5368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963908944326575e-05,
+      "loss": 1.0698,
+      "step": 5369
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896344948858606e-05,
+      "loss": 1.2285,
+      "step": 5370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89629899365634e-05,
+      "loss": 1.4121,
+      "step": 5371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962530288263536e-05,
+      "loss": 1.2324,
+      "step": 5372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962070543691408e-05,
+      "loss": 0.8047,
+      "step": 5373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896161070285195e-05,
+      "loss": 1.0273,
+      "step": 5374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961150765750104e-05,
+      "loss": 1.377,
+      "step": 5375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8960690732390808e-05,
+      "loss": 1.3799,
+      "step": 5376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8960230602779007e-05,
+      "loss": 1.0527,
+      "step": 5377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895977037691964e-05,
+      "loss": 1.4268,
+      "step": 5378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8959310054817652e-05,
+      "loss": 1.2344,
+      "step": 5379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895884963647799e-05,
+      "loss": 1.1748,
+      "step": 5380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8958389121905596e-05,
+      "loss": 1.0156,
+      "step": 5381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895792851110542e-05,
+      "loss": 1.293,
+      "step": 5382
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895746780408241e-05,
+      "loss": 1.3662,
+      "step": 5383
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8957007000841508e-05,
+      "loss": 0.9141,
+      "step": 5384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895654610138767e-05,
+      "loss": 1.3955,
+      "step": 5385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956085105725845e-05,
+      "loss": 1.1895,
+      "step": 5386
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8955624013860987e-05,
+      "loss": 1.3047,
+      "step": 5387
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895516282579805e-05,
+      "loss": 1.251,
+      "step": 5388
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895470154154198e-05,
+      "loss": 1.5654,
+      "step": 5389
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8954240161097737e-05,
+      "loss": 1.5449,
+      "step": 5390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895377868447028e-05,
+      "loss": 0.9863,
+      "step": 5391
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895331711166456e-05,
+      "loss": 1.0078,
+      "step": 5392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952855442685537e-05,
+      "loss": 1.2119,
+      "step": 5393
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952393677538174e-05,
+      "loss": 1.2422,
+      "step": 5394
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895193181622743e-05,
+      "loss": 1.0063,
+      "step": 5395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895146985875826e-05,
+      "loss": 0.937,
+      "step": 5396
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8951007805135636e-05,
+      "loss": 1.3555,
+      "step": 5397
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895054565536451e-05,
+      "loss": 1.0273,
+      "step": 5398
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8950083409449855e-05,
+      "loss": 1.3418,
+      "step": 5399
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894962106739663e-05,
+      "loss": 1.1748,
+      "step": 5400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894915862920981e-05,
+      "loss": 1.3213,
+      "step": 5401
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948696094894354e-05,
+      "loss": 1.0996,
+      "step": 5402
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948233464455232e-05,
+      "loss": 1.2188,
+      "step": 5403
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947770737897414e-05,
+      "loss": 1.2305,
+      "step": 5404
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947307915225876e-05,
+      "loss": 1.4785,
+      "step": 5405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894684499644558e-05,
+      "loss": 1.3521,
+      "step": 5406
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894638198156151e-05,
+      "loss": 1.2852,
+      "step": 5407
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945918870578627e-05,
+      "loss": 1.3438,
+      "step": 5408
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945455663501908e-05,
+      "loss": 0.9521,
+      "step": 5409
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8944992360336337e-05,
+      "loss": 1.4551,
+      "step": 5410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8944528961086887e-05,
+      "loss": 1.4766,
+      "step": 5411
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894406546575853e-05,
+      "loss": 1.4082,
+      "step": 5412
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894360187435625e-05,
+      "loss": 1.1738,
+      "step": 5413
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8943138186885025e-05,
+      "loss": 1.0972,
+      "step": 5414
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894267440334984e-05,
+      "loss": 1.2832,
+      "step": 5415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8942210523755668e-05,
+      "loss": 0.8984,
+      "step": 5416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89417465481075e-05,
+      "loss": 1.4004,
+      "step": 5417
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8941282476410315e-05,
+      "loss": 1.1807,
+      "step": 5418
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940818308669105e-05,
+      "loss": 1.3389,
+      "step": 5419
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940354044888843e-05,
+      "loss": 1.2344,
+      "step": 5420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893988968507453e-05,
+      "loss": 1.1621,
+      "step": 5421
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8939425229231143e-05,
+      "loss": 1.5449,
+      "step": 5422
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893896067736368e-05,
+      "loss": 1.4277,
+      "step": 5423
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8938496029477123e-05,
+      "loss": 0.7925,
+      "step": 5424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8938031285576466e-05,
+      "loss": 1.1904,
+      "step": 5425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8937566445666707e-05,
+      "loss": 0.9951,
+      "step": 5426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8937101509752828e-05,
+      "loss": 1.3301,
+      "step": 5427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893663647783983e-05,
+      "loss": 1.3105,
+      "step": 5428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8936171349932713e-05,
+      "loss": 0.8584,
+      "step": 5429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893570612603646e-05,
+      "loss": 1.1562,
+      "step": 5430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935240806156083e-05,
+      "loss": 0.8564,
+      "step": 5431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893477539029657e-05,
+      "loss": 1.3271,
+      "step": 5432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934309878462925e-05,
+      "loss": 1.166,
+      "step": 5433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933844270660144e-05,
+      "loss": 1.2871,
+      "step": 5434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893337856689323e-05,
+      "loss": 1.5625,
+      "step": 5435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932912767167192e-05,
+      "loss": 1.0947,
+      "step": 5436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932446871487023e-05,
+      "loss": 1.5449,
+      "step": 5437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931980879857737e-05,
+      "loss": 1.5527,
+      "step": 5438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893151479228433e-05,
+      "loss": 1.3271,
+      "step": 5439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893104860877182e-05,
+      "loss": 1.1411,
+      "step": 5440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930582329325206e-05,
+      "loss": 1.1348,
+      "step": 5441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930115953949496e-05,
+      "loss": 1.0439,
+      "step": 5442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929649482649704e-05,
+      "loss": 1.1787,
+      "step": 5443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892918291543084e-05,
+      "loss": 1.2031,
+      "step": 5444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928716252297912e-05,
+      "loss": 1.4277,
+      "step": 5445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892824949325594e-05,
+      "loss": 1.3643,
+      "step": 5446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892778263830993e-05,
+      "loss": 0.9453,
+      "step": 5447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89273156874649e-05,
+      "loss": 1.0117,
+      "step": 5448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892684864072587e-05,
+      "loss": 1.4629,
+      "step": 5449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892638149809785e-05,
+      "loss": 1.2275,
+      "step": 5450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925914259585857e-05,
+      "loss": 1.3086,
+      "step": 5451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892544692519492e-05,
+      "loss": 1.4473,
+      "step": 5452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892497949493005e-05,
+      "loss": 1.4639,
+      "step": 5453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892451196879627e-05,
+      "loss": 1.0059,
+      "step": 5454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8924044346798603e-05,
+      "loss": 1.126,
+      "step": 5455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923576628942072e-05,
+      "loss": 0.9463,
+      "step": 5456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89231088152317e-05,
+      "loss": 0.9878,
+      "step": 5457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892264090567251e-05,
+      "loss": 1.1333,
+      "step": 5458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8922172900269534e-05,
+      "loss": 1.0684,
+      "step": 5459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921704799027795e-05,
+      "loss": 1.0566,
+      "step": 5460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921236601952322e-05,
+      "loss": 1.3003,
+      "step": 5461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920768309048142e-05,
+      "loss": 1.0366,
+      "step": 5462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892029992032029e-05,
+      "loss": 1.416,
+      "step": 5463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919831435773795e-05,
+      "loss": 0.9336,
+      "step": 5464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891936285541369e-05,
+      "loss": 0.9961,
+      "step": 5465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8918894179245006e-05,
+      "loss": 1.416,
+      "step": 5466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891842540727278e-05,
+      "loss": 1.3281,
+      "step": 5467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8917956539502044e-05,
+      "loss": 0.9146,
+      "step": 5468
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8917487575937837e-05,
+      "loss": 1.1895,
+      "step": 5469
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8917018516585196e-05,
+      "loss": 1.125,
+      "step": 5470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891654936144916e-05,
+      "loss": 1.1895,
+      "step": 5471
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891608011053477e-05,
+      "loss": 1.1104,
+      "step": 5472
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891561076384706e-05,
+      "loss": 1.4121,
+      "step": 5473
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915141321391083e-05,
+      "loss": 1.084,
+      "step": 5474
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914671783171868e-05,
+      "loss": 1.6484,
+      "step": 5475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914202149194464e-05,
+      "loss": 0.8643,
+      "step": 5476
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891373241946392e-05,
+      "loss": 1.4629,
+      "step": 5477
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891326259398528e-05,
+      "loss": 1.3662,
+      "step": 5478
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891279267276359e-05,
+      "loss": 1.021,
+      "step": 5479
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891232265580389e-05,
+      "loss": 0.8564,
+      "step": 5480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8911852543111244e-05,
+      "loss": 0.7988,
+      "step": 5481
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8911382334690687e-05,
+      "loss": 0.9761,
+      "step": 5482
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8910912030547283e-05,
+      "loss": 1.8242,
+      "step": 5483
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8910441630686072e-05,
+      "loss": 1.0791,
+      "step": 5484
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890997113511211e-05,
+      "loss": 1.1006,
+      "step": 5485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909500543830456e-05,
+      "loss": 1.1621,
+      "step": 5486
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909029856846163e-05,
+      "loss": 1.2148,
+      "step": 5487
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908559074164282e-05,
+      "loss": 0.897,
+      "step": 5488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908088195789877e-05,
+      "loss": 1.0576,
+      "step": 5489
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8907617221728003e-05,
+      "loss": 1.1436,
+      "step": 5490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8907146151983718e-05,
+      "loss": 1.5254,
+      "step": 5491
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890667498656208e-05,
+      "loss": 1.2783,
+      "step": 5492
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8906203725468154e-05,
+      "loss": 1.3623,
+      "step": 5493
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905732368707003e-05,
+      "loss": 1.3896,
+      "step": 5494
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890526091628369e-05,
+      "loss": 1.293,
+      "step": 5495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8904789368203273e-05,
+      "loss": 0.8818,
+      "step": 5496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8904317724470825e-05,
+      "loss": 0.959,
+      "step": 5497
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903845985091406e-05,
+      "loss": 0.9883,
+      "step": 5498
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903374150070086e-05,
+      "loss": 1.2217,
+      "step": 5499
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902902219411935e-05,
+      "loss": 1.2314,
+      "step": 5500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890243019312202e-05,
+      "loss": 0.9844,
+      "step": 5501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890195807120541e-05,
+      "loss": 1.0146,
+      "step": 5502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8901485853667185e-05,
+      "loss": 1.1729,
+      "step": 5503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8901013540512404e-05,
+      "loss": 1.3354,
+      "step": 5504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8900541131746148e-05,
+      "loss": 1.3008,
+      "step": 5505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890006862737349e-05,
+      "loss": 1.2603,
+      "step": 5506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889959602739951e-05,
+      "loss": 0.8462,
+      "step": 5507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889912333182928e-05,
+      "loss": 1.2539,
+      "step": 5508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889865054066787e-05,
+      "loss": 0.8882,
+      "step": 5509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898177653920374e-05,
+      "loss": 1.2734,
+      "step": 5510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8897704671591866e-05,
+      "loss": 1.4902,
+      "step": 5511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889723159368742e-05,
+      "loss": 1.3203,
+      "step": 5512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8896758420212124e-05,
+      "loss": 1.2764,
+      "step": 5513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889628515117106e-05,
+      "loss": 0.9814,
+      "step": 5514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889581178656931e-05,
+      "loss": 0.8945,
+      "step": 5515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889533832641196e-05,
+      "loss": 1.2256,
+      "step": 5516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8894864770704094e-05,
+      "loss": 1.291,
+      "step": 5517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88943911194508e-05,
+      "loss": 1.314,
+      "step": 5518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893917372657165e-05,
+      "loss": 1.4746,
+      "step": 5519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889344353032828e-05,
+      "loss": 1.1768,
+      "step": 5520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892969592469235e-05,
+      "loss": 1.0918,
+      "step": 5521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889249555908512e-05,
+      "loss": 0.959,
+      "step": 5522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889202143018102e-05,
+      "loss": 0.9326,
+      "step": 5523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889154720576204e-05,
+      "loss": 1.4648,
+      "step": 5524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8891072885833263e-05,
+      "loss": 1.1133,
+      "step": 5525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889059847039979e-05,
+      "loss": 1.1055,
+      "step": 5526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890123959466722e-05,
+      "loss": 1.458,
+      "step": 5527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8889649353039145e-05,
+      "loss": 1.0469,
+      "step": 5528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8889174651122162e-05,
+      "loss": 0.9688,
+      "step": 5529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888699853720877e-05,
+      "loss": 0.7617,
+      "step": 5530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888822496084038e-05,
+      "loss": 1.2617,
+      "step": 5531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887749972485777e-05,
+      "loss": 1.0034,
+      "step": 5532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887274888662174e-05,
+      "loss": 1.4961,
+      "step": 5533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888679970937467e-05,
+      "loss": 0.9619,
+      "step": 5534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888632443462837e-05,
+      "loss": 1.4688,
+      "step": 5535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8885849064428377e-05,
+      "loss": 1.3066,
+      "step": 5536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8885373598779805e-05,
+      "loss": 1.1123,
+      "step": 5537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888489803768775e-05,
+      "loss": 1.2754,
+      "step": 5538
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8884422381157327e-05,
+      "loss": 1.2012,
+      "step": 5539
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8883946629193644e-05,
+      "loss": 1.3027,
+      "step": 5540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8883470781801813e-05,
+      "loss": 1.1289,
+      "step": 5541
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882994838986946e-05,
+      "loss": 1.4014,
+      "step": 5542
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888251880075415e-05,
+      "loss": 0.9053,
+      "step": 5543
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882042667108544e-05,
+      "loss": 1.0508,
+      "step": 5544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881566438055238e-05,
+      "loss": 0.8379,
+      "step": 5545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881090113599353e-05,
+      "loss": 1.1816,
+      "step": 5546
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8880613693746e-05,
+      "loss": 1.168,
+      "step": 5547
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88801371785003e-05,
+      "loss": 1.2861,
+      "step": 5548
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887966056786737e-05,
+      "loss": 0.917,
+      "step": 5549
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887918386185233e-05,
+      "loss": 0.9805,
+      "step": 5550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8878707060460302e-05,
+      "loss": 0.8955,
+      "step": 5551
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8878230163696405e-05,
+      "loss": 1.3066,
+      "step": 5552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8877753171565763e-05,
+      "loss": 1.1895,
+      "step": 5553
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88772760840735e-05,
+      "loss": 0.9551,
+      "step": 5554
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887679890122474e-05,
+      "loss": 1.2607,
+      "step": 5555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887632162302461e-05,
+      "loss": 1.4707,
+      "step": 5556
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8875844249478234e-05,
+      "loss": 1.0596,
+      "step": 5557
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8875366780590744e-05,
+      "loss": 1.3223,
+      "step": 5558
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874889216367266e-05,
+      "loss": 0.9541,
+      "step": 5559
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874411556812927e-05,
+      "loss": 1.0879,
+      "step": 5560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873933801932864e-05,
+      "loss": 1.0156,
+      "step": 5561
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873455951732205e-05,
+      "loss": 1.0557,
+      "step": 5562
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887297800621608e-05,
+      "loss": 1.0791,
+      "step": 5563
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8872499965389633e-05,
+      "loss": 1.4414,
+      "step": 5564
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887202182925799e-05,
+      "loss": 1.167,
+      "step": 5565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887154359782629e-05,
+      "loss": 1.0532,
+      "step": 5566
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8871065271099668e-05,
+      "loss": 0.9033,
+      "step": 5567
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870586849083265e-05,
+      "loss": 1.0029,
+      "step": 5568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887010833178222e-05,
+      "loss": 1.0459,
+      "step": 5569
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8869629719201668e-05,
+      "loss": 0.9678,
+      "step": 5570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8869151011346757e-05,
+      "loss": 1.1855,
+      "step": 5571
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868672208222624e-05,
+      "loss": 1.293,
+      "step": 5572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868193309834414e-05,
+      "loss": 1.0576,
+      "step": 5573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867714316187273e-05,
+      "loss": 1.4668,
+      "step": 5574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867235227286343e-05,
+      "loss": 1.3164,
+      "step": 5575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8866756043136772e-05,
+      "loss": 0.6655,
+      "step": 5576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8866276763743707e-05,
+      "loss": 1.3237,
+      "step": 5577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865797389112296e-05,
+      "loss": 1.334,
+      "step": 5578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865317919247687e-05,
+      "loss": 1.2246,
+      "step": 5579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8864838354155033e-05,
+      "loss": 0.8818,
+      "step": 5580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8864358693839487e-05,
+      "loss": 1.208,
+      "step": 5581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8863878938306193e-05,
+      "loss": 1.2861,
+      "step": 5582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886339908756031e-05,
+      "loss": 1.6289,
+      "step": 5583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8862919141606996e-05,
+      "loss": 1.0996,
+      "step": 5584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8862439100451397e-05,
+      "loss": 0.6265,
+      "step": 5585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886195896409868e-05,
+      "loss": 1.3535,
+      "step": 5586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861478732553994e-05,
+      "loss": 1.1084,
+      "step": 5587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860998405822503e-05,
+      "loss": 1.1982,
+      "step": 5588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886051798390936e-05,
+      "loss": 1.0518,
+      "step": 5589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860037466819732e-05,
+      "loss": 1.1025,
+      "step": 5590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8859556854558782e-05,
+      "loss": 0.853,
+      "step": 5591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8859076147131667e-05,
+      "loss": 1.3081,
+      "step": 5592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8858595344543552e-05,
+      "loss": 0.9033,
+      "step": 5593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88581144467996e-05,
+      "loss": 1.3926,
+      "step": 5594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885763345390498e-05,
+      "loss": 1.459,
+      "step": 5595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857152365864858e-05,
+      "loss": 1.543,
+      "step": 5596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88566711826844e-05,
+      "loss": 1.3105,
+      "step": 5597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856189904368774e-05,
+      "loss": 0.9775,
+      "step": 5598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8855708530923157e-05,
+      "loss": 1.2637,
+      "step": 5599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8855227062352707e-05,
+      "loss": 1.1895,
+      "step": 5600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885474549866261e-05,
+      "loss": 1.1006,
+      "step": 5601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8854263839858028e-05,
+      "loss": 1.6797,
+      "step": 5602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853782085944137e-05,
+      "loss": 1.0947,
+      "step": 5603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853300236926118e-05,
+      "loss": 1.0156,
+      "step": 5604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885281829280914e-05,
+      "loss": 1.1396,
+      "step": 5605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885233625359838e-05,
+      "loss": 1.0073,
+      "step": 5606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885185411929902e-05,
+      "loss": 0.9941,
+      "step": 5607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851371889916238e-05,
+      "loss": 0.9883,
+      "step": 5608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850889565455207e-05,
+      "loss": 1.3281,
+      "step": 5609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850407145921118e-05,
+      "loss": 1.5898,
+      "step": 5610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884992463131915e-05,
+      "loss": 1.2246,
+      "step": 5611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849442021654486e-05,
+      "loss": 1.4326,
+      "step": 5612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848959316932305e-05,
+      "loss": 0.9028,
+      "step": 5613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848476517157797e-05,
+      "loss": 1.3125,
+      "step": 5614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847993622336147e-05,
+      "loss": 1.2041,
+      "step": 5615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847510632472542e-05,
+      "loss": 1.0225,
+      "step": 5616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884702754757217e-05,
+      "loss": 1.4863,
+      "step": 5617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8846544367640218e-05,
+      "loss": 1.374,
+      "step": 5618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8846061092681883e-05,
+      "loss": 0.9775,
+      "step": 5619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845577722702346e-05,
+      "loss": 1.3081,
+      "step": 5620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884509425770681e-05,
+      "loss": 1.5195,
+      "step": 5621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884461069770046e-05,
+      "loss": 0.9453,
+      "step": 5622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844127042688494e-05,
+      "loss": 1.1133,
+      "step": 5623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843643292676105e-05,
+      "loss": 0.9385,
+      "step": 5624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884315944766849e-05,
+      "loss": 1.3154,
+      "step": 5625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884267550767085e-05,
+      "loss": 1.3662,
+      "step": 5626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884219147268838e-05,
+      "loss": 1.3516,
+      "step": 5627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841707342726277e-05,
+      "loss": 1.1074,
+      "step": 5628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841223117789745e-05,
+      "loss": 1.2715,
+      "step": 5629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884073879788398e-05,
+      "loss": 1.1143,
+      "step": 5630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840254383014195e-05,
+      "loss": 1.5215,
+      "step": 5631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839769873185588e-05,
+      "loss": 1.1396,
+      "step": 5632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883928526840336e-05,
+      "loss": 1.2852,
+      "step": 5633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838800568672716e-05,
+      "loss": 1.1406,
+      "step": 5634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838315773998866e-05,
+      "loss": 1.2285,
+      "step": 5635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837830884387022e-05,
+      "loss": 1.4355,
+      "step": 5636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883734589984238e-05,
+      "loss": 1.5176,
+      "step": 5637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836860820370162e-05,
+      "loss": 1.0918,
+      "step": 5638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836375645975572e-05,
+      "loss": 0.9209,
+      "step": 5639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8835890376663825e-05,
+      "loss": 1.2744,
+      "step": 5640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8835405012440132e-05,
+      "loss": 1.3867,
+      "step": 5641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834919553309704e-05,
+      "loss": 1.25,
+      "step": 5642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834433999277755e-05,
+      "loss": 1.1523,
+      "step": 5643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833948350349507e-05,
+      "loss": 1.1821,
+      "step": 5644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833462606530174e-05,
+      "loss": 1.1738,
+      "step": 5645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883297676782497e-05,
+      "loss": 1.1001,
+      "step": 5646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883249083423912e-05,
+      "loss": 0.9727,
+      "step": 5647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832004805777838e-05,
+      "loss": 1.0527,
+      "step": 5648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831518682446348e-05,
+      "loss": 1.2412,
+      "step": 5649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831032464249872e-05,
+      "loss": 0.8193,
+      "step": 5650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883054615119363e-05,
+      "loss": 1.2422,
+      "step": 5651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883005974328285e-05,
+      "loss": 1.2402,
+      "step": 5652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829573240522747e-05,
+      "loss": 1.3555,
+      "step": 5653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882908664291856e-05,
+      "loss": 1.0889,
+      "step": 5654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882859995047551e-05,
+      "loss": 0.9014,
+      "step": 5655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828113163198823e-05,
+      "loss": 1.1992,
+      "step": 5656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882762628109373e-05,
+      "loss": 1.0527,
+      "step": 5657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882713930416546e-05,
+      "loss": 1.3286,
+      "step": 5658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826652232419243e-05,
+      "loss": 1.3525,
+      "step": 5659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826165065860315e-05,
+      "loss": 1.2627,
+      "step": 5660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825677804493908e-05,
+      "loss": 1.4004,
+      "step": 5661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825190448325252e-05,
+      "loss": 0.9575,
+      "step": 5662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824702997359586e-05,
+      "loss": 1.2354,
+      "step": 5663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882421545160214e-05,
+      "loss": 0.8877,
+      "step": 5664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823727811058158e-05,
+      "loss": 1.124,
+      "step": 5665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882324007573288e-05,
+      "loss": 1.3965,
+      "step": 5666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822752245631536e-05,
+      "loss": 1.0723,
+      "step": 5667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822264320759372e-05,
+      "loss": 1.4512,
+      "step": 5668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821776301121627e-05,
+      "loss": 0.9395,
+      "step": 5669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821288186723544e-05,
+      "loss": 0.8477,
+      "step": 5670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882079997757037e-05,
+      "loss": 1.1748,
+      "step": 5671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820311673667336e-05,
+      "loss": 1.0293,
+      "step": 5672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819823275019704e-05,
+      "loss": 1.2749,
+      "step": 5673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881933478163271e-05,
+      "loss": 1.2861,
+      "step": 5674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818846193511605e-05,
+      "loss": 1.1357,
+      "step": 5675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818357510661637e-05,
+      "loss": 1.0664,
+      "step": 5676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817868733088053e-05,
+      "loss": 1.124,
+      "step": 5677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817379860796104e-05,
+      "loss": 1.1309,
+      "step": 5678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881689089379104e-05,
+      "loss": 1.5938,
+      "step": 5679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881640183207812e-05,
+      "loss": 1.2988,
+      "step": 5680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881591267566259e-05,
+      "loss": 1.1914,
+      "step": 5681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815423424549707e-05,
+      "loss": 1.0303,
+      "step": 5682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814934078744728e-05,
+      "loss": 1.207,
+      "step": 5683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814444638252903e-05,
+      "loss": 1.1943,
+      "step": 5684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88139551030795e-05,
+      "loss": 1.6621,
+      "step": 5685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813465473229765e-05,
+      "loss": 0.9414,
+      "step": 5686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881297574870897e-05,
+      "loss": 0.9141,
+      "step": 5687
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812485929522368e-05,
+      "loss": 1.0879,
+      "step": 5688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811996015675217e-05,
+      "loss": 1.0439,
+      "step": 5689
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881150600717279e-05,
+      "loss": 1.2314,
+      "step": 5690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811015904020342e-05,
+      "loss": 0.8501,
+      "step": 5691
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881052570622314e-05,
+      "loss": 1.4746,
+      "step": 5692
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881003541378645e-05,
+      "loss": 1.5176,
+      "step": 5693
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880954502671554e-05,
+      "loss": 1.0918,
+      "step": 5694
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809054545015676e-05,
+      "loss": 0.6787,
+      "step": 5695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808563968692124e-05,
+      "loss": 1.1963,
+      "step": 5696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808073297750157e-05,
+      "loss": 0.8926,
+      "step": 5697
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807582532195046e-05,
+      "loss": 1.3281,
+      "step": 5698
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880709167203206e-05,
+      "loss": 1.3496,
+      "step": 5699
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806600717266474e-05,
+      "loss": 1.1543,
+      "step": 5700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880610966790356e-05,
+      "loss": 1.1504,
+      "step": 5701
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880561852394859e-05,
+      "loss": 1.0361,
+      "step": 5702
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805127285406848e-05,
+      "loss": 1.1484,
+      "step": 5703
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804635952283603e-05,
+      "loss": 1.0283,
+      "step": 5704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804144524584135e-05,
+      "loss": 1.291,
+      "step": 5705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803653002313724e-05,
+      "loss": 1.1992,
+      "step": 5706
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880316138547765e-05,
+      "loss": 1.0376,
+      "step": 5707
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880266967408119e-05,
+      "loss": 1.1299,
+      "step": 5708
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880217786812963e-05,
+      "loss": 1.4609,
+      "step": 5709
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801685967628252e-05,
+      "loss": 1.6309,
+      "step": 5710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880119397258234e-05,
+      "loss": 1.3535,
+      "step": 5711
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800701882997177e-05,
+      "loss": 0.999,
+      "step": 5712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800209698878046e-05,
+      "loss": 1.5391,
+      "step": 5713
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799717420230242e-05,
+      "loss": 1.0771,
+      "step": 5714
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799225047059046e-05,
+      "loss": 1.207,
+      "step": 5715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879873257936975e-05,
+      "loss": 0.7495,
+      "step": 5716
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798240017167645e-05,
+      "loss": 0.9639,
+      "step": 5717
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797747360458018e-05,
+      "loss": 1.0469,
+      "step": 5718
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797254609246165e-05,
+      "loss": 1.2393,
+      "step": 5719
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796761763537375e-05,
+      "loss": 1.2285,
+      "step": 5720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796268823336947e-05,
+      "loss": 1.062,
+      "step": 5721
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795775788650174e-05,
+      "loss": 1.7617,
+      "step": 5722
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795282659482346e-05,
+      "loss": 1.0415,
+      "step": 5723
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879478943583877e-05,
+      "loss": 1.2236,
+      "step": 5724
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879429611772473e-05,
+      "loss": 1.4434,
+      "step": 5725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793802705145542e-05,
+      "loss": 1.2246,
+      "step": 5726
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793309198106498e-05,
+      "loss": 1.333,
+      "step": 5727
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87928155966129e-05,
+      "loss": 1.5342,
+      "step": 5728
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879232190067005e-05,
+      "loss": 1.2588,
+      "step": 5729
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791828110283244e-05,
+      "loss": 1.3135,
+      "step": 5730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87913342254578e-05,
+      "loss": 1.6113,
+      "step": 5731
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790840246199012e-05,
+      "loss": 1.4785,
+      "step": 5732
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879034617251219e-05,
+      "loss": 1.6191,
+      "step": 5733
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789852004402643e-05,
+      "loss": 1.3887,
+      "step": 5734
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789357741875673e-05,
+      "loss": 1.2168,
+      "step": 5735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788863384936594e-05,
+      "loss": 0.9199,
+      "step": 5736
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878836893359072e-05,
+      "loss": 1.1436,
+      "step": 5737
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787874387843358e-05,
+      "loss": 0.9609,
+      "step": 5738
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787379747699813e-05,
+      "loss": 0.8672,
+      "step": 5739
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786885013165413e-05,
+      "loss": 1.2715,
+      "step": 5740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878639018424546e-05,
+      "loss": 1.2764,
+      "step": 5741
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785895260945276e-05,
+      "loss": 0.9258,
+      "step": 5742
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785400243270175e-05,
+      "loss": 1.5742,
+      "step": 5743
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784905131225472e-05,
+      "loss": 1.3252,
+      "step": 5744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784409924816488e-05,
+      "loss": 1.5234,
+      "step": 5745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783914624048546e-05,
+      "loss": 1.0801,
+      "step": 5746
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878341922892696e-05,
+      "loss": 1.4424,
+      "step": 5747
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878292373945705e-05,
+      "loss": 1.0244,
+      "step": 5748
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878242815564415e-05,
+      "loss": 1.3301,
+      "step": 5749
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878193247749357e-05,
+      "loss": 1.416,
+      "step": 5750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781436705010643e-05,
+      "loss": 0.7539,
+      "step": 5751
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780940838200692e-05,
+      "loss": 1.0928,
+      "step": 5752
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780444877069038e-05,
+      "loss": 1.4346,
+      "step": 5753
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779948821621018e-05,
+      "loss": 1.2354,
+      "step": 5754
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877945267186195e-05,
+      "loss": 1.2246,
+      "step": 5755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778956427797175e-05,
+      "loss": 1.2539,
+      "step": 5756
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778460089432012e-05,
+      "loss": 1.2949,
+      "step": 5757
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87779636567718e-05,
+      "loss": 1.1318,
+      "step": 5758
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777467129821874e-05,
+      "loss": 1.2852,
+      "step": 5759
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776970508587558e-05,
+      "loss": 1.0596,
+      "step": 5760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776473793074196e-05,
+      "loss": 1.3438,
+      "step": 5761
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8775976983287117e-05,
+      "loss": 1.4209,
+      "step": 5762
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8775480079231657e-05,
+      "loss": 1.4053,
+      "step": 5763
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774983080913157e-05,
+      "loss": 1.4238,
+      "step": 5764
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877448598833696e-05,
+      "loss": 1.3633,
+      "step": 5765
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773988801508396e-05,
+      "loss": 1.0566,
+      "step": 5766
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877349152043281e-05,
+      "loss": 1.1807,
+      "step": 5767
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772994145115543e-05,
+      "loss": 1.291,
+      "step": 5768
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877249667556194e-05,
+      "loss": 1.3936,
+      "step": 5769
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771999111777344e-05,
+      "loss": 1.2515,
+      "step": 5770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771501453767097e-05,
+      "loss": 1.2451,
+      "step": 5771
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771003701536545e-05,
+      "loss": 0.8867,
+      "step": 5772
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770505855091036e-05,
+      "loss": 1.4023,
+      "step": 5773
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877000791443592e-05,
+      "loss": 1.5967,
+      "step": 5774
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769509879576544e-05,
+      "loss": 0.9717,
+      "step": 5775
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769011750518252e-05,
+      "loss": 0.9443,
+      "step": 5776
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768513527266402e-05,
+      "loss": 0.814,
+      "step": 5777
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768015209826346e-05,
+      "loss": 1.5234,
+      "step": 5778
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767516798203432e-05,
+      "loss": 1.002,
+      "step": 5779
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767018292403015e-05,
+      "loss": 1.3799,
+      "step": 5780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766519692430453e-05,
+      "loss": 1.0117,
+      "step": 5781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87660209982911e-05,
+      "loss": 1.2598,
+      "step": 5782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876552220999031e-05,
+      "loss": 1.3623,
+      "step": 5783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8765023327533445e-05,
+      "loss": 1.0381,
+      "step": 5784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876452435092586e-05,
+      "loss": 1.0684,
+      "step": 5785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876402528017292e-05,
+      "loss": 0.8135,
+      "step": 5786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876352611527998e-05,
+      "loss": 1.4199,
+      "step": 5787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763026856252404e-05,
+      "loss": 1.3975,
+      "step": 5788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762527503095557e-05,
+      "loss": 0.7393,
+      "step": 5789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87620280558148e-05,
+      "loss": 1.1719,
+      "step": 5790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761528514415496e-05,
+      "loss": 1.0605,
+      "step": 5791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761028878903018e-05,
+      "loss": 0.9038,
+      "step": 5792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876052914928273e-05,
+      "loss": 0.8945,
+      "step": 5793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760029325559992e-05,
+      "loss": 1.4268,
+      "step": 5794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875952940774019e-05,
+      "loss": 0.8369,
+      "step": 5795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759029395828675e-05,
+      "loss": 1.2588,
+      "step": 5796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758529289830828e-05,
+      "loss": 1.1787,
+      "step": 5797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758029089752023e-05,
+      "loss": 1.0771,
+      "step": 5798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757528795597625e-05,
+      "loss": 1.6855,
+      "step": 5799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875702840737302e-05,
+      "loss": 1.0889,
+      "step": 5800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756527925083567e-05,
+      "loss": 1.0029,
+      "step": 5801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756027348734654e-05,
+      "loss": 0.8472,
+      "step": 5802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755526678331653e-05,
+      "loss": 1.1807,
+      "step": 5803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755025913879945e-05,
+      "loss": 1.3115,
+      "step": 5804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875452505538491e-05,
+      "loss": 1.269,
+      "step": 5805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875402410285192e-05,
+      "loss": 1.0791,
+      "step": 5806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753523056286365e-05,
+      "loss": 0.9512,
+      "step": 5807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753021915693623e-05,
+      "loss": 1.3223,
+      "step": 5808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752520681079076e-05,
+      "loss": 1.0566,
+      "step": 5809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752019352448114e-05,
+      "loss": 0.9229,
+      "step": 5810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751517929806113e-05,
+      "loss": 1.1455,
+      "step": 5811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751016413158468e-05,
+      "loss": 1.1504,
+      "step": 5812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875051480251056e-05,
+      "loss": 1.249,
+      "step": 5813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750013097867783e-05,
+      "loss": 0.9961,
+      "step": 5814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749511299235518e-05,
+      "loss": 1.5156,
+      "step": 5815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874900940661916e-05,
+      "loss": 1.2461,
+      "step": 5816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748507420024104e-05,
+      "loss": 1.1943,
+      "step": 5817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748005339455732e-05,
+      "loss": 1.2363,
+      "step": 5818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747503164919444e-05,
+      "loss": 1.4355,
+      "step": 5819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747000896420635e-05,
+      "loss": 1.248,
+      "step": 5820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87464985339647e-05,
+      "loss": 1.3516,
+      "step": 5821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745996077557032e-05,
+      "loss": 0.9648,
+      "step": 5822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745493527203026e-05,
+      "loss": 0.915,
+      "step": 5823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744990882908086e-05,
+      "loss": 1.3574,
+      "step": 5824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744488144677612e-05,
+      "loss": 1.2344,
+      "step": 5825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743985312517e-05,
+      "loss": 1.627,
+      "step": 5826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743482386431652e-05,
+      "loss": 1.5137,
+      "step": 5827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874297936642697e-05,
+      "loss": 1.1055,
+      "step": 5828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874247625250836e-05,
+      "loss": 1.1494,
+      "step": 5829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741973044681226e-05,
+      "loss": 0.6592,
+      "step": 5830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874146974295097e-05,
+      "loss": 1.1631,
+      "step": 5831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740966347323e-05,
+      "loss": 0.9238,
+      "step": 5832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740462857802724e-05,
+      "loss": 1.1719,
+      "step": 5833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873995927439555e-05,
+      "loss": 1.165,
+      "step": 5834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739455597106883e-05,
+      "loss": 1.5781,
+      "step": 5835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738951825942143e-05,
+      "loss": 1.1719,
+      "step": 5836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873844796090673e-05,
+      "loss": 1.5215,
+      "step": 5837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737944002006065e-05,
+      "loss": 1.2559,
+      "step": 5838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737439949245556e-05,
+      "loss": 1.3018,
+      "step": 5839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736935802630622e-05,
+      "loss": 1.3633,
+      "step": 5840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736431562166677e-05,
+      "loss": 1.3691,
+      "step": 5841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735927227859136e-05,
+      "loss": 1.0889,
+      "step": 5842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735422799713415e-05,
+      "loss": 1.1553,
+      "step": 5843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734918277734932e-05,
+      "loss": 0.8428,
+      "step": 5844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734413661929113e-05,
+      "loss": 1.4336,
+      "step": 5845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873390895230137e-05,
+      "loss": 0.9736,
+      "step": 5846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873340414885713e-05,
+      "loss": 0.9351,
+      "step": 5847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873289925160181e-05,
+      "loss": 1.4238,
+      "step": 5848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732394260540844e-05,
+      "loss": 0.9834,
+      "step": 5849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731889175679647e-05,
+      "loss": 1.2129,
+      "step": 5850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731383997023646e-05,
+      "loss": 1.6533,
+      "step": 5851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873087872457827e-05,
+      "loss": 0.9531,
+      "step": 5852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730373358348942e-05,
+      "loss": 1.1816,
+      "step": 5853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87298678983411e-05,
+      "loss": 1.3906,
+      "step": 5854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872936234456016e-05,
+      "loss": 0.9424,
+      "step": 5855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728856697011563e-05,
+      "loss": 1.2256,
+      "step": 5856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728350955700734e-05,
+      "loss": 0.9365,
+      "step": 5857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727845120633108e-05,
+      "loss": 1.123,
+      "step": 5858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727339191814124e-05,
+      "loss": 0.9922,
+      "step": 5859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726833169249206e-05,
+      "loss": 1.085,
+      "step": 5860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726327052943792e-05,
+      "loss": 1.6504,
+      "step": 5861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725820842903327e-05,
+      "loss": 1.1768,
+      "step": 5862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725314539133242e-05,
+      "loss": 1.168,
+      "step": 5863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872480814163898e-05,
+      "loss": 1.3174,
+      "step": 5864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872430165042597e-05,
+      "loss": 1.082,
+      "step": 5865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723795065499663e-05,
+      "loss": 0.9893,
+      "step": 5866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723288386865498e-05,
+      "loss": 1.4297,
+      "step": 5867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722781614528913e-05,
+      "loss": 1.2578,
+      "step": 5868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872227474849536e-05,
+      "loss": 1.1465,
+      "step": 5869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721767788770274e-05,
+      "loss": 1.5049,
+      "step": 5870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721260735359107e-05,
+      "loss": 1.1763,
+      "step": 5871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720753588267303e-05,
+      "loss": 1.5703,
+      "step": 5872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872024634750031e-05,
+      "loss": 1.1787,
+      "step": 5873
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871973901306358e-05,
+      "loss": 1.4629,
+      "step": 5874
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871923158496256e-05,
+      "loss": 1.0947,
+      "step": 5875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718724063202698e-05,
+      "loss": 1.0684,
+      "step": 5876
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718216447789446e-05,
+      "loss": 1.126,
+      "step": 5877
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717708738728263e-05,
+      "loss": 1.3965,
+      "step": 5878
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717200936024598e-05,
+      "loss": 1.2646,
+      "step": 5879
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716693039683904e-05,
+      "loss": 1.5215,
+      "step": 5880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716185049711636e-05,
+      "loss": 0.9795,
+      "step": 5881
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715676966113256e-05,
+      "loss": 1.1709,
+      "step": 5882
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871516878889422e-05,
+      "loss": 1.0107,
+      "step": 5883
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871466051805998e-05,
+      "loss": 1.3164,
+      "step": 5884
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714152153616008e-05,
+      "loss": 1.2861,
+      "step": 5885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713643695567753e-05,
+      "loss": 1.249,
+      "step": 5886
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713135143920684e-05,
+      "loss": 0.8286,
+      "step": 5887
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871262649868026e-05,
+      "loss": 1.1182,
+      "step": 5888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8712117759851945e-05,
+      "loss": 1.665,
+      "step": 5889
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711608927441203e-05,
+      "loss": 1.126,
+      "step": 5890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711100001453503e-05,
+      "loss": 0.7822,
+      "step": 5891
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710590981894314e-05,
+      "loss": 1.3867,
+      "step": 5892
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710081868769094e-05,
+      "loss": 1.5078,
+      "step": 5893
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709572662083318e-05,
+      "loss": 1.5596,
+      "step": 5894
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709063361842455e-05,
+      "loss": 1.0186,
+      "step": 5895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708553968051974e-05,
+      "loss": 1.2002,
+      "step": 5896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708044480717352e-05,
+      "loss": 1.2197,
+      "step": 5897
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870753489984406e-05,
+      "loss": 1.5215,
+      "step": 5898
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8707025225437565e-05,
+      "loss": 1.2695,
+      "step": 5899
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706515457503348e-05,
+      "loss": 1.0322,
+      "step": 5900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706005596046884e-05,
+      "loss": 0.9971,
+      "step": 5901
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870549564107365e-05,
+      "loss": 1.4014,
+      "step": 5902
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870498559258912e-05,
+      "loss": 1.2207,
+      "step": 5903
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704475450598775e-05,
+      "loss": 0.9839,
+      "step": 5904
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703965215108096e-05,
+      "loss": 1.1914,
+      "step": 5905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703454886122568e-05,
+      "loss": 1.1992,
+      "step": 5906
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702944463647662e-05,
+      "loss": 1.748,
+      "step": 5907
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870243394768887e-05,
+      "loss": 1.416,
+      "step": 5908
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870192333825167e-05,
+      "loss": 1.082,
+      "step": 5909
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8701412635341555e-05,
+      "loss": 1.4863,
+      "step": 5910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700901838963998e-05,
+      "loss": 1.082,
+      "step": 5911
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700390949124498e-05,
+      "loss": 1.5781,
+      "step": 5912
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699879965828535e-05,
+      "loss": 1.5332,
+      "step": 5913
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86993688890816e-05,
+      "loss": 1.1211,
+      "step": 5914
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8698857718889182e-05,
+      "loss": 1.1084,
+      "step": 5915
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8698346455256776e-05,
+      "loss": 1.2539,
+      "step": 5916
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869783509818987e-05,
+      "loss": 1.3574,
+      "step": 5917
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697323647693956e-05,
+      "loss": 1.1992,
+      "step": 5918
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869681210377453e-05,
+      "loss": 1.2324,
+      "step": 5919
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869630046643709e-05,
+      "loss": 0.9668,
+      "step": 5920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695788735687124e-05,
+      "loss": 1.0781,
+      "step": 5921
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695276911530134e-05,
+      "loss": 1.1011,
+      "step": 5922
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8694764993971616e-05,
+      "loss": 1.0698,
+      "step": 5923
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869425298301707e-05,
+      "loss": 1.4541,
+      "step": 5924
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693740878671996e-05,
+      "loss": 1.228,
+      "step": 5925
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693228680941895e-05,
+      "loss": 1.0801,
+      "step": 5926
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692716389832266e-05,
+      "loss": 1.3027,
+      "step": 5927
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692204005348615e-05,
+      "loss": 1.5293,
+      "step": 5928
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8691691527496444e-05,
+      "loss": 1.2734,
+      "step": 5929
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869117895628126e-05,
+      "loss": 1.5469,
+      "step": 5930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8690666291708566e-05,
+      "loss": 0.9028,
+      "step": 5931
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8690153533783874e-05,
+      "loss": 1.1577,
+      "step": 5932
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689640682512685e-05,
+      "loss": 1.208,
+      "step": 5933
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868912773790051e-05,
+      "loss": 1.1426,
+      "step": 5934
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8688614699952863e-05,
+      "loss": 1.5146,
+      "step": 5935
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868810156867525e-05,
+      "loss": 0.5283,
+      "step": 5936
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8687588344073183e-05,
+      "loss": 1.4082,
+      "step": 5937
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868707502615218e-05,
+      "loss": 1.332,
+      "step": 5938
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868656161491775e-05,
+      "loss": 0.7686,
+      "step": 5939
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868604811037541e-05,
+      "loss": 1.3779,
+      "step": 5940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685534512530674e-05,
+      "loss": 0.6912,
+      "step": 5941
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685020821389058e-05,
+      "loss": 1.502,
+      "step": 5942
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8684507036956088e-05,
+      "loss": 1.2246,
+      "step": 5943
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868399315923727e-05,
+      "loss": 1.5625,
+      "step": 5944
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683479188238135e-05,
+      "loss": 1.3994,
+      "step": 5945
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86829651239642e-05,
+      "loss": 1.4102,
+      "step": 5946
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8682450966420987e-05,
+      "loss": 0.9697,
+      "step": 5947
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681936715614017e-05,
+      "loss": 1.2285,
+      "step": 5948
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681422371548814e-05,
+      "loss": 1.3662,
+      "step": 5949
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680907934230905e-05,
+      "loss": 1.3555,
+      "step": 5950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680393403665815e-05,
+      "loss": 0.9277,
+      "step": 5951
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867987877985907e-05,
+      "loss": 1.1792,
+      "step": 5952
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86793640628162e-05,
+      "loss": 0.9058,
+      "step": 5953
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678849252542733e-05,
+      "loss": 0.8018,
+      "step": 5954
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678334349044202e-05,
+      "loss": 1.1543,
+      "step": 5955
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867781935232613e-05,
+      "loss": 1.2305,
+      "step": 5956
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677304262394055e-05,
+      "loss": 1.2822,
+      "step": 5957
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867678907925351e-05,
+      "loss": 1.1748,
+      "step": 5958
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676273802910024e-05,
+      "loss": 1.1982,
+      "step": 5959
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675758433369137e-05,
+      "loss": 1.2246,
+      "step": 5960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675242970636388e-05,
+      "loss": 0.9805,
+      "step": 5961
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8674727414717302e-05,
+      "loss": 1.7246,
+      "step": 5962
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8674211765617426e-05,
+      "loss": 1.1416,
+      "step": 5963
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86736960233423e-05,
+      "loss": 0.8057,
+      "step": 5964
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673180187897456e-05,
+      "loss": 1.0068,
+      "step": 5965
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672664259288444e-05,
+      "loss": 1.1514,
+      "step": 5966
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672148237520796e-05,
+      "loss": 1.4199,
+      "step": 5967
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671632122600067e-05,
+      "loss": 1.2832,
+      "step": 5968
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671115914531792e-05,
+      "loss": 1.5605,
+      "step": 5969
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8670599613321518e-05,
+      "loss": 0.6147,
+      "step": 5970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867008321897479e-05,
+      "loss": 0.9688,
+      "step": 5971
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866956673149716e-05,
+      "loss": 1.2305,
+      "step": 5972
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866905015089417e-05,
+      "loss": 1.3018,
+      "step": 5973
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668533477171372e-05,
+      "loss": 1.0498,
+      "step": 5974
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668016710334313e-05,
+      "loss": 1.252,
+      "step": 5975
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667499850388547e-05,
+      "loss": 1.1123,
+      "step": 5976
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666982897339627e-05,
+      "loss": 1.6045,
+      "step": 5977
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86664658511931e-05,
+      "loss": 1.4629,
+      "step": 5978
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665948711954528e-05,
+      "loss": 0.7705,
+      "step": 5979
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665431479629463e-05,
+      "loss": 1.0679,
+      "step": 5980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664914154223456e-05,
+      "loss": 1.1064,
+      "step": 5981
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664396735742064e-05,
+      "loss": 0.9751,
+      "step": 5982
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8663879224190853e-05,
+      "loss": 0.9868,
+      "step": 5983
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8663361619575376e-05,
+      "loss": 1.1797,
+      "step": 5984
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662843921901198e-05,
+      "loss": 1.2227,
+      "step": 5985
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866232613117387e-05,
+      "loss": 1.0811,
+      "step": 5986
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8661808247398964e-05,
+      "loss": 0.9531,
+      "step": 5987
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866129027058204e-05,
+      "loss": 1.1924,
+      "step": 5988
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8660772200728658e-05,
+      "loss": 1.4375,
+      "step": 5989
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866025403784439e-05,
+      "loss": 0.9404,
+      "step": 5990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659735781934792e-05,
+      "loss": 1.1953,
+      "step": 5991
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865921743300544e-05,
+      "loss": 1.252,
+      "step": 5992
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86586989910619e-05,
+      "loss": 0.96,
+      "step": 5993
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658180456109738e-05,
+      "loss": 0.7749,
+      "step": 5994
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8657661828154527e-05,
+      "loss": 1.0972,
+      "step": 5995
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865714310720183e-05,
+      "loss": 0.8848,
+      "step": 5996
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8656624293257232e-05,
+      "loss": 1.6016,
+      "step": 5997
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86561053863263e-05,
+      "loss": 1.0293,
+      "step": 5998
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655586386414604e-05,
+      "loss": 1.3184,
+      "step": 5999
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865506729352772e-05,
+      "loss": 1.3438,
+      "step": 6000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865454810767123e-05,
+      "loss": 1.2451,
+      "step": 6001
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654028828850705e-05,
+      "loss": 1.2817,
+      "step": 6002
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653509457071726e-05,
+      "loss": 0.9932,
+      "step": 6003
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865298999233987e-05,
+      "loss": 1.4668,
+      "step": 6004
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865247043466072e-05,
+      "loss": 1.2764,
+      "step": 6005
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865195078403985e-05,
+      "loss": 1.584,
+      "step": 6006
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865143104048285e-05,
+      "loss": 1.2139,
+      "step": 6007
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650911203995296e-05,
+      "loss": 1.2363,
+      "step": 6008
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650391274582778e-05,
+      "loss": 1.3965,
+      "step": 6009
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649871252250874e-05,
+      "loss": 0.9824,
+      "step": 6010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649351137005178e-05,
+      "loss": 1.3223,
+      "step": 6011
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864883092885127e-05,
+      "loss": 0.999,
+      "step": 6012
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648310627794745e-05,
+      "loss": 1.0547,
+      "step": 6013
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647790233841183e-05,
+      "loss": 1.2749,
+      "step": 6014
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647269746996183e-05,
+      "loss": 1.3301,
+      "step": 6015
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8646749167265327e-05,
+      "loss": 1.2129,
+      "step": 6016
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8646228494654215e-05,
+      "loss": 1.4736,
+      "step": 6017
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864570772916843e-05,
+      "loss": 1.4414,
+      "step": 6018
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8645186870813578e-05,
+      "loss": 1.1484,
+      "step": 6019
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864466591959525e-05,
+      "loss": 1.6162,
+      "step": 6020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644144875519035e-05,
+      "loss": 1.0283,
+      "step": 6021
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8643623738590536e-05,
+      "loss": 1.0879,
+      "step": 6022
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8643102508815348e-05,
+      "loss": 0.8721,
+      "step": 6023
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864258118619907e-05,
+      "loss": 1.1807,
+      "step": 6024
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642059770747307e-05,
+      "loss": 1.3096,
+      "step": 6025
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641538262465656e-05,
+      "loss": 1.0078,
+      "step": 6026
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641016661359717e-05,
+      "loss": 0.9111,
+      "step": 6027
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640494967435095e-05,
+      "loss": 1.2188,
+      "step": 6028
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8639973180697394e-05,
+      "loss": 1.2559,
+      "step": 6029
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863945130115222e-05,
+      "loss": 1.0127,
+      "step": 6030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638929328805174e-05,
+      "loss": 1.3691,
+      "step": 6031
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638407263661865e-05,
+      "loss": 1.0645,
+      "step": 6032
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863788510572791e-05,
+      "loss": 1.2656,
+      "step": 6033
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637362855008902e-05,
+      "loss": 1.1123,
+      "step": 6034
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636840511510465e-05,
+      "loss": 0.9985,
+      "step": 6035
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86363180752382e-05,
+      "loss": 1.3008,
+      "step": 6036
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863579554619772e-05,
+      "loss": 1.1963,
+      "step": 6037
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635272924394643e-05,
+      "loss": 1.166,
+      "step": 6038
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634750209834577e-05,
+      "loss": 1.2061,
+      "step": 6039
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863422740252314e-05,
+      "loss": 1.3926,
+      "step": 6040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863370450246595e-05,
+      "loss": 0.9238,
+      "step": 6041
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633181509668623e-05,
+      "loss": 1.0674,
+      "step": 6042
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632658424136773e-05,
+      "loss": 1.1104,
+      "step": 6043
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863213524587602e-05,
+      "loss": 1.4131,
+      "step": 6044
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8631611974891984e-05,
+      "loss": 1.1191,
+      "step": 6045
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863108861119029e-05,
+      "loss": 1.3467,
+      "step": 6046
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630565154776557e-05,
+      "loss": 1.0625,
+      "step": 6047
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630041605656403e-05,
+      "loss": 1.3867,
+      "step": 6048
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862951796383546e-05,
+      "loss": 1.5527,
+      "step": 6049
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862899422931934e-05,
+      "loss": 1.6289,
+      "step": 6050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628470402113687e-05,
+      "loss": 1.2598,
+      "step": 6051
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8627946482224115e-05,
+      "loss": 1.0752,
+      "step": 6052
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8627422469656253e-05,
+      "loss": 1.1797,
+      "step": 6053
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626898364415733e-05,
+      "loss": 1.2236,
+      "step": 6054
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626374166508182e-05,
+      "loss": 1.3555,
+      "step": 6055
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862584987593924e-05,
+      "loss": 1.0088,
+      "step": 6056
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862532549271452e-05,
+      "loss": 1.1543,
+      "step": 6057
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862480101683967e-05,
+      "loss": 1.0586,
+      "step": 6058
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862427644832032e-05,
+      "loss": 1.0205,
+      "step": 6059
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623751787162106e-05,
+      "loss": 0.8643,
+      "step": 6060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623227033370657e-05,
+      "loss": 1.0146,
+      "step": 6061
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862270218695162e-05,
+      "loss": 0.7646,
+      "step": 6062
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622177247910623e-05,
+      "loss": 1.1875,
+      "step": 6063
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862165221625331e-05,
+      "loss": 1.0645,
+      "step": 6064
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862112709198532e-05,
+      "loss": 0.877,
+      "step": 6065
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620601875112295e-05,
+      "loss": 1.0225,
+      "step": 6066
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620076565639874e-05,
+      "loss": 1.0469,
+      "step": 6067
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8619551163573697e-05,
+      "loss": 1.0762,
+      "step": 6068
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8619025668919416e-05,
+      "loss": 1.2695,
+      "step": 6069
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861850008168267e-05,
+      "loss": 1.0605,
+      "step": 6070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8617974401869106e-05,
+      "loss": 1.0664,
+      "step": 6071
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861744862948437e-05,
+      "loss": 1.0459,
+      "step": 6072
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616922764534108e-05,
+      "loss": 0.8789,
+      "step": 6073
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616396807023975e-05,
+      "loss": 1.123,
+      "step": 6074
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8615870756959613e-05,
+      "loss": 1.2637,
+      "step": 6075
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861534461434668e-05,
+      "loss": 1.2666,
+      "step": 6076
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861481837919082e-05,
+      "loss": 1.1943,
+      "step": 6077
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614292051497686e-05,
+      "loss": 1.1152,
+      "step": 6078
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613765631272943e-05,
+      "loss": 1.207,
+      "step": 6079
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861323911852223e-05,
+      "loss": 1.2275,
+      "step": 6080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612712513251214e-05,
+      "loss": 1.3115,
+      "step": 6081
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612185815465547e-05,
+      "loss": 1.207,
+      "step": 6082
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861165902517089e-05,
+      "loss": 1.2832,
+      "step": 6083
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8611132142372896e-05,
+      "loss": 1.1279,
+      "step": 6084
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610605167077226e-05,
+      "loss": 0.9287,
+      "step": 6085
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610078099289544e-05,
+      "loss": 1.4941,
+      "step": 6086
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860955093901551e-05,
+      "loss": 1.0537,
+      "step": 6087
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8609023686260785e-05,
+      "loss": 1.0791,
+      "step": 6088
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8608496341031035e-05,
+      "loss": 1.1279,
+      "step": 6089
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860796890333192e-05,
+      "loss": 1.125,
+      "step": 6090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8607441373169116e-05,
+      "loss": 1.4287,
+      "step": 6091
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860691375054828e-05,
+      "loss": 0.9492,
+      "step": 6092
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860638603547508e-05,
+      "loss": 1.4238,
+      "step": 6093
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605858227955183e-05,
+      "loss": 0.9111,
+      "step": 6094
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605330327994266e-05,
+      "loss": 1.1729,
+      "step": 6095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604802335597995e-05,
+      "loss": 1.8242,
+      "step": 6096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604274250772042e-05,
+      "loss": 1.2285,
+      "step": 6097
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860374607352208e-05,
+      "loss": 1.3662,
+      "step": 6098
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860321780385378e-05,
+      "loss": 0.9092,
+      "step": 6099
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602689441772823e-05,
+      "loss": 1.1689,
+      "step": 6100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602160987284876e-05,
+      "loss": 0.8413,
+      "step": 6101
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860163244039562e-05,
+      "loss": 1.4707,
+      "step": 6102
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8601103801110733e-05,
+      "loss": 0.8623,
+      "step": 6103
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860057506943589e-05,
+      "loss": 1.1875,
+      "step": 6104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860004624537678e-05,
+      "loss": 1.5801,
+      "step": 6105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859951732893907e-05,
+      "loss": 0.7124,
+      "step": 6106
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598988320128445e-05,
+      "loss": 1.5117,
+      "step": 6107
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859845921895059e-05,
+      "loss": 1.0518,
+      "step": 6108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8597930025411194e-05,
+      "loss": 1.8906,
+      "step": 6109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8597400739515934e-05,
+      "loss": 1.4199,
+      "step": 6110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596871361270494e-05,
+      "loss": 1.4023,
+      "step": 6111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596341890680568e-05,
+      "loss": 0.8667,
+      "step": 6112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595812327751835e-05,
+      "loss": 1.082,
+      "step": 6113
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595282672489993e-05,
+      "loss": 1.3438,
+      "step": 6114
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594752924900722e-05,
+      "loss": 1.3164,
+      "step": 6115
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859422308498972e-05,
+      "loss": 1.2197,
+      "step": 6116
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859369315276267e-05,
+      "loss": 1.5312,
+      "step": 6117
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859316312822527e-05,
+      "loss": 1.1084,
+      "step": 6118
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859263301138321e-05,
+      "loss": 0.8726,
+      "step": 6119
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859210280224219e-05,
+      "loss": 0.8535,
+      "step": 6120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85915725008079e-05,
+      "loss": 0.835,
+      "step": 6121
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859104210708604e-05,
+      "loss": 1.0195,
+      "step": 6122
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590511621082302e-05,
+      "loss": 0.8477,
+      "step": 6123
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858998104280239e-05,
+      "loss": 1.1597,
+      "step": 6124
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589450372252002e-05,
+      "loss": 1.0615,
+      "step": 6125
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8588919609436837e-05,
+      "loss": 0.9404,
+      "step": 6126
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8588388754362596e-05,
+      "loss": 1.1338,
+      "step": 6127
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587857807034982e-05,
+      "loss": 1.7168,
+      "step": 6128
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587326767459696e-05,
+      "loss": 1.3027,
+      "step": 6129
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8586795635642444e-05,
+      "loss": 1.2363,
+      "step": 6130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8586264411588932e-05,
+      "loss": 1.3984,
+      "step": 6131
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858573309530487e-05,
+      "loss": 1.1797,
+      "step": 6132
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585201686795956e-05,
+      "loss": 1.5547,
+      "step": 6133
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584670186067907e-05,
+      "loss": 0.7554,
+      "step": 6134
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584138593126423e-05,
+      "loss": 1.1953,
+      "step": 6135
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8583606907977224e-05,
+      "loss": 1.0469,
+      "step": 6136
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8583075130626013e-05,
+      "loss": 1.0186,
+      "step": 6137
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858254326107851e-05,
+      "loss": 1.1802,
+      "step": 6138
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858201129934042e-05,
+      "loss": 1.1865,
+      "step": 6139
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8581479245417463e-05,
+      "loss": 1.1123,
+      "step": 6140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580947099315354e-05,
+      "loss": 1.0264,
+      "step": 6141
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580414861039803e-05,
+      "loss": 1.4043,
+      "step": 6142
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579882530596535e-05,
+      "loss": 0.7949,
+      "step": 6143
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579350107991263e-05,
+      "loss": 1.4792,
+      "step": 6144
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578817593229708e-05,
+      "loss": 0.9697,
+      "step": 6145
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578284986317592e-05,
+      "loss": 1.4463,
+      "step": 6146
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8577752287260628e-05,
+      "loss": 1.3193,
+      "step": 6147
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857721949606455e-05,
+      "loss": 1.1523,
+      "step": 6148
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576686612735072e-05,
+      "loss": 1.2559,
+      "step": 6149
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857615363727792e-05,
+      "loss": 1.1836,
+      "step": 6150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575620569698823e-05,
+      "loss": 1.2051,
+      "step": 6151
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575087410003502e-05,
+      "loss": 0.8711,
+      "step": 6152
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574554158197685e-05,
+      "loss": 1.0732,
+      "step": 6153
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574020814287104e-05,
+      "loss": 1.0127,
+      "step": 6154
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8573487378277483e-05,
+      "loss": 1.0146,
+      "step": 6155
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572953850174557e-05,
+      "loss": 1.2686,
+      "step": 6156
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572420229984048e-05,
+      "loss": 0.9731,
+      "step": 6157
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85718865177117e-05,
+      "loss": 1.3779,
+      "step": 6158
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571352713363232e-05,
+      "loss": 0.9976,
+      "step": 6159
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8570818816944393e-05,
+      "loss": 1.0947,
+      "step": 6160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857028482846091e-05,
+      "loss": 1.0967,
+      "step": 6161
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569750747918518e-05,
+      "loss": 1.5332,
+      "step": 6162
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569216575322954e-05,
+      "loss": 1.0127,
+      "step": 6163
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856868231067996e-05,
+      "loss": 0.9951,
+      "step": 6164
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856814795399527e-05,
+      "loss": 1.3145,
+      "step": 6165
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567613505274625e-05,
+      "loss": 1.2812,
+      "step": 6166
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856707896452377e-05,
+      "loss": 1.0557,
+      "step": 6167
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566544331748443e-05,
+      "loss": 1.4414,
+      "step": 6168
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856600960695439e-05,
+      "loss": 1.0518,
+      "step": 6169
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565474790147352e-05,
+      "loss": 0.8096,
+      "step": 6170
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856493988133307e-05,
+      "loss": 1.1104,
+      "step": 6171
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564404880517297e-05,
+      "loss": 1.4805,
+      "step": 6172
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563869787705776e-05,
+      "loss": 1.1143,
+      "step": 6173
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563334602904257e-05,
+      "loss": 1.1924,
+      "step": 6174
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562799326118485e-05,
+      "loss": 1.1104,
+      "step": 6175
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562263957354214e-05,
+      "loss": 1.1533,
+      "step": 6176
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8561728496617194e-05,
+      "loss": 1.0918,
+      "step": 6177
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8561192943913173e-05,
+      "loss": 0.8564,
+      "step": 6178
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856065729924791e-05,
+      "loss": 1.0117,
+      "step": 6179
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856012156262715e-05,
+      "loss": 1.2852,
+      "step": 6180
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8559585734056655e-05,
+      "loss": 0.7334,
+      "step": 6181
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855904981354218e-05,
+      "loss": 1.0078,
+      "step": 6182
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558513801089478e-05,
+      "loss": 1.0186,
+      "step": 6183
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557977696704313e-05,
+      "loss": 1.4551,
+      "step": 6184
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557441500392434e-05,
+      "loss": 1.2109,
+      "step": 6185
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556905212159607e-05,
+      "loss": 1.1367,
+      "step": 6186
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556368832011594e-05,
+      "loss": 1.1592,
+      "step": 6187
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8555832359954153e-05,
+      "loss": 1.3809,
+      "step": 6188
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855529579599305e-05,
+      "loss": 0.7783,
+      "step": 6189
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855475914013404e-05,
+      "loss": 1.2393,
+      "step": 6190
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85542223923829e-05,
+      "loss": 1.0586,
+      "step": 6191
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8553685552745387e-05,
+      "loss": 1.1582,
+      "step": 6192
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855314862122727e-05,
+      "loss": 1.5938,
+      "step": 6193
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855261159783432e-05,
+      "loss": 1.2715,
+      "step": 6194
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85520744825723e-05,
+      "loss": 1.1855,
+      "step": 6195
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551537275446978e-05,
+      "loss": 1.3398,
+      "step": 6196
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855099997646413e-05,
+      "loss": 1.2617,
+      "step": 6197
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550462585629527e-05,
+      "loss": 1.2588,
+      "step": 6198
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854992510294894e-05,
+      "loss": 1.1079,
+      "step": 6199
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854938752842814e-05,
+      "loss": 1.6709,
+      "step": 6200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8548849862072907e-05,
+      "loss": 1.4512,
+      "step": 6201
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8548312103889014e-05,
+      "loss": 1.2598,
+      "step": 6202
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547774253882233e-05,
+      "loss": 1.0732,
+      "step": 6203
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547236312058345e-05,
+      "loss": 0.8779,
+      "step": 6204
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854669827842313e-05,
+      "loss": 1.0605,
+      "step": 6205
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546160152982366e-05,
+      "loss": 1.3115,
+      "step": 6206
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8545621935741835e-05,
+      "loss": 1.0029,
+      "step": 6207
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8545083626707314e-05,
+      "loss": 1.0508,
+      "step": 6208
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8544545225884586e-05,
+      "loss": 1.0557,
+      "step": 6209
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854400673327944e-05,
+      "loss": 1.3574,
+      "step": 6210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543468148897657e-05,
+      "loss": 1.2241,
+      "step": 6211
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542929472745016e-05,
+      "loss": 1.1851,
+      "step": 6212
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542390704827316e-05,
+      "loss": 1.3145,
+      "step": 6213
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541851845150334e-05,
+      "loss": 1.6875,
+      "step": 6214
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854131289371986e-05,
+      "loss": 1.0869,
+      "step": 6215
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540773850541683e-05,
+      "loss": 1.1611,
+      "step": 6216
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540234715621598e-05,
+      "loss": 1.0469,
+      "step": 6217
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539695488965396e-05,
+      "loss": 1.4453,
+      "step": 6218
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539156170578863e-05,
+      "loss": 1.4863,
+      "step": 6219
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8538616760467792e-05,
+      "loss": 1.5186,
+      "step": 6220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853807725863799e-05,
+      "loss": 1.0283,
+      "step": 6221
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853753766509523e-05,
+      "loss": 0.9961,
+      "step": 6222
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853699797984533e-05,
+      "loss": 1.334,
+      "step": 6223
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8536458202894077e-05,
+      "loss": 1.1172,
+      "step": 6224
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853591833424727e-05,
+      "loss": 1.2783,
+      "step": 6225
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8535378373910706e-05,
+      "loss": 1.1416,
+      "step": 6226
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8534838321890187e-05,
+      "loss": 0.8926,
+      "step": 6227
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8534298178191515e-05,
+      "loss": 1.0186,
+      "step": 6228
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533757942820492e-05,
+      "loss": 1.0928,
+      "step": 6229
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533217615782918e-05,
+      "loss": 1.3271,
+      "step": 6230
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85326771970846e-05,
+      "loss": 1.3809,
+      "step": 6231
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8532136686731342e-05,
+      "loss": 0.8066,
+      "step": 6232
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531596084728953e-05,
+      "loss": 1.3994,
+      "step": 6233
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531055391083236e-05,
+      "loss": 1.1953,
+      "step": 6234
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85305146058e-05,
+      "loss": 1.4502,
+      "step": 6235
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852997372888505e-05,
+      "loss": 1.1914,
+      "step": 6236
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529432760344207e-05,
+      "loss": 1.2002,
+      "step": 6237
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852889170018327e-05,
+      "loss": 1.1357,
+      "step": 6238
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528350548408057e-05,
+      "loss": 1.3477,
+      "step": 6239
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8527809305024378e-05,
+      "loss": 1.4258,
+      "step": 6240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852726797003805e-05,
+      "loss": 1.1641,
+      "step": 6241
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8526726543454883e-05,
+      "loss": 1.1436,
+      "step": 6242
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85261850252807e-05,
+      "loss": 1.2651,
+      "step": 6243
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852564341552131e-05,
+      "loss": 1.04,
+      "step": 6244
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525101714182537e-05,
+      "loss": 1.3262,
+      "step": 6245
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85245599212702e-05,
+      "loss": 1.3623,
+      "step": 6246
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524018036790114e-05,
+      "loss": 1.0713,
+      "step": 6247
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85234760607481e-05,
+      "loss": 1.4668,
+      "step": 6248
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852293399314998e-05,
+      "loss": 1.2773,
+      "step": 6249
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8522391834001582e-05,
+      "loss": 1.1123,
+      "step": 6250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521849583308728e-05,
+      "loss": 1.0806,
+      "step": 6251
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521307241077236e-05,
+      "loss": 1.3301,
+      "step": 6252
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520764807312935e-05,
+      "loss": 0.6938,
+      "step": 6253
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520222282021656e-05,
+      "loss": 1.2275,
+      "step": 6254
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851967966520922e-05,
+      "loss": 1.1758,
+      "step": 6255
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519136956881464e-05,
+      "loss": 1.1865,
+      "step": 6256
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851859415704421e-05,
+      "loss": 1.4844,
+      "step": 6257
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518051265703293e-05,
+      "loss": 1.1094,
+      "step": 6258
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851750828286454e-05,
+      "loss": 1.1289,
+      "step": 6259
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851696520853379e-05,
+      "loss": 1.3984,
+      "step": 6260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851642204271687e-05,
+      "loss": 1.2666,
+      "step": 6261
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515878785419615e-05,
+      "loss": 1.1475,
+      "step": 6262
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515335436647864e-05,
+      "loss": 0.8896,
+      "step": 6263
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514791996407455e-05,
+      "loss": 1.2031,
+      "step": 6264
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514248464704217e-05,
+      "loss": 1.3477,
+      "step": 6265
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8513704841543997e-05,
+      "loss": 1.3691,
+      "step": 6266
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851316112693263e-05,
+      "loss": 1.2119,
+      "step": 6267
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851261732087596e-05,
+      "loss": 0.7612,
+      "step": 6268
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851207342337983e-05,
+      "loss": 1.3457,
+      "step": 6269
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8511529434450076e-05,
+      "loss": 0.9897,
+      "step": 6270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851098535409254e-05,
+      "loss": 1.1572,
+      "step": 6271
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851044118231307e-05,
+      "loss": 1.1719,
+      "step": 6272
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8509896919117517e-05,
+      "loss": 1.0234,
+      "step": 6273
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850935256451172e-05,
+      "loss": 1.2998,
+      "step": 6274
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508808118501525e-05,
+      "loss": 1.1338,
+      "step": 6275
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508263581092783e-05,
+      "loss": 1.1436,
+      "step": 6276
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507718952291345e-05,
+      "loss": 0.917,
+      "step": 6277
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850717423210306e-05,
+      "loss": 1.3486,
+      "step": 6278
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850662942053378e-05,
+      "loss": 1.248,
+      "step": 6279
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8506084517589354e-05,
+      "loss": 1.1523,
+      "step": 6280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505539523275638e-05,
+      "loss": 1.4863,
+      "step": 6281
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504994437598488e-05,
+      "loss": 1.0703,
+      "step": 6282
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504449260563753e-05,
+      "loss": 1.1348,
+      "step": 6283
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8503903992177293e-05,
+      "loss": 1.3232,
+      "step": 6284
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8503358632444965e-05,
+      "loss": 1.3232,
+      "step": 6285
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8502813181372628e-05,
+      "loss": 1.1836,
+      "step": 6286
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8502267638966135e-05,
+      "loss": 0.7451,
+      "step": 6287
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8501722005231356e-05,
+      "loss": 1.2383,
+      "step": 6288
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850117628017414e-05,
+      "loss": 1.0449,
+      "step": 6289
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500630463800364e-05,
+      "loss": 1.1982,
+      "step": 6290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850008455611588e-05,
+      "loss": 1.457,
+      "step": 6291
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849953855712655e-05,
+      "loss": 1.2842,
+      "step": 6292
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849899246683825e-05,
+      "loss": 1.5488,
+      "step": 6293
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8498446285256836e-05,
+      "loss": 1.0391,
+      "step": 6294
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849790001238818e-05,
+      "loss": 0.7529,
+      "step": 6295
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8497353648238148e-05,
+      "loss": 1.4766,
+      "step": 6296
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8496807192812608e-05,
+      "loss": 1.2451,
+      "step": 6297
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849626064611743e-05,
+      "loss": 1.1992,
+      "step": 6298
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8495714008158487e-05,
+      "loss": 0.9053,
+      "step": 6299
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849516727894165e-05,
+      "loss": 1.4678,
+      "step": 6300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849462045847279e-05,
+      "loss": 1.165,
+      "step": 6301
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849407354675778e-05,
+      "loss": 0.8701,
+      "step": 6302
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8493526543802497e-05,
+      "loss": 1.5391,
+      "step": 6303
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8492979449612823e-05,
+      "loss": 1.5557,
+      "step": 6304
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849243226419462e-05,
+      "loss": 0.9414,
+      "step": 6305
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849188498755378e-05,
+      "loss": 1.2402,
+      "step": 6306
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849133761969617e-05,
+      "loss": 1.0986,
+      "step": 6307
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8490790160627678e-05,
+      "loss": 1.1152,
+      "step": 6308
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8490242610354184e-05,
+      "loss": 1.0293,
+      "step": 6309
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848969496888156e-05,
+      "loss": 1.1768,
+      "step": 6310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84891472362157e-05,
+      "loss": 1.5859,
+      "step": 6311
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488599412362488e-05,
+      "loss": 1.3369,
+      "step": 6312
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488051497327798e-05,
+      "loss": 1.334,
+      "step": 6313
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8487503491117523e-05,
+      "loss": 1.0249,
+      "step": 6314
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8486955393737548e-05,
+      "loss": 0.9404,
+      "step": 6315
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848640720519376e-05,
+      "loss": 1.6797,
+      "step": 6316
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8485858925492048e-05,
+      "loss": 0.7803,
+      "step": 6317
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84853105546383e-05,
+      "loss": 1.1777,
+      "step": 6318
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848476209263841e-05,
+      "loss": 1.2061,
+      "step": 6319
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8484213539498266e-05,
+      "loss": 1.3892,
+      "step": 6320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848366489522376e-05,
+      "loss": 1.1611,
+      "step": 6321
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8483116159820785e-05,
+      "loss": 1.3105,
+      "step": 6322
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848256733329524e-05,
+      "loss": 1.0205,
+      "step": 6323
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848201841565302e-05,
+      "loss": 1.1094,
+      "step": 6324
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8481469406900016e-05,
+      "loss": 0.5781,
+      "step": 6325
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8480920307042124e-05,
+      "loss": 1.3457,
+      "step": 6326
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848037111608525e-05,
+      "loss": 1.248,
+      "step": 6327
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847982183403529e-05,
+      "loss": 1.5986,
+      "step": 6328
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847927246089814e-05,
+      "loss": 1.0508,
+      "step": 6329
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8478722996679708e-05,
+      "loss": 1.4277,
+      "step": 6330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8478173441385892e-05,
+      "loss": 1.251,
+      "step": 6331
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8477623795022594e-05,
+      "loss": 1.3828,
+      "step": 6332
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847707405759572e-05,
+      "loss": 0.7773,
+      "step": 6333
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8476524229111178e-05,
+      "loss": 0.9795,
+      "step": 6334
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847597430957487e-05,
+      "loss": 1.0293,
+      "step": 6335
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475424298992705e-05,
+      "loss": 1.0703,
+      "step": 6336
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8474874197370586e-05,
+      "loss": 1.0234,
+      "step": 6337
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847432400471443e-05,
+      "loss": 1.1025,
+      "step": 6338
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847377372103014e-05,
+      "loss": 1.2378,
+      "step": 6339
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473223346323635e-05,
+      "loss": 1.1172,
+      "step": 6340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847267288060082e-05,
+      "loss": 1.3896,
+      "step": 6341
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847212232386761e-05,
+      "loss": 1.2461,
+      "step": 6342
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8471571676129917e-05,
+      "loss": 0.9658,
+      "step": 6343
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8471020937393658e-05,
+      "loss": 1.335,
+      "step": 6344
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847047010766475e-05,
+      "loss": 0.8496,
+      "step": 6345
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469919186949105e-05,
+      "loss": 1.1787,
+      "step": 6346
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846936817525265e-05,
+      "loss": 1.0684,
+      "step": 6347
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8468817072581293e-05,
+      "loss": 1.208,
+      "step": 6348
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8468265878940958e-05,
+      "loss": 1.0811,
+      "step": 6349
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846771459433757e-05,
+      "loss": 1.2803,
+      "step": 6350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8467163218777044e-05,
+      "loss": 1.167,
+      "step": 6351
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846661175226531e-05,
+      "loss": 1.043,
+      "step": 6352
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8466060194808284e-05,
+      "loss": 1.3516,
+      "step": 6353
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8465508546411894e-05,
+      "loss": 1.0625,
+      "step": 6354
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464956807082066e-05,
+      "loss": 1.5166,
+      "step": 6355
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464404976824727e-05,
+      "loss": 1.2188,
+      "step": 6356
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463853055645805e-05,
+      "loss": 1.6152,
+      "step": 6357
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463301043551225e-05,
+      "loss": 1.3828,
+      "step": 6358
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846274894054692e-05,
+      "loss": 1.0186,
+      "step": 6359
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8462196746638818e-05,
+      "loss": 1.4512,
+      "step": 6360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846164446183285e-05,
+      "loss": 1.3457,
+      "step": 6361
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8461092086134953e-05,
+      "loss": 1.4102,
+      "step": 6362
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846053961955106e-05,
+      "loss": 1.1641,
+      "step": 6363
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84599870620871e-05,
+      "loss": 1.4219,
+      "step": 6364
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8459434413749013e-05,
+      "loss": 1.0898,
+      "step": 6365
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845888167454273e-05,
+      "loss": 0.9331,
+      "step": 6366
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8458328844474195e-05,
+      "loss": 1.416,
+      "step": 6367
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457775923549343e-05,
+      "loss": 1.1143,
+      "step": 6368
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457222911774112e-05,
+      "loss": 1.4844,
+      "step": 6369
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456669809154444e-05,
+      "loss": 1.3145,
+      "step": 6370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456116615696284e-05,
+      "loss": 1.0654,
+      "step": 6371
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845556333140557e-05,
+      "loss": 1.1572,
+      "step": 6372
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8455009956288244e-05,
+      "loss": 1.3008,
+      "step": 6373
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454456490350254e-05,
+      "loss": 1.6465,
+      "step": 6374
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453902933597538e-05,
+      "loss": 1.1382,
+      "step": 6375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453349286036046e-05,
+      "loss": 1.7012,
+      "step": 6376
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8452795547671727e-05,
+      "loss": 1.0781,
+      "step": 6377
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8452241718510532e-05,
+      "loss": 1.375,
+      "step": 6378
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8451687798558406e-05,
+      "loss": 1.207,
+      "step": 6379
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84511337878213e-05,
+      "loss": 1.0713,
+      "step": 6380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8450579686305157e-05,
+      "loss": 1.3184,
+      "step": 6381
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845002549401594e-05,
+      "loss": 1.3281,
+      "step": 6382
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84494712109596e-05,
+      "loss": 0.9922,
+      "step": 6383
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844891683714209e-05,
+      "loss": 0.9531,
+      "step": 6384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844836237256936e-05,
+      "loss": 0.9727,
+      "step": 6385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8447807817247368e-05,
+      "loss": 1.2231,
+      "step": 6386
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844725317118208e-05,
+      "loss": 1.4883,
+      "step": 6387
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446698434379436e-05,
+      "loss": 1.1621,
+      "step": 6388
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446143606845412e-05,
+      "loss": 1.5723,
+      "step": 6389
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844558868858596e-05,
+      "loss": 1.0898,
+      "step": 6390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844503367960704e-05,
+      "loss": 1.0908,
+      "step": 6391
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8444478579914618e-05,
+      "loss": 1.1055,
+      "step": 6392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8443923389514654e-05,
+      "loss": 0.9023,
+      "step": 6393
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844336810841311e-05,
+      "loss": 1.3691,
+      "step": 6394
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844281273661595e-05,
+      "loss": 1.4238,
+      "step": 6395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8442257274129142e-05,
+      "loss": 1.3555,
+      "step": 6396
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8441701720958655e-05,
+      "loss": 1.3906,
+      "step": 6397
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8441146077110455e-05,
+      "loss": 1.1006,
+      "step": 6398
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8440590342590506e-05,
+      "loss": 1.4316,
+      "step": 6399
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8440034517404783e-05,
+      "loss": 1.0903,
+      "step": 6400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439478601559254e-05,
+      "loss": 1.2295,
+      "step": 6401
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843892259505989e-05,
+      "loss": 1.2041,
+      "step": 6402
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8438366497912665e-05,
+      "loss": 1.1465,
+      "step": 6403
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8437810310123555e-05,
+      "loss": 1.3008,
+      "step": 6404
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8437254031698525e-05,
+      "loss": 1.3701,
+      "step": 6405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843669766264356e-05,
+      "loss": 1.4414,
+      "step": 6406
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8436141202964632e-05,
+      "loss": 1.4951,
+      "step": 6407
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843558465266772e-05,
+      "loss": 1.2412,
+      "step": 6408
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84350280117588e-05,
+      "loss": 1.1104,
+      "step": 6409
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434471280243854e-05,
+      "loss": 1.498,
+      "step": 6410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843391445812886e-05,
+      "loss": 1.2031,
+      "step": 6411
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84333575454198e-05,
+      "loss": 0.7461,
+      "step": 6412
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8432800542122656e-05,
+      "loss": 0.9355,
+      "step": 6413
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843224344824341e-05,
+      "loss": 0.9917,
+      "step": 6414
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843168626378805e-05,
+      "loss": 1.0771,
+      "step": 6415
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8431128988762556e-05,
+      "loss": 0.8789,
+      "step": 6416
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430571623172914e-05,
+      "loss": 1.4375,
+      "step": 6417
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843001416702512e-05,
+      "loss": 1.0547,
+      "step": 6418
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842945662032515e-05,
+      "loss": 1.208,
+      "step": 6419
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428898983079e-05,
+      "loss": 1.3457,
+      "step": 6420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428341255292657e-05,
+      "loss": 1.1562,
+      "step": 6421
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427783436972114e-05,
+      "loss": 1.3887,
+      "step": 6422
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427225528123366e-05,
+      "loss": 1.1826,
+      "step": 6423
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84266675287524e-05,
+      "loss": 1.2988,
+      "step": 6424
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842610943886521e-05,
+      "loss": 1.4824,
+      "step": 6425
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8425551258467792e-05,
+      "loss": 0.9023,
+      "step": 6426
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8424992987566144e-05,
+      "loss": 1.5742,
+      "step": 6427
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8424434626166263e-05,
+      "loss": 1.2437,
+      "step": 6428
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842387617427414e-05,
+      "loss": 1.1104,
+      "step": 6429
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423317631895784e-05,
+      "loss": 1.1758,
+      "step": 6430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8422758999037188e-05,
+      "loss": 1.3877,
+      "step": 6431
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842220027570435e-05,
+      "loss": 1.167,
+      "step": 6432
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842164146190328e-05,
+      "loss": 0.73,
+      "step": 6433
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8421082557639973e-05,
+      "loss": 1.6426,
+      "step": 6434
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420523562920435e-05,
+      "loss": 1.1602,
+      "step": 6435
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419964477750673e-05,
+      "loss": 1.293,
+      "step": 6436
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419405302136692e-05,
+      "loss": 1.249,
+      "step": 6437
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418846036084494e-05,
+      "loss": 1.334,
+      "step": 6438
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418286679600088e-05,
+      "loss": 0.8867,
+      "step": 6439
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8417727232689487e-05,
+      "loss": 1.1133,
+      "step": 6440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8417167695358698e-05,
+      "loss": 1.4277,
+      "step": 6441
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416608067613732e-05,
+      "loss": 0.9502,
+      "step": 6442
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416048349460592e-05,
+      "loss": 0.7036,
+      "step": 6443
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8415488540905303e-05,
+      "loss": 1.2881,
+      "step": 6444
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414928641953872e-05,
+      "loss": 1.0674,
+      "step": 6445
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841436865261231e-05,
+      "loss": 1.3223,
+      "step": 6446
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841380857288664e-05,
+      "loss": 1.0095,
+      "step": 6447
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413248402782872e-05,
+      "loss": 1.0127,
+      "step": 6448
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841268814230703e-05,
+      "loss": 1.1309,
+      "step": 6449
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841212779146512e-05,
+      "loss": 1.126,
+      "step": 6450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8411567350263172e-05,
+      "loss": 0.9932,
+      "step": 6451
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8411006818707203e-05,
+      "loss": 1.3496,
+      "step": 6452
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841044619680323e-05,
+      "loss": 1.6172,
+      "step": 6453
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409885484557283e-05,
+      "loss": 1.5664,
+      "step": 6454
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409324681975376e-05,
+      "loss": 1.2793,
+      "step": 6455
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840876378906354e-05,
+      "loss": 1.0254,
+      "step": 6456
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84082028058278e-05,
+      "loss": 2.0137,
+      "step": 6457
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8407641732274174e-05,
+      "loss": 1.1348,
+      "step": 6458
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84070805684087e-05,
+      "loss": 1.2178,
+      "step": 6459
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8406519314237395e-05,
+      "loss": 0.915,
+      "step": 6460
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8405957969766295e-05,
+      "loss": 1.2744,
+      "step": 6461
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8405396535001426e-05,
+      "loss": 1.4814,
+      "step": 6462
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840483500994882e-05,
+      "loss": 0.9902,
+      "step": 6463
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8404273394614512e-05,
+      "loss": 1.1982,
+      "step": 6464
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840371168900453e-05,
+      "loss": 1.3818,
+      "step": 6465
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403149893124907e-05,
+      "loss": 1.2383,
+      "step": 6466
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402588006981682e-05,
+      "loss": 1.1768,
+      "step": 6467
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402026030580887e-05,
+      "loss": 1.1494,
+      "step": 6468
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401463963928558e-05,
+      "loss": 1.2373,
+      "step": 6469
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840090180703074e-05,
+      "loss": 1.3789,
+      "step": 6470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840033955989346e-05,
+      "loss": 1.0981,
+      "step": 6471
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839977722252277e-05,
+      "loss": 1.0381,
+      "step": 6472
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8399214794924697e-05,
+      "loss": 1.1807,
+      "step": 6473
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398652277105293e-05,
+      "loss": 1.1416,
+      "step": 6474
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398089669070596e-05,
+      "loss": 1.4492,
+      "step": 6475
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839752697082665e-05,
+      "loss": 1.106,
+      "step": 6476
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396964182379497e-05,
+      "loss": 1.2246,
+      "step": 6477
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396401303735185e-05,
+      "loss": 1.1523,
+      "step": 6478
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839583833489976e-05,
+      "loss": 1.2568,
+      "step": 6479
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8395275275879273e-05,
+      "loss": 1.332,
+      "step": 6480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394712126679762e-05,
+      "loss": 1.2646,
+      "step": 6481
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394148887307286e-05,
+      "loss": 1.1211,
+      "step": 6482
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839358555776789e-05,
+      "loss": 1.084,
+      "step": 6483
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8393022138067626e-05,
+      "loss": 0.9434,
+      "step": 6484
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839245862821255e-05,
+      "loss": 1.3125,
+      "step": 6485
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391895028208707e-05,
+      "loss": 1.1172,
+      "step": 6486
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391331338062158e-05,
+      "loss": 1.2197,
+      "step": 6487
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390767557778954e-05,
+      "loss": 1.0654,
+      "step": 6488
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390203687365153e-05,
+      "loss": 0.8525,
+      "step": 6489
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838963972682681e-05,
+      "loss": 1.3184,
+      "step": 6490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389075676169988e-05,
+      "loss": 1.375,
+      "step": 6491
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388511535400736e-05,
+      "loss": 1.3789,
+      "step": 6492
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8387947304525125e-05,
+      "loss": 1.0928,
+      "step": 6493
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8387382983549207e-05,
+      "loss": 1.4219,
+      "step": 6494
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838681857247905e-05,
+      "loss": 1.334,
+      "step": 6495
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8386254071320714e-05,
+      "loss": 1.3438,
+      "step": 6496
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838568948008026e-05,
+      "loss": 1.4971,
+      "step": 6497
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8385124798763755e-05,
+      "loss": 1.25,
+      "step": 6498
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8384560027377264e-05,
+      "loss": 1.3906,
+      "step": 6499
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838399516592686e-05,
+      "loss": 1.2734,
+      "step": 6500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83834302144186e-05,
+      "loss": 1.0942,
+      "step": 6501
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8382865172858558e-05,
+      "loss": 1.3486,
+      "step": 6502
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8382300041252804e-05,
+      "loss": 0.9053,
+      "step": 6503
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838173481960741e-05,
+      "loss": 1.126,
+      "step": 6504
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838116950792844e-05,
+      "loss": 1.3252,
+      "step": 6505
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8380604106221973e-05,
+      "loss": 1.4727,
+      "step": 6506
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8380038614494083e-05,
+      "loss": 1.0439,
+      "step": 6507
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837947303275084e-05,
+      "loss": 1.2363,
+      "step": 6508
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837890736099832e-05,
+      "loss": 1.1064,
+      "step": 6509
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8378341599242602e-05,
+      "loss": 1.0127,
+      "step": 6510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377775747489758e-05,
+      "loss": 1.3848,
+      "step": 6511
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377209805745873e-05,
+      "loss": 1.1553,
+      "step": 6512
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8376643774017024e-05,
+      "loss": 0.9287,
+      "step": 6513
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8376077652309284e-05,
+      "loss": 1.0527,
+      "step": 6514
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8375511440628745e-05,
+      "loss": 0.9277,
+      "step": 6515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8374945138981483e-05,
+      "loss": 1.0605,
+      "step": 6516
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8374378747373583e-05,
+      "loss": 1.2158,
+      "step": 6517
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8373812265811126e-05,
+      "loss": 1.2412,
+      "step": 6518
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83732456943002e-05,
+      "loss": 1.248,
+      "step": 6519
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837267903284689e-05,
+      "loss": 0.6865,
+      "step": 6520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837211228145728e-05,
+      "loss": 0.9448,
+      "step": 6521
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371545440137467e-05,
+      "loss": 1.252,
+      "step": 6522
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837097850889353e-05,
+      "loss": 1.127,
+      "step": 6523
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370411487731562e-05,
+      "loss": 0.9795,
+      "step": 6524
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8369844376657654e-05,
+      "loss": 1.4707,
+      "step": 6525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8369277175677897e-05,
+      "loss": 1.3076,
+      "step": 6526
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8368709884798383e-05,
+      "loss": 1.2705,
+      "step": 6527
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836814250402521e-05,
+      "loss": 0.7993,
+      "step": 6528
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836757503336447e-05,
+      "loss": 0.9014,
+      "step": 6529
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8367007472822253e-05,
+      "loss": 0.6406,
+      "step": 6530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8366439822404666e-05,
+      "loss": 1.2466,
+      "step": 6531
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8365872082117798e-05,
+      "loss": 1.2988,
+      "step": 6532
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836530425196775e-05,
+      "loss": 1.3955,
+      "step": 6533
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8364736331960624e-05,
+      "loss": 1.0674,
+      "step": 6534
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836416832210252e-05,
+      "loss": 1.1807,
+      "step": 6535
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8363600222399534e-05,
+      "loss": 0.936,
+      "step": 6536
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836303203285777e-05,
+      "loss": 1.2539,
+      "step": 6537
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836246375348334e-05,
+      "loss": 1.2646,
+      "step": 6538
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836189538428234e-05,
+      "loss": 1.3154,
+      "step": 6539
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8361326925260873e-05,
+      "loss": 1.2275,
+      "step": 6540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8360758376425054e-05,
+      "loss": 1.2129,
+      "step": 6541
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836018973778098e-05,
+      "loss": 1.1807,
+      "step": 6542
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835962100933477e-05,
+      "loss": 1.1494,
+      "step": 6543
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8359052191092524e-05,
+      "loss": 1.3438,
+      "step": 6544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8358483283060353e-05,
+      "loss": 1.3848,
+      "step": 6545
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8357914285244373e-05,
+      "loss": 1.6299,
+      "step": 6546
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8357345197650694e-05,
+      "loss": 0.9346,
+      "step": 6547
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8356776020285427e-05,
+      "loss": 1.2363,
+      "step": 6548
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8356206753154692e-05,
+      "loss": 1.584,
+      "step": 6549
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8355637396264593e-05,
+      "loss": 1.2354,
+      "step": 6550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8355067949621254e-05,
+      "loss": 1.001,
+      "step": 6551
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835449841323079e-05,
+      "loss": 0.897,
+      "step": 6552
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835392878709932e-05,
+      "loss": 0.8633,
+      "step": 6553
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353359071232954e-05,
+      "loss": 1.0996,
+      "step": 6554
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352789265637824e-05,
+      "loss": 1.3496,
+      "step": 6555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835221937032005e-05,
+      "loss": 1.2148,
+      "step": 6556
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835164938528574e-05,
+      "loss": 1.0752,
+      "step": 6557
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8351079310541032e-05,
+      "loss": 1.4199,
+      "step": 6558
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835050914609204e-05,
+      "loss": 0.9736,
+      "step": 6559
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8349938891944893e-05,
+      "loss": 1.1113,
+      "step": 6560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8349368548105716e-05,
+      "loss": 1.0508,
+      "step": 6561
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834879811458063e-05,
+      "loss": 0.8271,
+      "step": 6562
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8348227591375773e-05,
+      "loss": 1.4883,
+      "step": 6563
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347656978497265e-05,
+      "loss": 1.1504,
+      "step": 6564
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347086275951236e-05,
+      "loss": 1.1025,
+      "step": 6565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8346515483743822e-05,
+      "loss": 1.2109,
+      "step": 6566
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8345944601881145e-05,
+      "loss": 1.1836,
+      "step": 6567
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8345373630369344e-05,
+      "loss": 1.3359,
+      "step": 6568
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8344802569214556e-05,
+      "loss": 1.1943,
+      "step": 6569
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8344231418422903e-05,
+      "loss": 1.1426,
+      "step": 6570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8343660178000527e-05,
+      "loss": 0.9961,
+      "step": 6571
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8343088847953568e-05,
+      "loss": 1.3477,
+      "step": 6572
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8342517428288155e-05,
+      "loss": 0.8091,
+      "step": 6573
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8341945919010434e-05,
+      "loss": 1.2148,
+      "step": 6574
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8341374320126537e-05,
+      "loss": 1.1621,
+      "step": 6575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8340802631642607e-05,
+      "loss": 1.0459,
+      "step": 6576
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8340230853564784e-05,
+      "loss": 1.2754,
+      "step": 6577
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833965898589921e-05,
+      "loss": 1.2158,
+      "step": 6578
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833908702865203e-05,
+      "loss": 0.9297,
+      "step": 6579
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8338514981829387e-05,
+      "loss": 0.9023,
+      "step": 6580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337942845437423e-05,
+      "loss": 0.8818,
+      "step": 6581
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337370619482282e-05,
+      "loss": 1.2798,
+      "step": 6582
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336798303970118e-05,
+      "loss": 1.1035,
+      "step": 6583
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336225898907075e-05,
+      "loss": 1.3945,
+      "step": 6584
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8335653404299298e-05,
+      "loss": 1.1953,
+      "step": 6585
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833508082015294e-05,
+      "loss": 1.0674,
+      "step": 6586
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8334508146474156e-05,
+      "loss": 1.1777,
+      "step": 6587
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833393538326909e-05,
+      "loss": 1.1289,
+      "step": 6588
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333362530543893e-05,
+      "loss": 0.9658,
+      "step": 6589
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8332789588304726e-05,
+      "loss": 1.1104,
+      "step": 6590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8332216556557744e-05,
+      "loss": 1.3467,
+      "step": 6591
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331643435309093e-05,
+      "loss": 0.707,
+      "step": 6592
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331070224564934e-05,
+      "loss": 1.3018,
+      "step": 6593
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8330496924331425e-05,
+      "loss": 1.2559,
+      "step": 6594
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8329923534614723e-05,
+      "loss": 1.1621,
+      "step": 6595
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832935005542099e-05,
+      "loss": 1.3027,
+      "step": 6596
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328776486756385e-05,
+      "loss": 1.207,
+      "step": 6597
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328202828627068e-05,
+      "loss": 1.3594,
+      "step": 6598
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8327629081039196e-05,
+      "loss": 1.3252,
+      "step": 6599
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832705524399894e-05,
+      "loss": 1.0498,
+      "step": 6600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8326481317512465e-05,
+      "loss": 1.0957,
+      "step": 6601
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8325907301585924e-05,
+      "loss": 1.2646,
+      "step": 6602
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8325333196225498e-05,
+      "loss": 1.1357,
+      "step": 6603
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324759001437346e-05,
+      "loss": 1.25,
+      "step": 6604
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324184717227634e-05,
+      "loss": 1.2412,
+      "step": 6605
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8323610343602533e-05,
+      "loss": 1.4033,
+      "step": 6606
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832303588056821e-05,
+      "loss": 0.7466,
+      "step": 6607
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322461328130846e-05,
+      "loss": 0.9131,
+      "step": 6608
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83218866862966e-05,
+      "loss": 1.0439,
+      "step": 6609
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832131195507165e-05,
+      "loss": 1.4961,
+      "step": 6610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832073713446217e-05,
+      "loss": 1.7656,
+      "step": 6611
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832016222447434e-05,
+      "loss": 1.0703,
+      "step": 6612
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831958722511432e-05,
+      "loss": 1.1719,
+      "step": 6613
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83190121363883e-05,
+      "loss": 0.8081,
+      "step": 6614
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8318436958302454e-05,
+      "loss": 1.0176,
+      "step": 6615
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8317861690862958e-05,
+      "loss": 1.2852,
+      "step": 6616
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8317286334075995e-05,
+      "loss": 1.2178,
+      "step": 6617
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831671088794774e-05,
+      "loss": 0.9971,
+      "step": 6618
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831613535248438e-05,
+      "loss": 0.9785,
+      "step": 6619
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8315559727692093e-05,
+      "loss": 1.2324,
+      "step": 6620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314984013577064e-05,
+      "loss": 1.0732,
+      "step": 6621
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314408210145476e-05,
+      "loss": 1.1318,
+      "step": 6622
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8313832317403518e-05,
+      "loss": 1.2812,
+      "step": 6623
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831325633535737e-05,
+      "loss": 1.0635,
+      "step": 6624
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312680264013224e-05,
+      "loss": 0.8936,
+      "step": 6625
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312104103377266e-05,
+      "loss": 1.4512,
+      "step": 6626
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8311527853455683e-05,
+      "loss": 1.0596,
+      "step": 6627
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8310951514254665e-05,
+      "loss": 0.9805,
+      "step": 6628
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8310375085780407e-05,
+      "loss": 1.2607,
+      "step": 6629
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83097985680391e-05,
+      "loss": 0.9443,
+      "step": 6630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8309221961036933e-05,
+      "loss": 1.0918,
+      "step": 6631
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83086452647801e-05,
+      "loss": 0.9497,
+      "step": 6632
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83080684792748e-05,
+      "loss": 1.0879,
+      "step": 6633
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8307491604527224e-05,
+      "loss": 1.2598,
+      "step": 6634
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830691464054357e-05,
+      "loss": 1.2246,
+      "step": 6635
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830633758733004e-05,
+      "loss": 1.1982,
+      "step": 6636
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8305760444892825e-05,
+      "loss": 1.3311,
+      "step": 6637
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830518321323813e-05,
+      "loss": 0.7393,
+      "step": 6638
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8304605892372152e-05,
+      "loss": 1.416,
+      "step": 6639
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83040284823011e-05,
+      "loss": 0.7803,
+      "step": 6640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8303450983031163e-05,
+      "loss": 0.9834,
+      "step": 6641
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8302873394568553e-05,
+      "loss": 1.5518,
+      "step": 6642
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8302295716919473e-05,
+      "loss": 1.3906,
+      "step": 6643
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830171795009013e-05,
+      "loss": 0.8208,
+      "step": 6644
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8301140094086725e-05,
+      "loss": 0.9385,
+      "step": 6645
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8300562148915468e-05,
+      "loss": 1.1562,
+      "step": 6646
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8299984114582572e-05,
+      "loss": 1.2988,
+      "step": 6647
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8299405991094237e-05,
+      "loss": 1.123,
+      "step": 6648
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829882777845668e-05,
+      "loss": 1.1484,
+      "step": 6649
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8298249476676107e-05,
+      "loss": 0.917,
+      "step": 6650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829767108575873e-05,
+      "loss": 1.0762,
+      "step": 6651
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829709260571077e-05,
+      "loss": 1.3418,
+      "step": 6652
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829651403653843e-05,
+      "loss": 0.9111,
+      "step": 6653
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8295935378247934e-05,
+      "loss": 1.2168,
+      "step": 6654
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829535663084549e-05,
+      "loss": 1.3486,
+      "step": 6655
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829477779433732e-05,
+      "loss": 1.2061,
+      "step": 6656
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829419886872964e-05,
+      "loss": 1.5317,
+      "step": 6657
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8293619854028666e-05,
+      "loss": 0.8774,
+      "step": 6658
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8293040750240624e-05,
+      "loss": 0.7363,
+      "step": 6659
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8292461557371726e-05,
+      "loss": 1.2559,
+      "step": 6660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82918822754282e-05,
+      "loss": 0.9824,
+      "step": 6661
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8291302904416266e-05,
+      "loss": 1.2373,
+      "step": 6662
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290723444342148e-05,
+      "loss": 1.2607,
+      "step": 6663
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829014389521207e-05,
+      "loss": 1.377,
+      "step": 6664
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828956425703226e-05,
+      "loss": 1.0801,
+      "step": 6665
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288984529808937e-05,
+      "loss": 1.2354,
+      "step": 6666
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288404713548334e-05,
+      "loss": 1.2588,
+      "step": 6667
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828782480825668e-05,
+      "loss": 1.1377,
+      "step": 6668
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82872448139402e-05,
+      "loss": 1.3154,
+      "step": 6669
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828666473060513e-05,
+      "loss": 1.3145,
+      "step": 6670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8286084558257695e-05,
+      "loss": 1.1016,
+      "step": 6671
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828550429690413e-05,
+      "loss": 1.0566,
+      "step": 6672
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8284923946550667e-05,
+      "loss": 1.2178,
+      "step": 6673
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828434350720354e-05,
+      "loss": 0.9551,
+      "step": 6674
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8283762978868982e-05,
+      "loss": 1.4141,
+      "step": 6675
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8283182361553234e-05,
+      "loss": 1.4316,
+      "step": 6676
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828260165526253e-05,
+      "loss": 1.6523,
+      "step": 6677
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282020860003106e-05,
+      "loss": 0.8926,
+      "step": 6678
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82814399757812e-05,
+      "loss": 1.1738,
+      "step": 6679
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828085900260306e-05,
+      "loss": 1.002,
+      "step": 6680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280277940474914e-05,
+      "loss": 0.8662,
+      "step": 6681
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8279696789403014e-05,
+      "loss": 0.8672,
+      "step": 6682
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8279115549393596e-05,
+      "loss": 1.1045,
+      "step": 6683
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8278534220452906e-05,
+      "loss": 1.0186,
+      "step": 6684
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8277952802587193e-05,
+      "loss": 0.9165,
+      "step": 6685
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8277371295802692e-05,
+      "loss": 1.4277,
+      "step": 6686
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8276789700105654e-05,
+      "loss": 1.1357,
+      "step": 6687
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8276208015502334e-05,
+      "loss": 0.7422,
+      "step": 6688
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827562624199897e-05,
+      "loss": 1.3965,
+      "step": 6689
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8275044379601814e-05,
+      "loss": 1.1152,
+      "step": 6690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827446242831712e-05,
+      "loss": 0.9404,
+      "step": 6691
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827388038815113e-05,
+      "loss": 1.0557,
+      "step": 6692
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8273298259110107e-05,
+      "loss": 0.9697,
+      "step": 6693
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272716041200302e-05,
+      "loss": 1.0605,
+      "step": 6694
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272133734427958e-05,
+      "loss": 1.1504,
+      "step": 6695
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8271551338799344e-05,
+      "loss": 1.0864,
+      "step": 6696
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827096885432071e-05,
+      "loss": 1.2354,
+      "step": 6697
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827038628099831e-05,
+      "loss": 1.5957,
+      "step": 6698
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8269803618838406e-05,
+      "loss": 1.0361,
+      "step": 6699
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8269220867847256e-05,
+      "loss": 1.5117,
+      "step": 6700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826863802803112e-05,
+      "loss": 1.3008,
+      "step": 6701
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8268055099396256e-05,
+      "loss": 0.999,
+      "step": 6702
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8267472081948927e-05,
+      "loss": 1.1943,
+      "step": 6703
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8266888975695395e-05,
+      "loss": 1.5879,
+      "step": 6704
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8266305780641924e-05,
+      "loss": 1.3555,
+      "step": 6705
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265722496794782e-05,
+      "loss": 1.0557,
+      "step": 6706
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826513912416023e-05,
+      "loss": 1.1963,
+      "step": 6707
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8264555662744533e-05,
+      "loss": 1.375,
+      "step": 6708
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8263972112553965e-05,
+      "loss": 0.8213,
+      "step": 6709
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826338847359479e-05,
+      "loss": 1.4375,
+      "step": 6710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8262804745873274e-05,
+      "loss": 1.375,
+      "step": 6711
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8262220929395694e-05,
+      "loss": 0.9717,
+      "step": 6712
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8261637024168316e-05,
+      "loss": 1.3789,
+      "step": 6713
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8261053030197414e-05,
+      "loss": 1.002,
+      "step": 6714
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8260468947489264e-05,
+      "loss": 1.0322,
+      "step": 6715
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259884776050133e-05,
+      "loss": 1.3184,
+      "step": 6716
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259300515886304e-05,
+      "loss": 1.2393,
+      "step": 6717
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8258716167004044e-05,
+      "loss": 0.9932,
+      "step": 6718
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825813172940964e-05,
+      "loss": 1.1226,
+      "step": 6719
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825754720310936e-05,
+      "loss": 1.1104,
+      "step": 6720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825696258810949e-05,
+      "loss": 1.1914,
+      "step": 6721
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8256377884416307e-05,
+      "loss": 0.9512,
+      "step": 6722
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825579309203609e-05,
+      "loss": 1.1533,
+      "step": 6723
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8255208210975124e-05,
+      "loss": 0.9941,
+      "step": 6724
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825462324123969e-05,
+      "loss": 1.1396,
+      "step": 6725
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825403818283607e-05,
+      "loss": 0.8716,
+      "step": 6726
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825345303577055e-05,
+      "loss": 1.2178,
+      "step": 6727
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8252867800049418e-05,
+      "loss": 1.0869,
+      "step": 6728
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8252282475678953e-05,
+      "loss": 1.2354,
+      "step": 6729
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825169706266545e-05,
+      "loss": 1.2617,
+      "step": 6730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8251111561015195e-05,
+      "loss": 1.1621,
+      "step": 6731
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250525970734475e-05,
+      "loss": 1.0684,
+      "step": 6732
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8249940291829585e-05,
+      "loss": 1.1699,
+      "step": 6733
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824935452430681e-05,
+      "loss": 1.167,
+      "step": 6734
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8248768668172447e-05,
+      "loss": 0.9512,
+      "step": 6735
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8248182723432788e-05,
+      "loss": 1.1182,
+      "step": 6736
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247596690094124e-05,
+      "loss": 1.2324,
+      "step": 6737
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247010568162752e-05,
+      "loss": 1.2578,
+      "step": 6738
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824642435764497e-05,
+      "loss": 1.0942,
+      "step": 6739
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824583805854707e-05,
+      "loss": 1.1855,
+      "step": 6740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824525167087536e-05,
+      "loss": 1.1338,
+      "step": 6741
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8244665194636122e-05,
+      "loss": 0.9062,
+      "step": 6742
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8244078629835668e-05,
+      "loss": 1.2354,
+      "step": 6743
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8243491976480298e-05,
+      "loss": 1.2295,
+      "step": 6744
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824290523457631e-05,
+      "loss": 1.0498,
+      "step": 6745
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824231840413001e-05,
+      "loss": 1.1104,
+      "step": 6746
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82417314851477e-05,
+      "loss": 1.3877,
+      "step": 6747
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824114447763568e-05,
+      "loss": 1.0137,
+      "step": 6748
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824055738160026e-05,
+      "loss": 1.6074,
+      "step": 6749
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823997019704775e-05,
+      "loss": 1.0918,
+      "step": 6750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823938292398445e-05,
+      "loss": 0.8955,
+      "step": 6751
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8238795562416672e-05,
+      "loss": 1.0249,
+      "step": 6752
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8238208112350727e-05,
+      "loss": 0.9854,
+      "step": 6753
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823762057379292e-05,
+      "loss": 1.4111,
+      "step": 6754
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8237032946749568e-05,
+      "loss": 1.1875,
+      "step": 6755
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8236445231226974e-05,
+      "loss": 1.0684,
+      "step": 6756
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235857427231464e-05,
+      "loss": 1.5059,
+      "step": 6757
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235269534769342e-05,
+      "loss": 1.0146,
+      "step": 6758
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8234681553846927e-05,
+      "loss": 1.1416,
+      "step": 6759
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8234093484470533e-05,
+      "loss": 1.1626,
+      "step": 6760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823350532664648e-05,
+      "loss": 1.3984,
+      "step": 6761
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823291708038108e-05,
+      "loss": 1.2979,
+      "step": 6762
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8232328745680656e-05,
+      "loss": 1.5303,
+      "step": 6763
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823174032255153e-05,
+      "loss": 1.1465,
+      "step": 6764
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8231151811000015e-05,
+      "loss": 0.8691,
+      "step": 6765
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8230563211032438e-05,
+      "loss": 0.7803,
+      "step": 6766
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8229974522655123e-05,
+      "loss": 1.1924,
+      "step": 6767
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8229385745874392e-05,
+      "loss": 0.8613,
+      "step": 6768
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8228796880696564e-05,
+      "loss": 1.6221,
+      "step": 6769
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822820792712797e-05,
+      "loss": 0.873,
+      "step": 6770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227618885174936e-05,
+      "loss": 1.1826,
+      "step": 6771
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822702975484379e-05,
+      "loss": 1.2451,
+      "step": 6772
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8226440536140863e-05,
+      "loss": 1.0371,
+      "step": 6773
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822585122907247e-05,
+      "loss": 1.1191,
+      "step": 6774
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8225261833644958e-05,
+      "loss": 1.2715,
+      "step": 6775
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822467234986465e-05,
+      "loss": 1.084,
+      "step": 6776
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8224082777737878e-05,
+      "loss": 0.9668,
+      "step": 6777
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8223493117270976e-05,
+      "loss": 0.8428,
+      "step": 6778
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822290336847028e-05,
+      "loss": 1.124,
+      "step": 6779
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822231353134212e-05,
+      "loss": 0.9434,
+      "step": 6780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8221723605892837e-05,
+      "loss": 0.8467,
+      "step": 6781
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8221133592128764e-05,
+      "loss": 1.1709,
+      "step": 6782
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220543490056238e-05,
+      "loss": 1.3164,
+      "step": 6783
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82199532996816e-05,
+      "loss": 0.916,
+      "step": 6784
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8219363021011197e-05,
+      "loss": 1.3105,
+      "step": 6785
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8218772654051357e-05,
+      "loss": 1.542,
+      "step": 6786
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8218182198808426e-05,
+      "loss": 1.0146,
+      "step": 6787
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821759165528875e-05,
+      "loss": 0.916,
+      "step": 6788
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8217001023498664e-05,
+      "loss": 1.2773,
+      "step": 6789
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821641030344452e-05,
+      "loss": 1.3506,
+      "step": 6790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8215819495132663e-05,
+      "loss": 1.3281,
+      "step": 6791
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8215228598569437e-05,
+      "loss": 1.1787,
+      "step": 6792
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821463761376119e-05,
+      "loss": 1.1279,
+      "step": 6793
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8214046540714268e-05,
+      "loss": 1.1064,
+      "step": 6794
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8213455379435024e-05,
+      "loss": 1.1689,
+      "step": 6795
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82128641299298e-05,
+      "loss": 1.5967,
+      "step": 6796
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8212272792204957e-05,
+      "loss": 1.2188,
+      "step": 6797
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211681366266843e-05,
+      "loss": 0.8613,
+      "step": 6798
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211089852121812e-05,
+      "loss": 0.7793,
+      "step": 6799
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8210498249776215e-05,
+      "loss": 1.2764,
+      "step": 6800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8209906559236407e-05,
+      "loss": 1.3018,
+      "step": 6801
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8209314780508745e-05,
+      "loss": 1.0605,
+      "step": 6802
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8208722913599587e-05,
+      "loss": 1.3506,
+      "step": 6803
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820813095851529e-05,
+      "loss": 1.124,
+      "step": 6804
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820753891526221e-05,
+      "loss": 1.0791,
+      "step": 6805
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206946783846708e-05,
+      "loss": 1.0166,
+      "step": 6806
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206354564275145e-05,
+      "loss": 1.1924,
+      "step": 6807
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8205762256553883e-05,
+      "loss": 1.2285,
+      "step": 6808
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820516986068928e-05,
+      "loss": 0.8311,
+      "step": 6809
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8204577376687707e-05,
+      "loss": 0.8857,
+      "step": 6810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8203984804555522e-05,
+      "loss": 1.3516,
+      "step": 6811
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8203392144299097e-05,
+      "loss": 1.2725,
+      "step": 6812
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8202799395924787e-05,
+      "loss": 1.3066,
+      "step": 6813
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820220655943897e-05,
+      "loss": 1.3574,
+      "step": 6814
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8201613634848007e-05,
+      "loss": 1.6113,
+      "step": 6815
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8201020622158272e-05,
+      "loss": 1.1729,
+      "step": 6816
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820042752137613e-05,
+      "loss": 0.9873,
+      "step": 6817
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8199834332507955e-05,
+      "loss": 1.5264,
+      "step": 6818
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819924105556012e-05,
+      "loss": 1.3672,
+      "step": 6819
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8198647690538993e-05,
+      "loss": 1.0674,
+      "step": 6820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8198054237450954e-05,
+      "loss": 1.2695,
+      "step": 6821
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819746069630237e-05,
+      "loss": 1.251,
+      "step": 6822
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8196867067099627e-05,
+      "loss": 1.2578,
+      "step": 6823
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819627334984909e-05,
+      "loss": 1.084,
+      "step": 6824
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8195679544557145e-05,
+      "loss": 1.1367,
+      "step": 6825
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8195085651230165e-05,
+      "loss": 1.0166,
+      "step": 6826
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8194491669874535e-05,
+      "loss": 1.4375,
+      "step": 6827
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819389760049663e-05,
+      "loss": 1.0664,
+      "step": 6828
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819330344310283e-05,
+      "loss": 0.8801,
+      "step": 6829
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192709197699527e-05,
+      "loss": 1.2051,
+      "step": 6830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192114864293097e-05,
+      "loss": 1.0098,
+      "step": 6831
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819152044288992e-05,
+      "loss": 1.1318,
+      "step": 6832
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8190925933496387e-05,
+      "loss": 1.1553,
+      "step": 6833
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8190331336118885e-05,
+      "loss": 1.2979,
+      "step": 6834
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81897366507638e-05,
+      "loss": 1.2793,
+      "step": 6835
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8189141877437517e-05,
+      "loss": 0.9783,
+      "step": 6836
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8188547016146426e-05,
+      "loss": 1.2998,
+      "step": 6837
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818795206689692e-05,
+      "loss": 1.2188,
+      "step": 6838
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8187357029695387e-05,
+      "loss": 1.541,
+      "step": 6839
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818676190454822e-05,
+      "loss": 1.4629,
+      "step": 6840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8186166691461803e-05,
+      "loss": 0.8857,
+      "step": 6841
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8185571390442542e-05,
+      "loss": 1.1855,
+      "step": 6842
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8184976001496826e-05,
+      "loss": 0.5918,
+      "step": 6843
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8184380524631056e-05,
+      "loss": 1.2075,
+      "step": 6844
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8183784959851617e-05,
+      "loss": 1.3086,
+      "step": 6845
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8183189307164913e-05,
+      "loss": 1.0449,
+      "step": 6846
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8182593566577345e-05,
+      "loss": 1.2441,
+      "step": 6847
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818199773809531e-05,
+      "loss": 1.6299,
+      "step": 6848
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8181401821725203e-05,
+      "loss": 1.541,
+      "step": 6849
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8180805817473436e-05,
+      "loss": 1.084,
+      "step": 6850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81802097253464e-05,
+      "loss": 1.3477,
+      "step": 6851
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81796135453505e-05,
+      "loss": 1.1465,
+      "step": 6852
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8179017277492147e-05,
+      "loss": 1.4072,
+      "step": 6853
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8178420921777743e-05,
+      "loss": 1.2061,
+      "step": 6854
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8177824478213687e-05,
+      "loss": 1.377,
+      "step": 6855
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8177227946806395e-05,
+      "loss": 0.7834,
+      "step": 6856
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817663132756227e-05,
+      "loss": 0.877,
+      "step": 6857
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817603462048772e-05,
+      "loss": 1.3828,
+      "step": 6858
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8175437825589156e-05,
+      "loss": 1.0322,
+      "step": 6859
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817484094287299e-05,
+      "loss": 1.2891,
+      "step": 6860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8174243972345635e-05,
+      "loss": 0.8799,
+      "step": 6861
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8173646914013498e-05,
+      "loss": 1.3018,
+      "step": 6862
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8173049767882995e-05,
+      "loss": 1.7891,
+      "step": 6863
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817245253396054e-05,
+      "loss": 1.2158,
+      "step": 6864
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8171855212252546e-05,
+      "loss": 0.9893,
+      "step": 6865
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8171257802765436e-05,
+      "loss": 1.3867,
+      "step": 6866
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817066030550562e-05,
+      "loss": 0.8662,
+      "step": 6867
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817006272047952e-05,
+      "loss": 0.999,
+      "step": 6868
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816946504769355e-05,
+      "loss": 1.3574,
+      "step": 6869
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816886728715414e-05,
+      "loss": 1.1055,
+      "step": 6870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8168269438867704e-05,
+      "loss": 1.2954,
+      "step": 6871
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8167671502840665e-05,
+      "loss": 1.1758,
+      "step": 6872
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8167073479079443e-05,
+      "loss": 1.2148,
+      "step": 6873
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8166475367590467e-05,
+      "loss": 1.1338,
+      "step": 6874
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816587716838016e-05,
+      "loss": 0.9072,
+      "step": 6875
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8165278881454943e-05,
+      "loss": 0.8564,
+      "step": 6876
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8164680506821247e-05,
+      "loss": 0.9043,
+      "step": 6877
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81640820444855e-05,
+      "loss": 1.1152,
+      "step": 6878
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816348349445413e-05,
+      "loss": 1.1011,
+      "step": 6879
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816288485673357e-05,
+      "loss": 1.2871,
+      "step": 6880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816228613133024e-05,
+      "loss": 0.7842,
+      "step": 6881
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8161687318250577e-05,
+      "loss": 1.5098,
+      "step": 6882
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8161088417501017e-05,
+      "loss": 1.1699,
+      "step": 6883
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8160489429087992e-05,
+      "loss": 1.0957,
+      "step": 6884
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815989035301793e-05,
+      "loss": 1.2676,
+      "step": 6885
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8159291189297267e-05,
+      "loss": 1.2363,
+      "step": 6886
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8158691937932445e-05,
+      "loss": 1.459,
+      "step": 6887
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81580925989299e-05,
+      "loss": 0.915,
+      "step": 6888
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8157493172296067e-05,
+      "loss": 1.0371,
+      "step": 6889
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8156893658037386e-05,
+      "loss": 1.2412,
+      "step": 6890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815629405616029e-05,
+      "loss": 1.2559,
+      "step": 6891
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155694366671235e-05,
+      "loss": 1.0049,
+      "step": 6892
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155094589576648e-05,
+      "loss": 1.3418,
+      "step": 6893
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815449472488298e-05,
+      "loss": 1.1309,
+      "step": 6894
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815389477259667e-05,
+      "loss": 0.8076,
+      "step": 6895
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815329473272417e-05,
+      "loss": 1.1631,
+      "step": 6896
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8152694605271912e-05,
+      "loss": 1.2227,
+      "step": 6897
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8152094390246353e-05,
+      "loss": 1.082,
+      "step": 6898
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8151494087653938e-05,
+      "loss": 1.3936,
+      "step": 6899
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8150893697501115e-05,
+      "loss": 0.9595,
+      "step": 6900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8150293219794332e-05,
+      "loss": 1.0762,
+      "step": 6901
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814969265454004e-05,
+      "loss": 1.1475,
+      "step": 6902
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814909200174469e-05,
+      "loss": 1.1445,
+      "step": 6903
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8148491261414733e-05,
+      "loss": 1.1084,
+      "step": 6904
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147890433556625e-05,
+      "loss": 0.9766,
+      "step": 6905
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814728951817682e-05,
+      "loss": 1.1768,
+      "step": 6906
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146688515281764e-05,
+      "loss": 1.0684,
+      "step": 6907
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146087424877927e-05,
+      "loss": 1.2217,
+      "step": 6908
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814548624697175e-05,
+      "loss": 1.6152,
+      "step": 6909
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8144884981569704e-05,
+      "loss": 0.8931,
+      "step": 6910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814428362867824e-05,
+      "loss": 1.125,
+      "step": 6911
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143682188303818e-05,
+      "loss": 1.3184,
+      "step": 6912
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143080660452905e-05,
+      "loss": 1.2637,
+      "step": 6913
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8142479045131956e-05,
+      "loss": 1.4668,
+      "step": 6914
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8141877342347434e-05,
+      "loss": 0.9512,
+      "step": 6915
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8141275552105803e-05,
+      "loss": 1.5137,
+      "step": 6916
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140673674413528e-05,
+      "loss": 1.1377,
+      "step": 6917
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140071709277075e-05,
+      "loss": 1.0225,
+      "step": 6918
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813946965670291e-05,
+      "loss": 1.6523,
+      "step": 6919
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8138867516697494e-05,
+      "loss": 1.0186,
+      "step": 6920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8138265289267304e-05,
+      "loss": 1.5059,
+      "step": 6921
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137662974418798e-05,
+      "loss": 1.3652,
+      "step": 6922
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137060572158457e-05,
+      "loss": 1.5879,
+      "step": 6923
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813645808249275e-05,
+      "loss": 1.1787,
+      "step": 6924
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8135855505428145e-05,
+      "loss": 1.2383,
+      "step": 6925
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8135252840971113e-05,
+      "loss": 1.1328,
+      "step": 6926
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134650089128128e-05,
+      "loss": 1.3838,
+      "step": 6927
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813404724990567e-05,
+      "loss": 1.3691,
+      "step": 6928
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813344432331021e-05,
+      "loss": 1.0928,
+      "step": 6929
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813284130934823e-05,
+      "loss": 1.6289,
+      "step": 6930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81322382080262e-05,
+      "loss": 0.8457,
+      "step": 6931
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8131635019350603e-05,
+      "loss": 0.9302,
+      "step": 6932
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8131031743327913e-05,
+      "loss": 1.4917,
+      "step": 6933
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130428379964616e-05,
+      "loss": 1.4844,
+      "step": 6934
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812982492926719e-05,
+      "loss": 1.1533,
+      "step": 6935
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8129221391242116e-05,
+      "loss": 1.3975,
+      "step": 6936
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8128617765895883e-05,
+      "loss": 1.1387,
+      "step": 6937
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812801405323497e-05,
+      "loss": 1.1758,
+      "step": 6938
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8127410253265863e-05,
+      "loss": 1.3945,
+      "step": 6939
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8126806365995046e-05,
+      "loss": 1.209,
+      "step": 6940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8126202391429007e-05,
+      "loss": 0.7559,
+      "step": 6941
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8125598329574234e-05,
+      "loss": 1.2031,
+      "step": 6942
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812499418043722e-05,
+      "loss": 1.2334,
+      "step": 6943
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8124389944024442e-05,
+      "loss": 1.1553,
+      "step": 6944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8123785620342404e-05,
+      "loss": 1.0029,
+      "step": 6945
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812318120939759e-05,
+      "loss": 1.3574,
+      "step": 6946
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8122576711196495e-05,
+      "loss": 1.0649,
+      "step": 6947
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812197212574561e-05,
+      "loss": 1.0449,
+      "step": 6948
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8121367453051434e-05,
+      "loss": 0.8311,
+      "step": 6949
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812076269312046e-05,
+      "loss": 1.1553,
+      "step": 6950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812015784595918e-05,
+      "loss": 0.8438,
+      "step": 6951
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8119552911574094e-05,
+      "loss": 1.4287,
+      "step": 6952
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8118947889971702e-05,
+      "loss": 1.0742,
+      "step": 6953
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81183427811585e-05,
+      "loss": 0.7568,
+      "step": 6954
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811773758514099e-05,
+      "loss": 1.249,
+      "step": 6955
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811713230192567e-05,
+      "loss": 1.2119,
+      "step": 6956
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8116526931519045e-05,
+      "loss": 0.9658,
+      "step": 6957
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115921473927617e-05,
+      "loss": 0.9248,
+      "step": 6958
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115315929157887e-05,
+      "loss": 1.0811,
+      "step": 6959
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8114710297216365e-05,
+      "loss": 1.3262,
+      "step": 6960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811410457810955e-05,
+      "loss": 0.9209,
+      "step": 6961
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811349877184395e-05,
+      "loss": 1.1045,
+      "step": 6962
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8112892878426078e-05,
+      "loss": 1.1836,
+      "step": 6963
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8112286897862433e-05,
+      "loss": 1.3574,
+      "step": 6964
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811168083015953e-05,
+      "loss": 1.458,
+      "step": 6965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8111074675323877e-05,
+      "loss": 1.3145,
+      "step": 6966
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8110468433361993e-05,
+      "loss": 1.332,
+      "step": 6967
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8109862104280374e-05,
+      "loss": 1.0664,
+      "step": 6968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810925568808555e-05,
+      "loss": 1.5869,
+      "step": 6969
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8108649184784024e-05,
+      "loss": 1.0752,
+      "step": 6970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8108042594382316e-05,
+      "loss": 1.7285,
+      "step": 6971
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810743591688694e-05,
+      "loss": 1.3584,
+      "step": 6972
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810682915230441e-05,
+      "loss": 0.918,
+      "step": 6973
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8106222300641246e-05,
+      "loss": 1.0957,
+      "step": 6974
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810561536190397e-05,
+      "loss": 0.9287,
+      "step": 6975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8105008336099098e-05,
+      "loss": 0.8691,
+      "step": 6976
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810440122323315e-05,
+      "loss": 1.1562,
+      "step": 6977
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8103794023312645e-05,
+      "loss": 1.2852,
+      "step": 6978
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8103186736344113e-05,
+      "loss": 1.0742,
+      "step": 6979
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8102579362334072e-05,
+      "loss": 0.9746,
+      "step": 6980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8101971901289044e-05,
+      "loss": 1.0396,
+      "step": 6981
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810136435321556e-05,
+      "loss": 1.1016,
+      "step": 6982
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8100756718120144e-05,
+      "loss": 1.375,
+      "step": 6983
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810014899600932e-05,
+      "loss": 1.1543,
+      "step": 6984
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8099541186889618e-05,
+      "loss": 1.1367,
+      "step": 6985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809893329076757e-05,
+      "loss": 1.0381,
+      "step": 6986
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.80983253076497e-05,
+      "loss": 1.2266,
+      "step": 6987
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809771723754254e-05,
+      "loss": 1.2266,
+      "step": 6988
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8097109080452627e-05,
+      "loss": 1.1821,
+      "step": 6989
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8096500836386487e-05,
+      "loss": 1.4541,
+      "step": 6990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8095892505350656e-05,
+      "loss": 1.1782,
+      "step": 6991
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809528408735167e-05,
+      "loss": 0.833,
+      "step": 6992
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8094675582396063e-05,
+      "loss": 1.083,
+      "step": 6993
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809406699049037e-05,
+      "loss": 1.1172,
+      "step": 6994
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809345831164113e-05,
+      "loss": 1.2393,
+      "step": 6995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8092849545854884e-05,
+      "loss": 1.3711,
+      "step": 6996
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8092240693138166e-05,
+      "loss": 1.29,
+      "step": 6997
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809163175349752e-05,
+      "loss": 1.0195,
+      "step": 6998
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809102272693948e-05,
+      "loss": 1.083,
+      "step": 6999
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.80904136134706e-05,
+      "loss": 1.2305,
+      "step": 7000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089804413097415e-05,
+      "loss": 1.124,
+      "step": 7001
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089195125826466e-05,
+      "loss": 1.668,
+      "step": 7002
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088585751664305e-05,
+      "loss": 0.9204,
+      "step": 7003
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808797629061747e-05,
+      "loss": 1.2998,
+      "step": 7004
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8087366742692518e-05,
+      "loss": 1.3301,
+      "step": 7005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8086757107895985e-05,
+      "loss": 1.127,
+      "step": 7006
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808614738623443e-05,
+      "loss": 1.252,
+      "step": 7007
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8085537577714393e-05,
+      "loss": 1.042,
+      "step": 7008
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808492768234243e-05,
+      "loss": 1.0342,
+      "step": 7009
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808431770012509e-05,
+      "loss": 1.3379,
+      "step": 7010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808370763106893e-05,
+      "loss": 1.6748,
+      "step": 7011
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8083097475180494e-05,
+      "loss": 0.8936,
+      "step": 7012
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8082487232466345e-05,
+      "loss": 0.9678,
+      "step": 7013
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8081876902933034e-05,
+      "loss": 1.0059,
+      "step": 7014
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8081266486587116e-05,
+      "loss": 0.9004,
+      "step": 7015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808065598343515e-05,
+      "loss": 0.7178,
+      "step": 7016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808004539348369e-05,
+      "loss": 1.1504,
+      "step": 7017
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.80794347167393e-05,
+      "loss": 1.0669,
+      "step": 7018
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807882395320854e-05,
+      "loss": 0.9707,
+      "step": 7019
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8078213102897967e-05,
+      "loss": 1.1279,
+      "step": 7020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077602165814143e-05,
+      "loss": 1.5938,
+      "step": 7021
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807699114196363e-05,
+      "loss": 1.1455,
+      "step": 7022
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8076380031352993e-05,
+      "loss": 0.98,
+      "step": 7023
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8075768833988795e-05,
+      "loss": 1.4014,
+      "step": 7024
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8075157549877603e-05,
+      "loss": 1.0811,
+      "step": 7025
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807454617902598e-05,
+      "loss": 0.7979,
+      "step": 7026
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.80739347214405e-05,
+      "loss": 1.1582,
+      "step": 7027
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073323177127723e-05,
+      "loss": 1.7754,
+      "step": 7028
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807271154609422e-05,
+      "loss": 1.1914,
+      "step": 7029
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8072099828346563e-05,
+      "loss": 1.2305,
+      "step": 7030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8071488023891325e-05,
+      "loss": 1.0664,
+      "step": 7031
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8070876132735075e-05,
+      "loss": 1.0049,
+      "step": 7032
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.807026415488438e-05,
+      "loss": 1.1768,
+      "step": 7033
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8069652090345825e-05,
+      "loss": 0.9922,
+      "step": 7034
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8069039939125977e-05,
+      "loss": 1.4102,
+      "step": 7035
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8068427701231413e-05,
+      "loss": 1.0361,
+      "step": 7036
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806781537666871e-05,
+      "loss": 1.416,
+      "step": 7037
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8067202965444444e-05,
+      "loss": 1.0024,
+      "step": 7038
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8066590467565196e-05,
+      "loss": 1.2969,
+      "step": 7039
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8065977883037543e-05,
+      "loss": 1.0352,
+      "step": 7040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8065365211868066e-05,
+      "loss": 1.2266,
+      "step": 7041
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8064752454063347e-05,
+      "loss": 1.041,
+      "step": 7042
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8064139609629963e-05,
+      "loss": 1.3965,
+      "step": 7043
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8063526678574508e-05,
+      "loss": 1.0664,
+      "step": 7044
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8062913660903554e-05,
+      "loss": 1.1133,
+      "step": 7045
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806230055662369e-05,
+      "loss": 1.1914,
+      "step": 7046
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8061687365741506e-05,
+      "loss": 1.5117,
+      "step": 7047
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8061074088263582e-05,
+      "loss": 1.251,
+      "step": 7048
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8060460724196505e-05,
+      "loss": 1.0918,
+      "step": 7049
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8059847273546876e-05,
+      "loss": 1.4316,
+      "step": 7050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805923373632127e-05,
+      "loss": 1.167,
+      "step": 7051
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8058620112526286e-05,
+      "loss": 1.1523,
+      "step": 7052
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8058006402168508e-05,
+      "loss": 1.5938,
+      "step": 7053
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8057392605254537e-05,
+      "loss": 1.3184,
+      "step": 7054
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805677872179096e-05,
+      "loss": 1.4961,
+      "step": 7055
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8056164751784373e-05,
+      "loss": 1.3232,
+      "step": 7056
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8055550695241374e-05,
+      "loss": 1.2383,
+      "step": 7057
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8054936552168548e-05,
+      "loss": 1.3252,
+      "step": 7058
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8054322322572506e-05,
+      "loss": 1.1846,
+      "step": 7059
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8053708006459837e-05,
+      "loss": 1.1582,
+      "step": 7060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8053093603837146e-05,
+      "loss": 1.1201,
+      "step": 7061
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805247911471103e-05,
+      "loss": 1.2305,
+      "step": 7062
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8051864539088086e-05,
+      "loss": 1.3916,
+      "step": 7063
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8051249876974916e-05,
+      "loss": 0.958,
+      "step": 7064
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8050635128378128e-05,
+      "loss": 1.6729,
+      "step": 7065
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805002029330432e-05,
+      "loss": 0.7559,
+      "step": 7066
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8049405371760098e-05,
+      "loss": 0.9766,
+      "step": 7067
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804879036375207e-05,
+      "loss": 0.9873,
+      "step": 7068
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048175269286837e-05,
+      "loss": 1.7129,
+      "step": 7069
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8047560088371012e-05,
+      "loss": 1.0381,
+      "step": 7070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80469448210112e-05,
+      "loss": 1.25,
+      "step": 7071
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8046329467214012e-05,
+      "loss": 1.293,
+      "step": 7072
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045714026986054e-05,
+      "loss": 1.1709,
+      "step": 7073
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045098500333938e-05,
+      "loss": 0.9712,
+      "step": 7074
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8044482887264275e-05,
+      "loss": 1.3105,
+      "step": 7075
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8043867187783682e-05,
+      "loss": 1.2207,
+      "step": 7076
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8043251401898773e-05,
+      "loss": 1.1201,
+      "step": 7077
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8042635529616154e-05,
+      "loss": 1.1816,
+      "step": 7078
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804201957094245e-05,
+      "loss": 1.04,
+      "step": 7079
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804140352588427e-05,
+      "loss": 1.2568,
+      "step": 7080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8040787394448237e-05,
+      "loss": 0.9629,
+      "step": 7081
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8040171176640967e-05,
+      "loss": 0.8184,
+      "step": 7082
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8039554872469076e-05,
+      "loss": 1.3096,
+      "step": 7083
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803893848193919e-05,
+      "loss": 0.9346,
+      "step": 7084
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803832200505793e-05,
+      "loss": 0.958,
+      "step": 7085
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8037705441831913e-05,
+      "loss": 1.0205,
+      "step": 7086
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8037088792267762e-05,
+      "loss": 1.1836,
+      "step": 7087
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8036472056372102e-05,
+      "loss": 1.2559,
+      "step": 7088
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035855234151563e-05,
+      "loss": 1.2861,
+      "step": 7089
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035238325612764e-05,
+      "loss": 1.124,
+      "step": 7090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8034621330762337e-05,
+      "loss": 1.5781,
+      "step": 7091
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8034004249606903e-05,
+      "loss": 1.3633,
+      "step": 7092
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80333870821531e-05,
+      "loss": 1.1777,
+      "step": 7093
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8032769828407543e-05,
+      "loss": 0.9609,
+      "step": 7094
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803215248837688e-05,
+      "loss": 0.9658,
+      "step": 7095
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8031535062067726e-05,
+      "loss": 1.3262,
+      "step": 7096
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803091754948672e-05,
+      "loss": 1.2061,
+      "step": 7097
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8030299950640496e-05,
+      "loss": 1.2266,
+      "step": 7098
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8029682265535693e-05,
+      "loss": 1.5137,
+      "step": 7099
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8029064494178935e-05,
+      "loss": 1.2373,
+      "step": 7100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8028446636576865e-05,
+      "loss": 1.0508,
+      "step": 7101
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8027828692736115e-05,
+      "loss": 1.2666,
+      "step": 7102
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802721066266333e-05,
+      "loss": 1.2285,
+      "step": 7103
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802659254636514e-05,
+      "loss": 1.0654,
+      "step": 7104
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025974343848192e-05,
+      "loss": 0.9326,
+      "step": 7105
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025356055119122e-05,
+      "loss": 1.5596,
+      "step": 7106
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8024737680184575e-05,
+      "loss": 1.1895,
+      "step": 7107
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8024119219051188e-05,
+      "loss": 0.8921,
+      "step": 7108
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8023500671725612e-05,
+      "loss": 1.3359,
+      "step": 7109
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802288203821448e-05,
+      "loss": 1.0586,
+      "step": 7110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802226331852445e-05,
+      "loss": 1.1641,
+      "step": 7111
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802164451266216e-05,
+      "loss": 1.1523,
+      "step": 7112
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802102562063426e-05,
+      "loss": 1.5293,
+      "step": 7113
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8020406642447393e-05,
+      "loss": 1.1055,
+      "step": 7114
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8019787578108214e-05,
+      "loss": 1.1484,
+      "step": 7115
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801916842762337e-05,
+      "loss": 1.293,
+      "step": 7116
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8018549190999514e-05,
+      "loss": 1.3154,
+      "step": 7117
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8017929868243298e-05,
+      "loss": 1.1748,
+      "step": 7118
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8017310459361365e-05,
+      "loss": 1.4053,
+      "step": 7119
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8016690964360383e-05,
+      "loss": 1.2129,
+      "step": 7120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8016071383246995e-05,
+      "loss": 1.0098,
+      "step": 7121
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8015451716027863e-05,
+      "loss": 1.4277,
+      "step": 7122
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801483196270964e-05,
+      "loss": 0.5249,
+      "step": 7123
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801421212329898e-05,
+      "loss": 1.2461,
+      "step": 7124
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013592197802547e-05,
+      "loss": 1.291,
+      "step": 7125
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8012972186227e-05,
+      "loss": 1.0234,
+      "step": 7126
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8012352088578995e-05,
+      "loss": 1.4688,
+      "step": 7127
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8011731904865196e-05,
+      "loss": 1.0693,
+      "step": 7128
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801111163509226e-05,
+      "loss": 1.2559,
+      "step": 7129
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801049127926686e-05,
+      "loss": 1.1914,
+      "step": 7130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800987083739565e-05,
+      "loss": 0.999,
+      "step": 7131
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8009250309485297e-05,
+      "loss": 0.8501,
+      "step": 7132
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008629695542465e-05,
+      "loss": 1.0547,
+      "step": 7133
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008008995573825e-05,
+      "loss": 1.1943,
+      "step": 7134
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8007388209586038e-05,
+      "loss": 1.4922,
+      "step": 7135
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8006767337585782e-05,
+      "loss": 1.3223,
+      "step": 7136
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8006146379579715e-05,
+      "loss": 1.0068,
+      "step": 7137
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8005525335574515e-05,
+      "loss": 1.2002,
+      "step": 7138
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8004904205576853e-05,
+      "loss": 1.1221,
+      "step": 7139
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8004282989593394e-05,
+      "loss": 0.8076,
+      "step": 7140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8003661687630815e-05,
+      "loss": 0.8838,
+      "step": 7141
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800304029969579e-05,
+      "loss": 0.9414,
+      "step": 7142
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8002418825794997e-05,
+      "loss": 0.9023,
+      "step": 7143
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8001797265935104e-05,
+      "loss": 1.0195,
+      "step": 7144
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8001175620122793e-05,
+      "loss": 1.1328,
+      "step": 7145
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8000553888364743e-05,
+      "loss": 1.1865,
+      "step": 7146
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7999932070667628e-05,
+      "loss": 1.0322,
+      "step": 7147
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7999310167038126e-05,
+      "loss": 1.2285,
+      "step": 7148
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7998688177482925e-05,
+      "loss": 1.3291,
+      "step": 7149
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7998066102008703e-05,
+      "loss": 1.2559,
+      "step": 7150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799744394062214e-05,
+      "loss": 1.2324,
+      "step": 7151
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996821693329913e-05,
+      "loss": 1.6826,
+      "step": 7152
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996199360138717e-05,
+      "loss": 0.9331,
+      "step": 7153
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7995576941055236e-05,
+      "loss": 1.1348,
+      "step": 7154
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7994954436086148e-05,
+      "loss": 1.1523,
+      "step": 7155
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7994331845238145e-05,
+      "loss": 1.4277,
+      "step": 7156
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799370916851791e-05,
+      "loss": 1.2739,
+      "step": 7157
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799308640593214e-05,
+      "loss": 1.5879,
+      "step": 7158
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799246355748752e-05,
+      "loss": 1.6406,
+      "step": 7159
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799184062319074e-05,
+      "loss": 1.3906,
+      "step": 7160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7991217603048493e-05,
+      "loss": 1.1113,
+      "step": 7161
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799059449706747e-05,
+      "loss": 0.9951,
+      "step": 7162
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798997130525436e-05,
+      "loss": 1.208,
+      "step": 7163
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798934802761586e-05,
+      "loss": 1.0381,
+      "step": 7164
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798872466415867e-05,
+      "loss": 1.1328,
+      "step": 7165
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798810121488948e-05,
+      "loss": 1.2383,
+      "step": 7166
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798747767981499e-05,
+      "loss": 1.1787,
+      "step": 7167
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79868540589419e-05,
+      "loss": 1.1484,
+      "step": 7168
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79862303522769e-05,
+      "loss": 1.2773,
+      "step": 7169
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79856065598267e-05,
+      "loss": 1.3926,
+      "step": 7170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798498268159799e-05,
+      "loss": 1.2285,
+      "step": 7171
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798435871759748e-05,
+      "loss": 1.1274,
+      "step": 7172
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7983734667831867e-05,
+      "loss": 1.2979,
+      "step": 7173
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7983110532307862e-05,
+      "loss": 1.3066,
+      "step": 7174
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7982486311032163e-05,
+      "loss": 1.2168,
+      "step": 7175
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798186200401147e-05,
+      "loss": 1.4229,
+      "step": 7176
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79812376112525e-05,
+      "loss": 0.8604,
+      "step": 7177
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7980613132761956e-05,
+      "loss": 1.123,
+      "step": 7178
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7979988568546544e-05,
+      "loss": 1.4316,
+      "step": 7179
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7979363918612972e-05,
+      "loss": 1.0771,
+      "step": 7180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7978739182967954e-05,
+      "loss": 0.9932,
+      "step": 7181
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79781143616182e-05,
+      "loss": 1.1768,
+      "step": 7182
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7977489454570418e-05,
+      "loss": 1.0977,
+      "step": 7183
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7976864461831323e-05,
+      "loss": 0.834,
+      "step": 7184
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7976239383407626e-05,
+      "loss": 1.6514,
+      "step": 7185
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797561421930605e-05,
+      "loss": 1.3984,
+      "step": 7186
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7974988969533295e-05,
+      "loss": 1.0654,
+      "step": 7187
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7974363634096095e-05,
+      "loss": 0.9053,
+      "step": 7188
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797373821300115e-05,
+      "loss": 1.4766,
+      "step": 7189
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7973112706255192e-05,
+      "loss": 1.165,
+      "step": 7190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797248711386493e-05,
+      "loss": 0.8501,
+      "step": 7191
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797186143583709e-05,
+      "loss": 0.9424,
+      "step": 7192
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7971235672178392e-05,
+      "loss": 1.541,
+      "step": 7193
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970609822895552e-05,
+      "loss": 1.416,
+      "step": 7194
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79699838879953e-05,
+      "loss": 1.3301,
+      "step": 7195
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7969357867484358e-05,
+      "loss": 1.4492,
+      "step": 7196
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7968731761369447e-05,
+      "loss": 1.1123,
+      "step": 7197
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7968105569657296e-05,
+      "loss": 1.3047,
+      "step": 7198
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7967479292354627e-05,
+      "loss": 0.8955,
+      "step": 7199
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7966852929468173e-05,
+      "loss": 1.0908,
+      "step": 7200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796622648100466e-05,
+      "loss": 1.4541,
+      "step": 7201
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965599946970814e-05,
+      "loss": 1.1123,
+      "step": 7202
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796497332737337e-05,
+      "loss": 1.2285,
+      "step": 7203
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7964346622219055e-05,
+      "loss": 1.4082,
+      "step": 7204
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79637198315146e-05,
+      "loss": 1.0732,
+      "step": 7205
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796309295526674e-05,
+      "loss": 1.0205,
+      "step": 7206
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7962465993482213e-05,
+      "loss": 1.1562,
+      "step": 7207
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7961838946167747e-05,
+      "loss": 1.2734,
+      "step": 7208
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7961211813330082e-05,
+      "loss": 1.1211,
+      "step": 7209
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796058459497595e-05,
+      "loss": 1.3193,
+      "step": 7210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795995729111209e-05,
+      "loss": 1.082,
+      "step": 7211
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795932990174524e-05,
+      "loss": 1.0869,
+      "step": 7212
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958702426882146e-05,
+      "loss": 1.1416,
+      "step": 7213
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958074866529537e-05,
+      "loss": 1.1797,
+      "step": 7214
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7957447220694154e-05,
+      "loss": 1.292,
+      "step": 7215
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7956819489382754e-05,
+      "loss": 1.1572,
+      "step": 7216
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7956191672602065e-05,
+      "loss": 0.8672,
+      "step": 7217
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7955563770358834e-05,
+      "loss": 1.167,
+      "step": 7218
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795493578265981e-05,
+      "loss": 1.1025,
+      "step": 7219
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7954307709511735e-05,
+      "loss": 1.1094,
+      "step": 7220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795367955092136e-05,
+      "loss": 1.0645,
+      "step": 7221
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7953051306895424e-05,
+      "loss": 1.2744,
+      "step": 7222
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795242297744068e-05,
+      "loss": 1.6387,
+      "step": 7223
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7951794562563884e-05,
+      "loss": 1.0859,
+      "step": 7224
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7951166062271777e-05,
+      "loss": 1.3613,
+      "step": 7225
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795053747657111e-05,
+      "loss": 1.2539,
+      "step": 7226
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794990880546864e-05,
+      "loss": 1.1968,
+      "step": 7227
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7949280048971118e-05,
+      "loss": 1.123,
+      "step": 7228
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79486512070853e-05,
+      "loss": 1.0371,
+      "step": 7229
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7948022279817936e-05,
+      "loss": 1.5107,
+      "step": 7230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7947393267175785e-05,
+      "loss": 1.5156,
+      "step": 7231
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7946764169165606e-05,
+      "loss": 0.8086,
+      "step": 7232
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794613498579415e-05,
+      "loss": 1.1729,
+      "step": 7233
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7945505717068176e-05,
+      "loss": 0.9551,
+      "step": 7234
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794487636299445e-05,
+      "loss": 1.0356,
+      "step": 7235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794424692357973e-05,
+      "loss": 1.2705,
+      "step": 7236
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7943617398830774e-05,
+      "loss": 1.1055,
+      "step": 7237
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7942987788754348e-05,
+      "loss": 1.2871,
+      "step": 7238
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794235809335721e-05,
+      "loss": 0.9961,
+      "step": 7239
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794172831264613e-05,
+      "loss": 1.291,
+      "step": 7240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794109844662787e-05,
+      "loss": 1.0957,
+      "step": 7241
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7940468495309192e-05,
+      "loss": 1.0586,
+      "step": 7242
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7939838458696874e-05,
+      "loss": 1.3223,
+      "step": 7243
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7939208336797665e-05,
+      "loss": 0.8584,
+      "step": 7244
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7938578129618353e-05,
+      "loss": 1.3916,
+      "step": 7245
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7937947837165696e-05,
+      "loss": 1.0107,
+      "step": 7246
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7937317459446468e-05,
+      "loss": 0.8203,
+      "step": 7247
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793668699646744e-05,
+      "loss": 1.2617,
+      "step": 7248
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936056448235384e-05,
+      "loss": 1.249,
+      "step": 7249
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793542581475707e-05,
+      "loss": 1.377,
+      "step": 7250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7934795096039282e-05,
+      "loss": 1.2402,
+      "step": 7251
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793416429208878e-05,
+      "loss": 1.2188,
+      "step": 7252
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7933533402912354e-05,
+      "loss": 1.5469,
+      "step": 7253
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793290242851677e-05,
+      "loss": 1.1211,
+      "step": 7254
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7932271368908814e-05,
+      "loss": 1.2256,
+      "step": 7255
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793164022409526e-05,
+      "loss": 1.1309,
+      "step": 7256
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7931008994082888e-05,
+      "loss": 1.1172,
+      "step": 7257
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793037767887848e-05,
+      "loss": 1.4248,
+      "step": 7258
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7929746278488815e-05,
+      "loss": 1.1016,
+      "step": 7259
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7929114792920677e-05,
+      "loss": 1.3828,
+      "step": 7260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7928483222180845e-05,
+      "loss": 0.915,
+      "step": 7261
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7927851566276112e-05,
+      "loss": 1.3018,
+      "step": 7262
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792721982521326e-05,
+      "loss": 1.1924,
+      "step": 7263
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7926587998999067e-05,
+      "loss": 1.043,
+      "step": 7264
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7925956087640328e-05,
+      "loss": 0.9717,
+      "step": 7265
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792532409114383e-05,
+      "loss": 0.9746,
+      "step": 7266
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7924692009516357e-05,
+      "loss": 1.2207,
+      "step": 7267
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7924059842764702e-05,
+      "loss": 1.2764,
+      "step": 7268
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7923427590895658e-05,
+      "loss": 1.3047,
+      "step": 7269
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7922795253916012e-05,
+      "loss": 0.8813,
+      "step": 7270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792216283183256e-05,
+      "loss": 1.2412,
+      "step": 7271
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7921530324652092e-05,
+      "loss": 1.0352,
+      "step": 7272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79208977323814e-05,
+      "loss": 1.4512,
+      "step": 7273
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7920265055027285e-05,
+      "loss": 1.2998,
+      "step": 7274
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7919632292596542e-05,
+      "loss": 1.1748,
+      "step": 7275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791899944509597e-05,
+      "loss": 0.9805,
+      "step": 7276
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7918366512532353e-05,
+      "loss": 1.2227,
+      "step": 7277
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7917733494912508e-05,
+      "loss": 1.022,
+      "step": 7278
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7917100392243223e-05,
+      "loss": 0.8474,
+      "step": 7279
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79164672045313e-05,
+      "loss": 0.9883,
+      "step": 7280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791583393178355e-05,
+      "loss": 1.292,
+      "step": 7281
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791520057400676e-05,
+      "loss": 0.9585,
+      "step": 7282
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914567131207747e-05,
+      "loss": 1.4238,
+      "step": 7283
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7913933603393307e-05,
+      "loss": 0.8052,
+      "step": 7284
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791329999057025e-05,
+      "loss": 1.5371,
+      "step": 7285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912666292745377e-05,
+      "loss": 1.2969,
+      "step": 7286
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912032509925498e-05,
+      "loss": 1.2251,
+      "step": 7287
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911398642117423e-05,
+      "loss": 1.2764,
+      "step": 7288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7910764689327955e-05,
+      "loss": 0.9883,
+      "step": 7289
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791013065156391e-05,
+      "loss": 1.0742,
+      "step": 7290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7909496528832092e-05,
+      "loss": 0.9985,
+      "step": 7291
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7908862321139318e-05,
+      "loss": 0.9258,
+      "step": 7292
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7908228028492402e-05,
+      "loss": 1.1006,
+      "step": 7293
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7907593650898147e-05,
+      "loss": 1.1592,
+      "step": 7294
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7906959188363378e-05,
+      "loss": 1.1509,
+      "step": 7295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7906324640894907e-05,
+      "loss": 1.2998,
+      "step": 7296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7905690008499545e-05,
+      "loss": 1.3057,
+      "step": 7297
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790505529118412e-05,
+      "loss": 1.0312,
+      "step": 7298
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790442048895544e-05,
+      "loss": 1.3027,
+      "step": 7299
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7903785601820326e-05,
+      "loss": 1.1455,
+      "step": 7300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79031506297856e-05,
+      "loss": 1.1641,
+      "step": 7301
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790251557285808e-05,
+      "loss": 1.0898,
+      "step": 7302
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7901880431044588e-05,
+      "loss": 1.3398,
+      "step": 7303
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790124520435195e-05,
+      "loss": 1.0615,
+      "step": 7304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7900609892786988e-05,
+      "loss": 1.2959,
+      "step": 7305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7899974496356523e-05,
+      "loss": 1.0596,
+      "step": 7306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7899339015067384e-05,
+      "loss": 0.9536,
+      "step": 7307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7898703448926395e-05,
+      "loss": 1.1143,
+      "step": 7308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7898067797940385e-05,
+      "loss": 1.2598,
+      "step": 7309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789743206211618e-05,
+      "loss": 1.1846,
+      "step": 7310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896796241460603e-05,
+      "loss": 1.1719,
+      "step": 7311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896160335980494e-05,
+      "loss": 1.0459,
+      "step": 7312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789552434568268e-05,
+      "loss": 0.9658,
+      "step": 7313
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894888270573997e-05,
+      "loss": 1.2002,
+      "step": 7314
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894252110661264e-05,
+      "loss": 1.0977,
+      "step": 7315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7893615865951332e-05,
+      "loss": 1.248,
+      "step": 7316
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789297953645102e-05,
+      "loss": 1.1309,
+      "step": 7317
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789234312216717e-05,
+      "loss": 1.1396,
+      "step": 7318
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7891706623106618e-05,
+      "loss": 1.1641,
+      "step": 7319
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7891070039276203e-05,
+      "loss": 1.1982,
+      "step": 7320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789043337068276e-05,
+      "loss": 1.2627,
+      "step": 7321
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788979661733313e-05,
+      "loss": 1.5801,
+      "step": 7322
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7889159779234153e-05,
+      "loss": 0.8369,
+      "step": 7323
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788852285639266e-05,
+      "loss": 1.251,
+      "step": 7324
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788788584881551e-05,
+      "loss": 0.7305,
+      "step": 7325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7887248756509528e-05,
+      "loss": 1.0142,
+      "step": 7326
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788661157948157e-05,
+      "loss": 1.502,
+      "step": 7327
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7885974317738477e-05,
+      "loss": 0.9131,
+      "step": 7328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7885336971287092e-05,
+      "loss": 1.3633,
+      "step": 7329
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788469954013426e-05,
+      "loss": 1.2314,
+      "step": 7330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7884062024286833e-05,
+      "loss": 1.2256,
+      "step": 7331
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788342442375166e-05,
+      "loss": 0.8267,
+      "step": 7332
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788278673853558e-05,
+      "loss": 1.0801,
+      "step": 7333
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7882148968645453e-05,
+      "loss": 1.0684,
+      "step": 7334
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7881511114088123e-05,
+      "loss": 1.2412,
+      "step": 7335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880873174870446e-05,
+      "loss": 1.0635,
+      "step": 7336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880235150999274e-05,
+      "loss": 1.1025,
+      "step": 7337
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7879597042481456e-05,
+      "loss": 0.7275,
+      "step": 7338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787895884932385e-05,
+      "loss": 1.2471,
+      "step": 7339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7878320571533315e-05,
+      "loss": 1.375,
+      "step": 7340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7877682209116698e-05,
+      "loss": 1.2891,
+      "step": 7341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7877043762080865e-05,
+      "loss": 1.2842,
+      "step": 7342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7876405230432668e-05,
+      "loss": 0.7407,
+      "step": 7343
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7875766614178968e-05,
+      "loss": 1.125,
+      "step": 7344
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7875127913326623e-05,
+      "loss": 0.9756,
+      "step": 7345
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7874489127882496e-05,
+      "loss": 1.001,
+      "step": 7346
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787385025785345e-05,
+      "loss": 0.9961,
+      "step": 7347
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7873211303246344e-05,
+      "loss": 1.0869,
+      "step": 7348
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7872572264068043e-05,
+      "loss": 0.9355,
+      "step": 7349
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7871933140325412e-05,
+      "loss": 0.8618,
+      "step": 7350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7871293932025314e-05,
+      "loss": 0.917,
+      "step": 7351
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787065463917462e-05,
+      "loss": 1.2461,
+      "step": 7352
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787001526178019e-05,
+      "loss": 0.9502,
+      "step": 7353
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869375799848896e-05,
+      "loss": 1.3037,
+      "step": 7354
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7868736253387608e-05,
+      "loss": 0.9434,
+      "step": 7355
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7868096622403196e-05,
+      "loss": 1.2637,
+      "step": 7356
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7867456906902524e-05,
+      "loss": 0.8359,
+      "step": 7357
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786681710689247e-05,
+      "loss": 1.0996,
+      "step": 7358
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7866177222379907e-05,
+      "loss": 1.1387,
+      "step": 7359
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7865537253371707e-05,
+      "loss": 1.2158,
+      "step": 7360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786489719987474e-05,
+      "loss": 0.9658,
+      "step": 7361
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786425706189589e-05,
+      "loss": 0.9619,
+      "step": 7362
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7863616839442024e-05,
+      "loss": 0.9717,
+      "step": 7363
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7862976532520025e-05,
+      "loss": 1.0942,
+      "step": 7364
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786233614113677e-05,
+      "loss": 1.4219,
+      "step": 7365
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7861695665299133e-05,
+      "loss": 1.3184,
+      "step": 7366
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7861055105014004e-05,
+      "loss": 1.6055,
+      "step": 7367
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860414460288252e-05,
+      "loss": 1.1572,
+      "step": 7368
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859773731128768e-05,
+      "loss": 1.1914,
+      "step": 7369
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859132917542427e-05,
+      "loss": 1.2305,
+      "step": 7370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858492019536117e-05,
+      "loss": 1.1313,
+      "step": 7371
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785785103711672e-05,
+      "loss": 0.8076,
+      "step": 7372
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7857209970291122e-05,
+      "loss": 1.3955,
+      "step": 7373
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7856568819066213e-05,
+      "loss": 1.0332,
+      "step": 7374
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7855927583448874e-05,
+      "loss": 1.002,
+      "step": 7375
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7855286263445997e-05,
+      "loss": 1.0156,
+      "step": 7376
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7854644859064468e-05,
+      "loss": 1.0049,
+      "step": 7377
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785400337031118e-05,
+      "loss": 1.1851,
+      "step": 7378
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7853361797193018e-05,
+      "loss": 1.0098,
+      "step": 7379
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785272013971688e-05,
+      "loss": 1.1406,
+      "step": 7380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7852078397889652e-05,
+      "loss": 1.6621,
+      "step": 7381
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7851436571718235e-05,
+      "loss": 1.4961,
+      "step": 7382
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785079466120952e-05,
+      "loss": 1.1416,
+      "step": 7383
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7850152666370397e-05,
+      "loss": 1.3428,
+      "step": 7384
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784951058720777e-05,
+      "loss": 1.1543,
+      "step": 7385
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7848868423728527e-05,
+      "loss": 1.3008,
+      "step": 7386
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7848226175939576e-05,
+      "loss": 1.1562,
+      "step": 7387
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784758384384781e-05,
+      "loss": 1.0557,
+      "step": 7388
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7846941427460133e-05,
+      "loss": 1.3809,
+      "step": 7389
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7846298926783437e-05,
+      "loss": 1.1035,
+      "step": 7390
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7845656341824632e-05,
+      "loss": 1.0146,
+      "step": 7391
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784501367259062e-05,
+      "loss": 1.2373,
+      "step": 7392
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7844370919088298e-05,
+      "loss": 0.8954,
+      "step": 7393
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7843728081324575e-05,
+      "loss": 0.9385,
+      "step": 7394
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784308515930635e-05,
+      "loss": 0.8374,
+      "step": 7395
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784244215304054e-05,
+      "loss": 0.8872,
+      "step": 7396
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7841799062534043e-05,
+      "loss": 1.0615,
+      "step": 7397
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7841155887793773e-05,
+      "loss": 0.9814,
+      "step": 7398
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7840512628826637e-05,
+      "loss": 1.0146,
+      "step": 7399
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783986928563954e-05,
+      "loss": 1.2949,
+      "step": 7400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783922585823939e-05,
+      "loss": 1.0557,
+      "step": 7401
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7838582346633116e-05,
+      "loss": 1.3418,
+      "step": 7402
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837938750827614e-05,
+      "loss": 0.8506,
+      "step": 7403
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78372950708298e-05,
+      "loss": 0.9883,
+      "step": 7404
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783665130664659e-05,
+      "loss": 0.96,
+      "step": 7405
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7836007458284902e-05,
+      "loss": 0.9727,
+      "step": 7406
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7835363525751647e-05,
+      "loss": 0.9482,
+      "step": 7407
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7834719509053743e-05,
+      "loss": 1.3418,
+      "step": 7408
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783407540819811e-05,
+      "loss": 1.3965,
+      "step": 7409
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7833431223191666e-05,
+      "loss": 0.8062,
+      "step": 7410
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783278695404133e-05,
+      "loss": 1.0259,
+      "step": 7411
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7832142600754025e-05,
+      "loss": 1.0283,
+      "step": 7412
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7831498163336668e-05,
+      "loss": 0.9707,
+      "step": 7413
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7830853641796184e-05,
+      "loss": 1.0566,
+      "step": 7414
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7830209036139493e-05,
+      "loss": 1.0186,
+      "step": 7415
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7829564346373525e-05,
+      "loss": 1.2207,
+      "step": 7416
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78289195725052e-05,
+      "loss": 1.375,
+      "step": 7417
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7828274714541445e-05,
+      "loss": 1.1758,
+      "step": 7418
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7827629772489187e-05,
+      "loss": 1.0645,
+      "step": 7419
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7826984746355357e-05,
+      "loss": 1.3398,
+      "step": 7420
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782633963614688e-05,
+      "loss": 1.2988,
+      "step": 7421
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7825694441870686e-05,
+      "loss": 0.8726,
+      "step": 7422
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7825049163533707e-05,
+      "loss": 0.9775,
+      "step": 7423
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782440380114287e-05,
+      "loss": 0.8535,
+      "step": 7424
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7823758354705115e-05,
+      "loss": 1.021,
+      "step": 7425
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782311282422737e-05,
+      "loss": 1.3374,
+      "step": 7426
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7822467209716566e-05,
+      "loss": 1.3223,
+      "step": 7427
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7821821511179644e-05,
+      "loss": 1.4199,
+      "step": 7428
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7821175728623538e-05,
+      "loss": 1.0615,
+      "step": 7429
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7820529862055184e-05,
+      "loss": 1.168,
+      "step": 7430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7819883911481518e-05,
+      "loss": 0.8057,
+      "step": 7431
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781923787690948e-05,
+      "loss": 1.7266,
+      "step": 7432
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7818591758346017e-05,
+      "loss": 1.0049,
+      "step": 7433
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7817945555798054e-05,
+      "loss": 1.1797,
+      "step": 7434
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7817299269272547e-05,
+      "loss": 1.3828,
+      "step": 7435
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781665289877643e-05,
+      "loss": 1.2617,
+      "step": 7436
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781600644431664e-05,
+      "loss": 1.1758,
+      "step": 7437
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7815359905900138e-05,
+      "loss": 1.1006,
+      "step": 7438
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7814713283533857e-05,
+      "loss": 1.208,
+      "step": 7439
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7814066577224747e-05,
+      "loss": 1.0547,
+      "step": 7440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781341978697975e-05,
+      "loss": 1.2734,
+      "step": 7441
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781277291280582e-05,
+      "loss": 1.1309,
+      "step": 7442
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78121259547099e-05,
+      "loss": 1.2212,
+      "step": 7443
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7811478912698943e-05,
+      "loss": 1.6895,
+      "step": 7444
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78108317867799e-05,
+      "loss": 1.0723,
+      "step": 7445
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7810184576959718e-05,
+      "loss": 1.166,
+      "step": 7446
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780953728324535e-05,
+      "loss": 1.2129,
+      "step": 7447
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7808889905643753e-05,
+      "loss": 1.0674,
+      "step": 7448
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7808242444161873e-05,
+      "loss": 0.9409,
+      "step": 7449
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7807594898806676e-05,
+      "loss": 1.1709,
+      "step": 7450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7806947269585107e-05,
+      "loss": 1.3467,
+      "step": 7451
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780629955650413e-05,
+      "loss": 1.3633,
+      "step": 7452
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7805651759570697e-05,
+      "loss": 1.0078,
+      "step": 7453
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7805003878791775e-05,
+      "loss": 0.8223,
+      "step": 7454
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780435591417431e-05,
+      "loss": 0.8672,
+      "step": 7455
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7803707865725277e-05,
+      "loss": 1.6914,
+      "step": 7456
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7803059733451623e-05,
+      "loss": 0.875,
+      "step": 7457
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7802411517360322e-05,
+      "loss": 1.4238,
+      "step": 7458
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7801763217458326e-05,
+      "loss": 1.0674,
+      "step": 7459
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7801114833752612e-05,
+      "loss": 1.0303,
+      "step": 7460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780046636625013e-05,
+      "loss": 0.6499,
+      "step": 7461
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7799817814957858e-05,
+      "loss": 0.7134,
+      "step": 7462
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7799169179882753e-05,
+      "loss": 0.9189,
+      "step": 7463
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779852046103179e-05,
+      "loss": 1.1426,
+      "step": 7464
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797871658411932e-05,
+      "loss": 1.3818,
+      "step": 7465
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797222772030155e-05,
+      "loss": 1.1914,
+      "step": 7466
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7796573801893417e-05,
+      "loss": 1.1562,
+      "step": 7467
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77959247480087e-05,
+      "loss": 1.2754,
+      "step": 7468
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779527561038297e-05,
+      "loss": 0.9424,
+      "step": 7469
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7794626389023207e-05,
+      "loss": 1.207,
+      "step": 7470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7793977083936377e-05,
+      "loss": 1.1104,
+      "step": 7471
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7793327695129455e-05,
+      "loss": 0.9248,
+      "step": 7472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779267822260942e-05,
+      "loss": 1.1221,
+      "step": 7473
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7792028666383247e-05,
+      "loss": 0.9932,
+      "step": 7474
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791379026457914e-05,
+      "loss": 0.813,
+      "step": 7475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77907293028404e-05,
+      "loss": 0.8984,
+      "step": 7476
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7790079495537683e-05,
+      "loss": 1.1885,
+      "step": 7477
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778942960455674e-05,
+      "loss": 1.3057,
+      "step": 7478
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7788779629904556e-05,
+      "loss": 0.8496,
+      "step": 7479
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7788129571588112e-05,
+      "loss": 0.8335,
+      "step": 7480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778747942961439e-05,
+      "loss": 1.041,
+      "step": 7481
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7786829203990375e-05,
+      "loss": 0.9746,
+      "step": 7482
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7786178894723046e-05,
+      "loss": 1.0488,
+      "step": 7483
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785528501819395e-05,
+      "loss": 1.0635,
+      "step": 7484
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7784878025286406e-05,
+      "loss": 0.999,
+      "step": 7485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778422746513107e-05,
+      "loss": 1.4805,
+      "step": 7486
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7783576821360366e-05,
+      "loss": 1.4082,
+      "step": 7487
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778292609398129e-05,
+      "loss": 0.8398,
+      "step": 7488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778227528300083e-05,
+      "loss": 1.0449,
+      "step": 7489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7781624388425974e-05,
+      "loss": 0.9541,
+      "step": 7490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778097341026372e-05,
+      "loss": 1.2559,
+      "step": 7491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7780322348521054e-05,
+      "loss": 1.3965,
+      "step": 7492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7779671203204975e-05,
+      "loss": 1.3018,
+      "step": 7493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7779019974322476e-05,
+      "loss": 0.7949,
+      "step": 7494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777836866188055e-05,
+      "loss": 1.2109,
+      "step": 7495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7777717265886194e-05,
+      "loss": 1.1724,
+      "step": 7496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7777065786346405e-05,
+      "loss": 0.9736,
+      "step": 7497
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7776414223268184e-05,
+      "loss": 1.2812,
+      "step": 7498
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7775762576658526e-05,
+      "loss": 1.0957,
+      "step": 7499
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7775110846524432e-05,
+      "loss": 1.6855,
+      "step": 7500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77744590328729e-05,
+      "loss": 1.3369,
+      "step": 7501
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7773807135710933e-05,
+      "loss": 1.1914,
+      "step": 7502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777315515504554e-05,
+      "loss": 1.3916,
+      "step": 7503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7772503090883716e-05,
+      "loss": 1.3623,
+      "step": 7504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777185094323247e-05,
+      "loss": 1.167,
+      "step": 7505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77711987120988e-05,
+      "loss": 0.5659,
+      "step": 7506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7770546397489724e-05,
+      "loss": 1.1221,
+      "step": 7507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7769893999412238e-05,
+      "loss": 1.3096,
+      "step": 7508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7769241517873353e-05,
+      "loss": 0.8936,
+      "step": 7509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7768588952880087e-05,
+      "loss": 1.0449,
+      "step": 7510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7767936304439435e-05,
+      "loss": 1.001,
+      "step": 7511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7767283572558414e-05,
+      "loss": 0.9834,
+      "step": 7512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7766630757244035e-05,
+      "loss": 1.1396,
+      "step": 7513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7765977858503314e-05,
+      "loss": 1.1445,
+      "step": 7514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776532487634326e-05,
+      "loss": 1.4062,
+      "step": 7515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7764671810770884e-05,
+      "loss": 1.2383,
+      "step": 7516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776401866179321e-05,
+      "loss": 0.9502,
+      "step": 7517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776336542941724e-05,
+      "loss": 1.2812,
+      "step": 7518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776271211365001e-05,
+      "loss": 1.6543,
+      "step": 7519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7762058714498524e-05,
+      "loss": 1.125,
+      "step": 7520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7761405231969804e-05,
+      "loss": 0.9336,
+      "step": 7521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7760751666070867e-05,
+      "loss": 1.1973,
+      "step": 7522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776009801680874e-05,
+      "loss": 1.2656,
+      "step": 7523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7759444284190438e-05,
+      "loss": 1.2402,
+      "step": 7524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7758790468222984e-05,
+      "loss": 1.3525,
+      "step": 7525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7758136568913403e-05,
+      "loss": 0.6753,
+      "step": 7526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757482586268716e-05,
+      "loss": 1.3594,
+      "step": 7527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7756828520295953e-05,
+      "loss": 1.46,
+      "step": 7528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7756174371002135e-05,
+      "loss": 1.1377,
+      "step": 7529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7755520138394292e-05,
+      "loss": 1.2871,
+      "step": 7530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775486582247945e-05,
+      "loss": 1.0127,
+      "step": 7531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754211423264634e-05,
+      "loss": 1.416,
+      "step": 7532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775355694075688e-05,
+      "loss": 1.0566,
+      "step": 7533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7752902374963213e-05,
+      "loss": 1.2627,
+      "step": 7534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775224772589067e-05,
+      "loss": 1.0723,
+      "step": 7535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7751592993546275e-05,
+      "loss": 1.147,
+      "step": 7536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7750938177937065e-05,
+      "loss": 1.0449,
+      "step": 7537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7750283279070077e-05,
+      "loss": 1.4102,
+      "step": 7538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774962829695234e-05,
+      "loss": 1.2036,
+      "step": 7539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774897323159089e-05,
+      "loss": 1.29,
+      "step": 7540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774831808299277e-05,
+      "loss": 1.0264,
+      "step": 7541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774766285116501e-05,
+      "loss": 1.7188,
+      "step": 7542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774700753611465e-05,
+      "loss": 1.4375,
+      "step": 7543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7746352137848732e-05,
+      "loss": 1.126,
+      "step": 7544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7745696656374295e-05,
+      "loss": 1.3516,
+      "step": 7545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774504109169838e-05,
+      "loss": 1.3008,
+      "step": 7546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7744385443828024e-05,
+      "loss": 1.1895,
+      "step": 7547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7743729712770273e-05,
+      "loss": 0.9023,
+      "step": 7548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7743073898532177e-05,
+      "loss": 1.1309,
+      "step": 7549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774241800112077e-05,
+      "loss": 1.0488,
+      "step": 7550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741762020543107e-05,
+      "loss": 1.1406,
+      "step": 7551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741105956806227e-05,
+      "loss": 0.9736,
+      "step": 7552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774044980991718e-05,
+      "loss": 1.3252,
+      "step": 7553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7739793579883015e-05,
+      "loss": 0.9355,
+      "step": 7554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773913726671078e-05,
+      "loss": 1.4434,
+      "step": 7555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7738480870407524e-05,
+      "loss": 1.3335,
+      "step": 7556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77378243909803e-05,
+      "loss": 1.0479,
+      "step": 7557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7737167828436156e-05,
+      "loss": 0.7837,
+      "step": 7558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7736511182782146e-05,
+      "loss": 1.061,
+      "step": 7559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7735854454025328e-05,
+      "loss": 0.8477,
+      "step": 7560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773519764217275e-05,
+      "loss": 0.937,
+      "step": 7561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773454074723147e-05,
+      "loss": 1.0845,
+      "step": 7562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7733883769208543e-05,
+      "loss": 1.3564,
+      "step": 7563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7733226708111032e-05,
+      "loss": 0.9492,
+      "step": 7564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7732569563945983e-05,
+      "loss": 1.0488,
+      "step": 7565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7731912336720467e-05,
+      "loss": 1.2432,
+      "step": 7566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7731255026441537e-05,
+      "loss": 0.8438,
+      "step": 7567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7730597633116255e-05,
+      "loss": 1.4683,
+      "step": 7568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772994015675168e-05,
+      "loss": 1.2832,
+      "step": 7569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772928259735488e-05,
+      "loss": 0.9844,
+      "step": 7570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7728624954932914e-05,
+      "loss": 1.5059,
+      "step": 7571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7727967229492847e-05,
+      "loss": 1.248,
+      "step": 7572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7727309421041745e-05,
+      "loss": 1.0635,
+      "step": 7573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7726651529586673e-05,
+      "loss": 0.8398,
+      "step": 7574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7725993555134698e-05,
+      "loss": 1.1484,
+      "step": 7575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7725335497692887e-05,
+      "loss": 1.1914,
+      "step": 7576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772467735726831e-05,
+      "loss": 1.4199,
+      "step": 7577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7724019133868035e-05,
+      "loss": 1.25,
+      "step": 7578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7723360827499135e-05,
+      "loss": 1.0801,
+      "step": 7579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772270243816868e-05,
+      "loss": 0.9863,
+      "step": 7580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7722043965883743e-05,
+      "loss": 1.1377,
+      "step": 7581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7721385410651393e-05,
+      "loss": 1.1221,
+      "step": 7582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7720726772478705e-05,
+      "loss": 1.2812,
+      "step": 7583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772006805137276e-05,
+      "loss": 1.0781,
+      "step": 7584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771940924734063e-05,
+      "loss": 0.9639,
+      "step": 7585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771875036038939e-05,
+      "loss": 0.8574,
+      "step": 7586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7718091390526117e-05,
+      "loss": 1.0439,
+      "step": 7587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771743233775789e-05,
+      "loss": 1.0576,
+      "step": 7588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7716773202091794e-05,
+      "loss": 1.4697,
+      "step": 7589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77161139835349e-05,
+      "loss": 1.3125,
+      "step": 7590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7715454682094295e-05,
+      "loss": 1.0479,
+      "step": 7591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771479529777706e-05,
+      "loss": 1.2363,
+      "step": 7592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771413583059028e-05,
+      "loss": 1.3037,
+      "step": 7593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7713476280541032e-05,
+      "loss": 1.0303,
+      "step": 7594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7712816647636405e-05,
+      "loss": 0.8662,
+      "step": 7595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7712156931883483e-05,
+      "loss": 1.0586,
+      "step": 7596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771149713328936e-05,
+      "loss": 1.3418,
+      "step": 7597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771083725186111e-05,
+      "loss": 1.2441,
+      "step": 7598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7710177287605835e-05,
+      "loss": 1.4375,
+      "step": 7599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7709517240530615e-05,
+      "loss": 1.4414,
+      "step": 7600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708857110642542e-05,
+      "loss": 1.2529,
+      "step": 7601
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708196897948708e-05,
+      "loss": 0.9116,
+      "step": 7602
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7707536602456203e-05,
+      "loss": 1.1475,
+      "step": 7603
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7706876224172124e-05,
+      "loss": 0.9385,
+      "step": 7604
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7706215763103557e-05,
+      "loss": 1.3916,
+      "step": 7605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77055552192576e-05,
+      "loss": 0.8931,
+      "step": 7606
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7704894592641354e-05,
+      "loss": 1.0742,
+      "step": 7607
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7704233883261908e-05,
+      "loss": 0.8887,
+      "step": 7608
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770357309112636e-05,
+      "loss": 0.8389,
+      "step": 7609
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770291221624181e-05,
+      "loss": 0.6343,
+      "step": 7610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770225125861536e-05,
+      "loss": 1.248,
+      "step": 7611
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77015902182541e-05,
+      "loss": 0.8838,
+      "step": 7612
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770092909516514e-05,
+      "loss": 1.0771,
+      "step": 7613
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770026788935558e-05,
+      "loss": 1.0381,
+      "step": 7614
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7699606600832517e-05,
+      "loss": 0.9023,
+      "step": 7615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7698945229603057e-05,
+      "loss": 1.2559,
+      "step": 7616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7698283775674307e-05,
+      "loss": 1.2275,
+      "step": 7617
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769762223905337e-05,
+      "loss": 0.9717,
+      "step": 7618
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7696960619747354e-05,
+      "loss": 1.0215,
+      "step": 7619
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7696298917763366e-05,
+      "loss": 1.248,
+      "step": 7620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769563713310851e-05,
+      "loss": 0.8662,
+      "step": 7621
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7694975265789893e-05,
+      "loss": 1.0215,
+      "step": 7622
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769431331581463e-05,
+      "loss": 1.0361,
+      "step": 7623
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7693651283189832e-05,
+      "loss": 1.3047,
+      "step": 7624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7692989167922606e-05,
+      "loss": 1.4238,
+      "step": 7625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7692326970020066e-05,
+      "loss": 1.1436,
+      "step": 7626
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7691664689489326e-05,
+      "loss": 1.041,
+      "step": 7627
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7691002326337498e-05,
+      "loss": 0.9248,
+      "step": 7628
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690339880571697e-05,
+      "loss": 1.166,
+      "step": 7629
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7689677352199042e-05,
+      "loss": 0.9028,
+      "step": 7630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7689014741226648e-05,
+      "loss": 1.041,
+      "step": 7631
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768835204766163e-05,
+      "loss": 1.168,
+      "step": 7632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768768927151111e-05,
+      "loss": 1.4336,
+      "step": 7633
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.76870264127822e-05,
+      "loss": 1.0479,
+      "step": 7634
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7686363471482035e-05,
+      "loss": 0.9326,
+      "step": 7635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768570044761772e-05,
+      "loss": 0.9375,
+      "step": 7636
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7685037341196386e-05,
+      "loss": 0.7168,
+      "step": 7637
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7684374152225155e-05,
+      "loss": 1.5078,
+      "step": 7638
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683710880711145e-05,
+      "loss": 1.084,
+      "step": 7639
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683047526661494e-05,
+      "loss": 1.2578,
+      "step": 7640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768238409008331e-05,
+      "loss": 1.3887,
+      "step": 7641
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7681720570983733e-05,
+      "loss": 1.1143,
+      "step": 7642
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768105696936988e-05,
+      "loss": 1.3125,
+      "step": 7643
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7680393285248887e-05,
+      "loss": 1.5381,
+      "step": 7644
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767972951862788e-05,
+      "loss": 1.1709,
+      "step": 7645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7679065669513994e-05,
+      "loss": 1.0391,
+      "step": 7646
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767840173791435e-05,
+      "loss": 1.0244,
+      "step": 7647
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7677737723836087e-05,
+      "loss": 1.0244,
+      "step": 7648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7677073627286336e-05,
+      "loss": 1.208,
+      "step": 7649
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7676409448272228e-05,
+      "loss": 1.2207,
+      "step": 7650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.76757451868009e-05,
+      "loss": 1.2832,
+      "step": 7651
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767508084287949e-05,
+      "loss": 1.2559,
+      "step": 7652
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7674416416515128e-05,
+      "loss": 1.1641,
+      "step": 7653
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767375190771495e-05,
+      "loss": 1.3711,
+      "step": 7654
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7673087316486105e-05,
+      "loss": 1.2949,
+      "step": 7655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767242264283572e-05,
+      "loss": 0.7812,
+      "step": 7656
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.767175788677094e-05,
+      "loss": 1.2363,
+      "step": 7657
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7671093048298905e-05,
+      "loss": 1.1191,
+      "step": 7658
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.767042812742676e-05,
+      "loss": 0.9385,
+      "step": 7659
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7669763124161637e-05,
+      "loss": 0.7812,
+      "step": 7660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766909803851069e-05,
+      "loss": 1.4688,
+      "step": 7661
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7668432870481058e-05,
+      "loss": 0.9688,
+      "step": 7662
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7667767620079885e-05,
+      "loss": 1.2383,
+      "step": 7663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7667102287314322e-05,
+      "loss": 1.1719,
+      "step": 7664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7666436872191512e-05,
+      "loss": 0.9971,
+      "step": 7665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7665771374718604e-05,
+      "loss": 1.3291,
+      "step": 7666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7665105794902743e-05,
+      "loss": 1.1182,
+      "step": 7667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7664440132751082e-05,
+      "loss": 1.1143,
+      "step": 7668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7663774388270773e-05,
+      "loss": 0.7297,
+      "step": 7669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766310856146896e-05,
+      "loss": 1.752,
+      "step": 7670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7662442652352806e-05,
+      "loss": 1.4004,
+      "step": 7671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7661776660929453e-05,
+      "loss": 1.0801,
+      "step": 7672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7661110587206058e-05,
+      "loss": 1.2324,
+      "step": 7673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766044443118978e-05,
+      "loss": 1.0479,
+      "step": 7674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659778192887772e-05,
+      "loss": 1.2969,
+      "step": 7675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659111872307193e-05,
+      "loss": 0.9531,
+      "step": 7676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7658445469455194e-05,
+      "loss": 1.5049,
+      "step": 7677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7657778984338938e-05,
+      "loss": 1.4375,
+      "step": 7678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7657112416965582e-05,
+      "loss": 0.8232,
+      "step": 7679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765644576734229e-05,
+      "loss": 1.3203,
+      "step": 7680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7655779035476217e-05,
+      "loss": 0.9375,
+      "step": 7681
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765511222137453e-05,
+      "loss": 1.3057,
+      "step": 7682
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765444532504439e-05,
+      "loss": 0.9951,
+      "step": 7683
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7653778346492956e-05,
+      "loss": 1.1816,
+      "step": 7684
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76531112857274e-05,
+      "loss": 1.6875,
+      "step": 7685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765244414275488e-05,
+      "loss": 0.9551,
+      "step": 7686
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7651776917582572e-05,
+      "loss": 0.9858,
+      "step": 7687
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7651109610217635e-05,
+      "loss": 1.0361,
+      "step": 7688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765044222066724e-05,
+      "loss": 0.8511,
+      "step": 7689
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649774748938553e-05,
+      "loss": 1.2305,
+      "step": 7690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649107195038748e-05,
+      "loss": 1.4453,
+      "step": 7691
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764843955897499e-05,
+      "loss": 0.7549,
+      "step": 7692
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7647771840754456e-05,
+      "loss": 1.2529,
+      "step": 7693
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764710404038432e-05,
+      "loss": 1.0811,
+      "step": 7694
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764643615787174e-05,
+      "loss": 1.0195,
+      "step": 7695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7645768193223915e-05,
+      "loss": 0.9424,
+      "step": 7696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7645100146447997e-05,
+      "loss": 0.9805,
+      "step": 7697
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7644432017551177e-05,
+      "loss": 1.3379,
+      "step": 7698
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7643763806540626e-05,
+      "loss": 0.8291,
+      "step": 7699
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764309551342352e-05,
+      "loss": 1.0859,
+      "step": 7700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7642427138207042e-05,
+      "loss": 1.2803,
+      "step": 7701
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7641758680898368e-05,
+      "loss": 0.9131,
+      "step": 7702
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764109014150468e-05,
+      "loss": 1.0928,
+      "step": 7703
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7640421520033156e-05,
+      "loss": 1.417,
+      "step": 7704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763975281649098e-05,
+      "loss": 1.2734,
+      "step": 7705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763908403088534e-05,
+      "loss": 1.2559,
+      "step": 7706
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7638415163223416e-05,
+      "loss": 1.1748,
+      "step": 7707
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7637746213512393e-05,
+      "loss": 0.999,
+      "step": 7708
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763707718175945e-05,
+      "loss": 1.3115,
+      "step": 7709
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7636408067971783e-05,
+      "loss": 1.0781,
+      "step": 7710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635738872156577e-05,
+      "loss": 1.1152,
+      "step": 7711
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635069594321018e-05,
+      "loss": 1.5625,
+      "step": 7712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7634400234472294e-05,
+      "loss": 1.1709,
+      "step": 7713
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76337307926176e-05,
+      "loss": 1.1699,
+      "step": 7714
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7633061268764126e-05,
+      "loss": 1.4824,
+      "step": 7715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7632391662919057e-05,
+      "loss": 1.1543,
+      "step": 7716
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631721975089594e-05,
+      "loss": 1.0029,
+      "step": 7717
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631052205282927e-05,
+      "loss": 1.2129,
+      "step": 7718
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7630382353506247e-05,
+      "loss": 1.4355,
+      "step": 7719
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7629712419766755e-05,
+      "loss": 1.1982,
+      "step": 7720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7629042404071648e-05,
+      "loss": 1.4629,
+      "step": 7721
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7628372306428113e-05,
+      "loss": 1.4316,
+      "step": 7722
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7627702126843363e-05,
+      "loss": 1.2695,
+      "step": 7723
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7627031865324587e-05,
+      "loss": 1.0166,
+      "step": 7724
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7626361521878986e-05,
+      "loss": 1.2021,
+      "step": 7725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7625691096513758e-05,
+      "loss": 1.4375,
+      "step": 7726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762502058923611e-05,
+      "loss": 0.8955,
+      "step": 7727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7624350000053245e-05,
+      "loss": 1.3838,
+      "step": 7728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7623679328972355e-05,
+      "loss": 0.7471,
+      "step": 7729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762300857600066e-05,
+      "loss": 1.377,
+      "step": 7730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7622337741145355e-05,
+      "loss": 1.1665,
+      "step": 7731
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762166682441365e-05,
+      "loss": 1.3662,
+      "step": 7732
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7620995825812745e-05,
+      "loss": 1.0615,
+      "step": 7733
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7620324745349858e-05,
+      "loss": 0.9653,
+      "step": 7734
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7619653583032186e-05,
+      "loss": 1.2598,
+      "step": 7735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761898233886695e-05,
+      "loss": 1.4717,
+      "step": 7736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7618311012861352e-05,
+      "loss": 1.3916,
+      "step": 7737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7617639605022606e-05,
+      "loss": 1.3174,
+      "step": 7738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7616968115357925e-05,
+      "loss": 1.5947,
+      "step": 7739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761629654387452e-05,
+      "loss": 1.165,
+      "step": 7740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7615624890579606e-05,
+      "loss": 1.4023,
+      "step": 7741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76149531554804e-05,
+      "loss": 1.0303,
+      "step": 7742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761428133858411e-05,
+      "loss": 1.3252,
+      "step": 7743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7613609439897957e-05,
+      "loss": 1.0918,
+      "step": 7744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7612937459429163e-05,
+      "loss": 1.4727,
+      "step": 7745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7612265397184938e-05,
+      "loss": 1.3896,
+      "step": 7746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7611593253172507e-05,
+      "loss": 1.1768,
+      "step": 7747
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7610921027399085e-05,
+      "loss": 1.2783,
+      "step": 7748
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7610248719871897e-05,
+      "loss": 1.2588,
+      "step": 7749
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760957633059816e-05,
+      "loss": 1.0527,
+      "step": 7750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7608903859585103e-05,
+      "loss": 1.21,
+      "step": 7751
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760823130683995e-05,
+      "loss": 0.7773,
+      "step": 7752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7607558672369914e-05,
+      "loss": 1.0117,
+      "step": 7753
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606885956182228e-05,
+      "loss": 1.3828,
+      "step": 7754
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606213158284122e-05,
+      "loss": 1.0679,
+      "step": 7755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7605540278682817e-05,
+      "loss": 1.1064,
+      "step": 7756
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760486731738554e-05,
+      "loss": 1.0752,
+      "step": 7757
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7604194274399525e-05,
+      "loss": 1.2109,
+      "step": 7758
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7603521149732002e-05,
+      "loss": 1.207,
+      "step": 7759
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7602847943390194e-05,
+      "loss": 0.8965,
+      "step": 7760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7602174655381336e-05,
+      "loss": 0.8447,
+      "step": 7761
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760150128571266e-05,
+      "loss": 1.3477,
+      "step": 7762
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7600827834391404e-05,
+      "loss": 0.8604,
+      "step": 7763
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7600154301424794e-05,
+      "loss": 1.3789,
+      "step": 7764
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759948068682007e-05,
+      "loss": 0.9541,
+      "step": 7765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7598806990584467e-05,
+      "loss": 0.9312,
+      "step": 7766
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759813321272522e-05,
+      "loss": 1.2402,
+      "step": 7767
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7597459353249568e-05,
+      "loss": 1.1992,
+      "step": 7768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759678541216475e-05,
+      "loss": 0.8716,
+      "step": 7769
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7596111389478003e-05,
+      "loss": 1.1494,
+      "step": 7770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7595437285196568e-05,
+      "loss": 1.3125,
+      "step": 7771
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594763099327686e-05,
+      "loss": 0.9531,
+      "step": 7772
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594088831878598e-05,
+      "loss": 1.124,
+      "step": 7773
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759341448285655e-05,
+      "loss": 1.4648,
+      "step": 7774
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7592740052268785e-05,
+      "loss": 0.8584,
+      "step": 7775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7592065540122544e-05,
+      "loss": 1.6797,
+      "step": 7776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7591390946425072e-05,
+      "loss": 1.0,
+      "step": 7777
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759071627118362e-05,
+      "loss": 1.0312,
+      "step": 7778
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7590041514405435e-05,
+      "loss": 1.2598,
+      "step": 7779
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758936667609776e-05,
+      "loss": 1.333,
+      "step": 7780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7588691756267843e-05,
+      "loss": 0.8379,
+      "step": 7781
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758801675492294e-05,
+      "loss": 0.8838,
+      "step": 7782
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75873416720703e-05,
+      "loss": 0.8662,
+      "step": 7783
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7586666507717172e-05,
+      "loss": 1.0664,
+      "step": 7784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758599126187081e-05,
+      "loss": 1.2402,
+      "step": 7785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7585315934538468e-05,
+      "loss": 1.1318,
+      "step": 7786
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7584640525727398e-05,
+      "loss": 1.0117,
+      "step": 7787
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7583965035444857e-05,
+      "loss": 1.1826,
+      "step": 7788
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75832894636981e-05,
+      "loss": 1.1104,
+      "step": 7789
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758261381049438e-05,
+      "loss": 1.2578,
+      "step": 7790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758193807584096e-05,
+      "loss": 1.3418,
+      "step": 7791
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75812622597451e-05,
+      "loss": 1.2539,
+      "step": 7792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7580586362214054e-05,
+      "loss": 0.8105,
+      "step": 7793
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7579910383255085e-05,
+      "loss": 1.2168,
+      "step": 7794
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7579234322875453e-05,
+      "loss": 1.0977,
+      "step": 7795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757855818108242e-05,
+      "loss": 0.9365,
+      "step": 7796
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757788195788325e-05,
+      "loss": 1.2012,
+      "step": 7797
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7577205653285205e-05,
+      "loss": 0.9395,
+      "step": 7798
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757652926729555e-05,
+      "loss": 1.1172,
+      "step": 7799
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7575852799921555e-05,
+      "loss": 1.4248,
+      "step": 7800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7575176251170482e-05,
+      "loss": 1.1182,
+      "step": 7801
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75744996210496e-05,
+      "loss": 1.3145,
+      "step": 7802
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573822909566172e-05,
+      "loss": 1.6445,
+      "step": 7803
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573146116727475e-05,
+      "loss": 0.8511,
+      "step": 7804
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7572469242540773e-05,
+      "loss": 0.9189,
+      "step": 7805
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7571792287013338e-05,
+      "loss": 1.3994,
+      "step": 7806
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757111525015244e-05,
+      "loss": 1.0068,
+      "step": 7807
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757043813196536e-05,
+      "loss": 1.124,
+      "step": 7808
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756976093245936e-05,
+      "loss": 0.8154,
+      "step": 7809
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756908365164172e-05,
+      "loss": 1.1465,
+      "step": 7810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7568406289519716e-05,
+      "loss": 1.2803,
+      "step": 7811
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756772884610062e-05,
+      "loss": 1.4912,
+      "step": 7812
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7567051321391716e-05,
+      "loss": 1.1494,
+      "step": 7813
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7566373715400274e-05,
+      "loss": 1.0234,
+      "step": 7814
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7565696028133575e-05,
+      "loss": 1.0625,
+      "step": 7815
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75650182595989e-05,
+      "loss": 1.1846,
+      "step": 7816
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7564340409803525e-05,
+      "loss": 1.2178,
+      "step": 7817
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7563662478754737e-05,
+      "loss": 1.6797,
+      "step": 7818
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7562984466459815e-05,
+      "loss": 1.3721,
+      "step": 7819
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7562306372926044e-05,
+      "loss": 1.0269,
+      "step": 7820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561628198160706e-05,
+      "loss": 1.1328,
+      "step": 7821
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756094994217109e-05,
+      "loss": 1.1758,
+      "step": 7822
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7560271604964473e-05,
+      "loss": 1.2432,
+      "step": 7823
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7559593186548146e-05,
+      "loss": 1.29,
+      "step": 7824
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75589146869294e-05,
+      "loss": 1.2227,
+      "step": 7825
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7558236106115514e-05,
+      "loss": 1.4238,
+      "step": 7826
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7557557444113786e-05,
+      "loss": 1.0312,
+      "step": 7827
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7556878700931505e-05,
+      "loss": 0.957,
+      "step": 7828
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7556199876575956e-05,
+      "loss": 1.2109,
+      "step": 7829
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7555520971054437e-05,
+      "loss": 1.0781,
+      "step": 7830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554841984374236e-05,
+      "loss": 1.1035,
+      "step": 7831
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755416291654265e-05,
+      "loss": 1.3965,
+      "step": 7832
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7553483767566973e-05,
+      "loss": 1.1504,
+      "step": 7833
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7552804537454494e-05,
+      "loss": 1.1328,
+      "step": 7834
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7552125226212518e-05,
+      "loss": 1.0854,
+      "step": 7835
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7551445833848336e-05,
+      "loss": 1.2354,
+      "step": 7836
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7550766360369244e-05,
+      "loss": 1.6152,
+      "step": 7837
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755008680578255e-05,
+      "loss": 1.1367,
+      "step": 7838
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7549407170095543e-05,
+      "loss": 0.9526,
+      "step": 7839
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7548727453315534e-05,
+      "loss": 1.3711,
+      "step": 7840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7548047655449813e-05,
+      "loss": 1.3086,
+      "step": 7841
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7547367776505688e-05,
+      "loss": 1.1084,
+      "step": 7842
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754668781649046e-05,
+      "loss": 1.2129,
+      "step": 7843
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754600777541144e-05,
+      "loss": 1.3535,
+      "step": 7844
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7545327653275922e-05,
+      "loss": 0.8545,
+      "step": 7845
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754464745009122e-05,
+      "loss": 1.4199,
+      "step": 7846
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7543967165864636e-05,
+      "loss": 0.8994,
+      "step": 7847
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7543286800603477e-05,
+      "loss": 1.1348,
+      "step": 7848
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7542606354315058e-05,
+      "loss": 1.0332,
+      "step": 7849
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754192582700668e-05,
+      "loss": 1.1035,
+      "step": 7850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7541245218685656e-05,
+      "loss": 1.3623,
+      "step": 7851
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7540564529359293e-05,
+      "loss": 0.8984,
+      "step": 7852
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7539883759034913e-05,
+      "loss": 1.2363,
+      "step": 7853
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7539202907719817e-05,
+      "loss": 1.1895,
+      "step": 7854
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7538521975421325e-05,
+      "loss": 1.1699,
+      "step": 7855
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753784096214675e-05,
+      "loss": 0.8687,
+      "step": 7856
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753715986790341e-05,
+      "loss": 1.0752,
+      "step": 7857
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7536478692698615e-05,
+      "loss": 1.0825,
+      "step": 7858
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7535797436539684e-05,
+      "loss": 1.5449,
+      "step": 7859
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753511609943394e-05,
+      "loss": 1.5664,
+      "step": 7860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7534434681388694e-05,
+      "loss": 1.2959,
+      "step": 7861
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7533753182411272e-05,
+      "loss": 1.2529,
+      "step": 7862
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753307160250899e-05,
+      "loss": 0.9663,
+      "step": 7863
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7532389941689172e-05,
+      "loss": 1.3164,
+      "step": 7864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7531708199959134e-05,
+      "loss": 1.791,
+      "step": 7865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753102637732621e-05,
+      "loss": 0.8359,
+      "step": 7866
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7530344473797716e-05,
+      "loss": 1.3613,
+      "step": 7867
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7529662489380975e-05,
+      "loss": 1.252,
+      "step": 7868
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752898042408332e-05,
+      "loss": 1.3672,
+      "step": 7869
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7528298277912077e-05,
+      "loss": 0.8945,
+      "step": 7870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527616050874566e-05,
+      "loss": 1.334,
+      "step": 7871
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752693374297812e-05,
+      "loss": 1.335,
+      "step": 7872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7526251354230067e-05,
+      "loss": 1.2939,
+      "step": 7873
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752556888463774e-05,
+      "loss": 1.1855,
+      "step": 7874
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7524886334208464e-05,
+      "loss": 1.3203,
+      "step": 7875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7524203702949576e-05,
+      "loss": 1.1758,
+      "step": 7876
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7523520990868405e-05,
+      "loss": 1.125,
+      "step": 7877
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7522838197972286e-05,
+      "loss": 1.1592,
+      "step": 7878
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7522155324268556e-05,
+      "loss": 1.459,
+      "step": 7879
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7521472369764547e-05,
+      "loss": 1.2881,
+      "step": 7880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520789334467595e-05,
+      "loss": 1.0928,
+      "step": 7881
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520106218385038e-05,
+      "loss": 1.1699,
+      "step": 7882
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751942302152421e-05,
+      "loss": 1.375,
+      "step": 7883
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7518739743892458e-05,
+      "loss": 1.6133,
+      "step": 7884
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7518056385497116e-05,
+      "loss": 1.1631,
+      "step": 7885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751737294634552e-05,
+      "loss": 0.9321,
+      "step": 7886
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751668942644502e-05,
+      "loss": 1.3535,
+      "step": 7887
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7516005825802954e-05,
+      "loss": 0.9424,
+      "step": 7888
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7515322144426665e-05,
+      "loss": 0.8809,
+      "step": 7889
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7514638382323497e-05,
+      "loss": 1.0215,
+      "step": 7890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513954539500796e-05,
+      "loss": 1.0,
+      "step": 7891
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75132706159659e-05,
+      "loss": 1.2891,
+      "step": 7892
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751258661172617e-05,
+      "loss": 0.8872,
+      "step": 7893
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7511902526788943e-05,
+      "loss": 1.3145,
+      "step": 7894
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751121836116157e-05,
+      "loss": 1.5381,
+      "step": 7895
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7510534114851397e-05,
+      "loss": 1.2588,
+      "step": 7896
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7509849787865775e-05,
+      "loss": 0.9229,
+      "step": 7897
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750916538021206e-05,
+      "loss": 0.9521,
+      "step": 7898
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7508480891897598e-05,
+      "loss": 1.1543,
+      "step": 7899
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7507796322929743e-05,
+      "loss": 1.1934,
+      "step": 7900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7507111673315852e-05,
+      "loss": 1.0537,
+      "step": 7901
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750642694306327e-05,
+      "loss": 0.8701,
+      "step": 7902
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750574213217936e-05,
+      "loss": 0.9922,
+      "step": 7903
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750505724067148e-05,
+      "loss": 1.2178,
+      "step": 7904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7504372268546983e-05,
+      "loss": 1.0117,
+      "step": 7905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750368721581322e-05,
+      "loss": 0.9854,
+      "step": 7906
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7503002082477567e-05,
+      "loss": 1.1895,
+      "step": 7907
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7502316868547368e-05,
+      "loss": 1.3271,
+      "step": 7908
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7501631574029983e-05,
+      "loss": 1.1729,
+      "step": 7909
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7500946198932785e-05,
+      "loss": 1.6035,
+      "step": 7910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7500260743263124e-05,
+      "loss": 1.5391,
+      "step": 7911
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7499575207028374e-05,
+      "loss": 1.4365,
+      "step": 7912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7498889590235888e-05,
+      "loss": 1.4141,
+      "step": 7913
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749820389289304e-05,
+      "loss": 1.2803,
+      "step": 7914
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7497518115007185e-05,
+      "loss": 1.2319,
+      "step": 7915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74968322565857e-05,
+      "loss": 1.3186,
+      "step": 7916
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749614631763595e-05,
+      "loss": 1.3184,
+      "step": 7917
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7495460298165294e-05,
+      "loss": 0.9551,
+      "step": 7918
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749477419818111e-05,
+      "loss": 0.9619,
+      "step": 7919
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749408801769077e-05,
+      "loss": 1.1738,
+      "step": 7920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7493401756701637e-05,
+      "loss": 1.1338,
+      "step": 7921
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492715415221087e-05,
+      "loss": 0.6777,
+      "step": 7922
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492028993256493e-05,
+      "loss": 1.2617,
+      "step": 7923
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7491342490815225e-05,
+      "loss": 1.5488,
+      "step": 7924
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749065590790466e-05,
+      "loss": 1.4668,
+      "step": 7925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489969244532173e-05,
+      "loss": 1.3926,
+      "step": 7926
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489282500705135e-05,
+      "loss": 1.3564,
+      "step": 7927
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7488595676430927e-05,
+      "loss": 0.9004,
+      "step": 7928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7487908771716927e-05,
+      "loss": 0.8994,
+      "step": 7929
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7487221786570517e-05,
+      "loss": 1.2002,
+      "step": 7930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748653472099907e-05,
+      "loss": 1.0435,
+      "step": 7931
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7485847575009963e-05,
+      "loss": 1.2598,
+      "step": 7932
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7485160348610586e-05,
+      "loss": 1.002,
+      "step": 7933
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7484473041808318e-05,
+      "loss": 1.1543,
+      "step": 7934
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748378565461054e-05,
+      "loss": 0.9441,
+      "step": 7935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7483098187024637e-05,
+      "loss": 1.1641,
+      "step": 7936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482410639057994e-05,
+      "loss": 1.4248,
+      "step": 7937
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7481723010717994e-05,
+      "loss": 0.8115,
+      "step": 7938
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7481035302012027e-05,
+      "loss": 1.4023,
+      "step": 7939
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7480347512947474e-05,
+      "loss": 1.2871,
+      "step": 7940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7479659643531732e-05,
+      "loss": 0.9551,
+      "step": 7941
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7478971693772182e-05,
+      "loss": 0.9707,
+      "step": 7942
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7478283663676215e-05,
+      "loss": 1.4902,
+      "step": 7943
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7477595553251228e-05,
+      "loss": 1.3564,
+      "step": 7944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7476907362504604e-05,
+      "loss": 1.2949,
+      "step": 7945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7476219091443738e-05,
+      "loss": 0.999,
+      "step": 7946
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475530740076028e-05,
+      "loss": 1.5625,
+      "step": 7947
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7474842308408863e-05,
+      "loss": 1.0225,
+      "step": 7948
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7474153796449635e-05,
+      "loss": 1.3574,
+      "step": 7949
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7473465204205748e-05,
+      "loss": 1.2471,
+      "step": 7950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747277653168459e-05,
+      "loss": 1.4951,
+      "step": 7951
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7472087778893568e-05,
+      "loss": 1.252,
+      "step": 7952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747139894584007e-05,
+      "loss": 1.4385,
+      "step": 7953
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7470710032531502e-05,
+      "loss": 1.1216,
+      "step": 7954
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7470021038975263e-05,
+      "loss": 1.1436,
+      "step": 7955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7469331965178753e-05,
+      "loss": 1.1953,
+      "step": 7956
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7468642811149378e-05,
+      "loss": 1.1802,
+      "step": 7957
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746795357689453e-05,
+      "loss": 1.2344,
+      "step": 7958
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746726426242162e-05,
+      "loss": 1.4199,
+      "step": 7959
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7466574867738056e-05,
+      "loss": 1.1074,
+      "step": 7960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7465885392851235e-05,
+      "loss": 1.0537,
+      "step": 7961
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7465195837768568e-05,
+      "loss": 1.3906,
+      "step": 7962
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746450620249746e-05,
+      "loss": 1.3262,
+      "step": 7963
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746381648704532e-05,
+      "loss": 0.8438,
+      "step": 7964
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7463126691419558e-05,
+      "loss": 1.1768,
+      "step": 7965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746243681562758e-05,
+      "loss": 1.0166,
+      "step": 7966
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74617468596768e-05,
+      "loss": 0.9854,
+      "step": 7967
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7461056823574628e-05,
+      "loss": 1.2002,
+      "step": 7968
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7460366707328473e-05,
+      "loss": 0.9678,
+      "step": 7969
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745967651094575e-05,
+      "loss": 1.0117,
+      "step": 7970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7458986234433876e-05,
+      "loss": 1.3789,
+      "step": 7971
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7458295877800265e-05,
+      "loss": 1.3994,
+      "step": 7972
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7457605441052326e-05,
+      "loss": 1.7695,
+      "step": 7973
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745691492419748e-05,
+      "loss": 1.2578,
+      "step": 7974
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745622432724315e-05,
+      "loss": 1.0391,
+      "step": 7975
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7455533650196744e-05,
+      "loss": 0.958,
+      "step": 7976
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454842893065687e-05,
+      "loss": 0.9248,
+      "step": 7977
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454152055857396e-05,
+      "loss": 1.3213,
+      "step": 7978
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7453461138579294e-05,
+      "loss": 1.0752,
+      "step": 7979
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74527701412388e-05,
+      "loss": 1.1494,
+      "step": 7980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745207906384334e-05,
+      "loss": 1.2363,
+      "step": 7981
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745138790640034e-05,
+      "loss": 1.1318,
+      "step": 7982
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7450696668917214e-05,
+      "loss": 1.5215,
+      "step": 7983
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7450005351401394e-05,
+      "loss": 1.1895,
+      "step": 7984
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7449313953860303e-05,
+      "loss": 0.916,
+      "step": 7985
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744862247630137e-05,
+      "loss": 0.7783,
+      "step": 7986
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744793091873202e-05,
+      "loss": 1.1855,
+      "step": 7987
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744723928115969e-05,
+      "loss": 1.0957,
+      "step": 7988
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7446547563591794e-05,
+      "loss": 1.416,
+      "step": 7989
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7445855766035776e-05,
+      "loss": 0.7368,
+      "step": 7990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7445163888499065e-05,
+      "loss": 1.2754,
+      "step": 7991
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7444471930989084e-05,
+      "loss": 1.3926,
+      "step": 7992
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443779893513274e-05,
+      "loss": 0.6694,
+      "step": 7993
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443087776079068e-05,
+      "loss": 0.9297,
+      "step": 7994
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7442395578693893e-05,
+      "loss": 0.6616,
+      "step": 7995
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7441703301365196e-05,
+      "loss": 0.8896,
+      "step": 7996
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7441010944100404e-05,
+      "loss": 1.1846,
+      "step": 7997
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7440318506906958e-05,
+      "loss": 1.124,
+      "step": 7998
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7439625989792293e-05,
+      "loss": 1.1133,
+      "step": 7999
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7438933392763852e-05,
+      "loss": 1.252,
+      "step": 8000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7438240715829073e-05,
+      "loss": 1.0425,
+      "step": 8001
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7437547958995394e-05,
+      "loss": 0.9204,
+      "step": 8002
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743685512227026e-05,
+      "loss": 0.6924,
+      "step": 8003
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743616220566111e-05,
+      "loss": 1.1548,
+      "step": 8004
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743546920917539e-05,
+      "loss": 1.1973,
+      "step": 8005
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7434776132820545e-05,
+      "loss": 0.8042,
+      "step": 8006
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7434082976604015e-05,
+      "loss": 0.9834,
+      "step": 8007
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7433389740533243e-05,
+      "loss": 1.1396,
+      "step": 8008
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432696424615688e-05,
+      "loss": 1.043,
+      "step": 8009
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432003028858785e-05,
+      "loss": 1.1924,
+      "step": 8010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743130955326999e-05,
+      "loss": 0.791,
+      "step": 8011
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743061599785675e-05,
+      "loss": 1.1328,
+      "step": 8012
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742992236262651e-05,
+      "loss": 1.3945,
+      "step": 8013
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7429228647586726e-05,
+      "loss": 1.2783,
+      "step": 8014
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742853485274485e-05,
+      "loss": 0.8955,
+      "step": 8015
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7427840978108333e-05,
+      "loss": 1.4023,
+      "step": 8016
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7427147023684627e-05,
+      "loss": 1.043,
+      "step": 8017
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7426452989481187e-05,
+      "loss": 1.3242,
+      "step": 8018
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742575887550547e-05,
+      "loss": 1.4805,
+      "step": 8019
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742506468176493e-05,
+      "loss": 0.6396,
+      "step": 8020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742437040826702e-05,
+      "loss": 0.8867,
+      "step": 8021
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7423676055019206e-05,
+      "loss": 1.1699,
+      "step": 8022
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7422981622028942e-05,
+      "loss": 0.8164,
+      "step": 8023
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7422287109303688e-05,
+      "loss": 0.9277,
+      "step": 8024
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7421592516850906e-05,
+      "loss": 1.2861,
+      "step": 8025
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742089784467805e-05,
+      "loss": 0.9746,
+      "step": 8026
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742020309279259e-05,
+      "loss": 0.9146,
+      "step": 8027
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7419508261201986e-05,
+      "loss": 0.9238,
+      "step": 8028
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74188133499137e-05,
+      "loss": 0.9463,
+      "step": 8029
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7418118358935196e-05,
+      "loss": 1.6973,
+      "step": 8030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741742328827394e-05,
+      "loss": 0.8618,
+      "step": 8031
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7416728137937407e-05,
+      "loss": 1.1533,
+      "step": 8032
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7416032907933052e-05,
+      "loss": 1.1201,
+      "step": 8033
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7415337598268346e-05,
+      "loss": 1.416,
+      "step": 8034
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414642208950762e-05,
+      "loss": 0.8711,
+      "step": 8035
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7413946739987765e-05,
+      "loss": 0.9912,
+      "step": 8036
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7413251191386828e-05,
+      "loss": 1.2637,
+      "step": 8037
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741255556315542e-05,
+      "loss": 1.6172,
+      "step": 8038
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7411859855301015e-05,
+      "loss": 1.2559,
+      "step": 8039
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7411164067831085e-05,
+      "loss": 1.2607,
+      "step": 8040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7410468200753105e-05,
+      "loss": 0.9814,
+      "step": 8041
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740977225407455e-05,
+      "loss": 0.9766,
+      "step": 8042
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740907622780289e-05,
+      "loss": 0.9434,
+      "step": 8043
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7408380121945615e-05,
+      "loss": 1.3564,
+      "step": 8044
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7407683936510187e-05,
+      "loss": 0.8301,
+      "step": 8045
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7406987671504097e-05,
+      "loss": 1.3125,
+      "step": 8046
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7406291326934814e-05,
+      "loss": 0.9414,
+      "step": 8047
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740559490280982e-05,
+      "loss": 1.2275,
+      "step": 8048
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7404898399136603e-05,
+      "loss": 1.1514,
+      "step": 8049
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7404201815922635e-05,
+      "loss": 1.2363,
+      "step": 8050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7403505153175403e-05,
+      "loss": 1.2344,
+      "step": 8051
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7402808410902392e-05,
+      "loss": 1.0967,
+      "step": 8052
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740211158911108e-05,
+      "loss": 0.9707,
+      "step": 8053
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740141468780896e-05,
+      "loss": 1.1953,
+      "step": 8054
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740071770700351e-05,
+      "loss": 0.7388,
+      "step": 8055
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7400020646702226e-05,
+      "loss": 1.1104,
+      "step": 8056
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7399323506912587e-05,
+      "loss": 1.165,
+      "step": 8057
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7398626287642083e-05,
+      "loss": 1.458,
+      "step": 8058
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739792898889821e-05,
+      "loss": 1.1475,
+      "step": 8059
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739723161068845e-05,
+      "loss": 1.0977,
+      "step": 8060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7396534153020298e-05,
+      "loss": 1.1719,
+      "step": 8061
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7395836615901243e-05,
+      "loss": 1.1289,
+      "step": 8062
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7395138999338778e-05,
+      "loss": 1.0566,
+      "step": 8063
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7394441303340404e-05,
+      "loss": 1.209,
+      "step": 8064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7393743527913603e-05,
+      "loss": 1.1797,
+      "step": 8065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739304567306588e-05,
+      "loss": 1.0413,
+      "step": 8066
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739234773880473e-05,
+      "loss": 1.1318,
+      "step": 8067
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7391649725137644e-05,
+      "loss": 1.292,
+      "step": 8068
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7390951632072126e-05,
+      "loss": 0.7837,
+      "step": 8069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739025345961567e-05,
+      "loss": 1.2822,
+      "step": 8070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738955520777578e-05,
+      "loss": 0.917,
+      "step": 8071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7388856876559955e-05,
+      "loss": 0.9619,
+      "step": 8072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7388158465975693e-05,
+      "loss": 1.0977,
+      "step": 8073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73874599760305e-05,
+      "loss": 1.4707,
+      "step": 8074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7386761406731873e-05,
+      "loss": 1.0759,
+      "step": 8075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7386062758087324e-05,
+      "loss": 1.0947,
+      "step": 8076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7385364030104353e-05,
+      "loss": 1.1807,
+      "step": 8077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738466522279047e-05,
+      "loss": 1.3096,
+      "step": 8078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7383966336153172e-05,
+      "loss": 0.9224,
+      "step": 8079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7383267370199974e-05,
+      "loss": 1.2637,
+      "step": 8080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382568324938384e-05,
+      "loss": 1.0518,
+      "step": 8081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738186920037591e-05,
+      "loss": 1.1826,
+      "step": 8082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7381169996520057e-05,
+      "loss": 0.709,
+      "step": 8083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7380470713378336e-05,
+      "loss": 1.1699,
+      "step": 8084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7379771350958266e-05,
+      "loss": 1.3291,
+      "step": 8085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7379071909267353e-05,
+      "loss": 0.9307,
+      "step": 8086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378372388313113e-05,
+      "loss": 1.4434,
+      "step": 8087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737767278810306e-05,
+      "loss": 1.5889,
+      "step": 8088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376973108644707e-05,
+      "loss": 1.1553,
+      "step": 8089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376273349945568e-05,
+      "loss": 1.2158,
+      "step": 8090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7375573512013164e-05,
+      "loss": 1.0083,
+      "step": 8091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7374873594855015e-05,
+      "loss": 1.2739,
+      "step": 8092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7374173598478627e-05,
+      "loss": 0.8643,
+      "step": 8093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737347352289153e-05,
+      "loss": 1.6523,
+      "step": 8094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737277336810124e-05,
+      "loss": 1.1895,
+      "step": 8095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7372073134115282e-05,
+      "loss": 1.2383,
+      "step": 8096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7371372820941173e-05,
+      "loss": 1.0056,
+      "step": 8097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7370672428586436e-05,
+      "loss": 1.0869,
+      "step": 8098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7369971957058592e-05,
+      "loss": 1.3652,
+      "step": 8099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7369271406365175e-05,
+      "loss": 1.0371,
+      "step": 8100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73685707765137e-05,
+      "loss": 1.4697,
+      "step": 8101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7367870067511695e-05,
+      "loss": 1.2988,
+      "step": 8102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736716927936669e-05,
+      "loss": 1.2793,
+      "step": 8103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736646841208621e-05,
+      "loss": 1.3887,
+      "step": 8104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7365767465677785e-05,
+      "loss": 1.1016,
+      "step": 8105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7365066440148942e-05,
+      "loss": 0.8765,
+      "step": 8106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7364365335507213e-05,
+      "loss": 0.7378,
+      "step": 8107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736366415176013e-05,
+      "loss": 1.5186,
+      "step": 8108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7362962888915222e-05,
+      "loss": 1.1016,
+      "step": 8109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7362261546980025e-05,
+      "loss": 1.0596,
+      "step": 8110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736156012596207e-05,
+      "loss": 1.2119,
+      "step": 8111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7360858625868894e-05,
+      "loss": 1.2275,
+      "step": 8112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736015704670803e-05,
+      "loss": 1.1777,
+      "step": 8113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7359455388487008e-05,
+      "loss": 1.3008,
+      "step": 8114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735875365121338e-05,
+      "loss": 1.2148,
+      "step": 8115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7358051834894672e-05,
+      "loss": 0.9668,
+      "step": 8116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357349939538426e-05,
+      "loss": 1.0352,
+      "step": 8117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7356647965152184e-05,
+      "loss": 1.5645,
+      "step": 8118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735594591174348e-05,
+      "loss": 1.082,
+      "step": 8119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735524377931986e-05,
+      "loss": 0.8862,
+      "step": 8120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7354541567888866e-05,
+      "loss": 0.9873,
+      "step": 8121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735383927745804e-05,
+      "loss": 1.1318,
+      "step": 8122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7353136908034925e-05,
+      "loss": 1.207,
+      "step": 8123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7352434459627067e-05,
+      "loss": 1.2246,
+      "step": 8124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7351731932242012e-05,
+      "loss": 1.2539,
+      "step": 8125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7351029325887303e-05,
+      "loss": 1.2549,
+      "step": 8126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7350326640570493e-05,
+      "loss": 1.2451,
+      "step": 8127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7349623876299122e-05,
+      "loss": 1.2305,
+      "step": 8128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7348921033080746e-05,
+      "loss": 1.4189,
+      "step": 8129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734821811092291e-05,
+      "loss": 1.0664,
+      "step": 8130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7347515109833167e-05,
+      "loss": 1.0615,
+      "step": 8131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7346812029819067e-05,
+      "loss": 1.187,
+      "step": 8132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734610887088816e-05,
+      "loss": 0.9971,
+      "step": 8133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7345405633048012e-05,
+      "loss": 1.0537,
+      "step": 8134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734470231630616e-05,
+      "loss": 0.8916,
+      "step": 8135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7343998920670166e-05,
+      "loss": 0.9727,
+      "step": 8136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7343295446147585e-05,
+      "loss": 1.0527,
+      "step": 8137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7342591892745978e-05,
+      "loss": 0.5967,
+      "step": 8138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7341888260472895e-05,
+      "loss": 1.2822,
+      "step": 8139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7341184549335895e-05,
+      "loss": 0.8486,
+      "step": 8140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7340480759342543e-05,
+      "loss": 1.3193,
+      "step": 8141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7339776890500392e-05,
+      "loss": 1.252,
+      "step": 8142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7339072942817007e-05,
+      "loss": 1.208,
+      "step": 8143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733836891629995e-05,
+      "loss": 1.3203,
+      "step": 8144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7337664810956784e-05,
+      "loss": 1.375,
+      "step": 8145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7336960626795064e-05,
+      "loss": 0.8765,
+      "step": 8146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7336256363822364e-05,
+      "loss": 1.332,
+      "step": 8147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7335552022046244e-05,
+      "loss": 0.9326,
+      "step": 8148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733484760147427e-05,
+      "loss": 1.0117,
+      "step": 8149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7334143102114015e-05,
+      "loss": 1.416,
+      "step": 8150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7333438523973033e-05,
+      "loss": 1.1484,
+      "step": 8151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7332733867058905e-05,
+      "loss": 1.0176,
+      "step": 8152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7332029131379196e-05,
+      "loss": 0.9331,
+      "step": 8153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7331324316941474e-05,
+      "loss": 0.8271,
+      "step": 8154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733061942375331e-05,
+      "loss": 1.1777,
+      "step": 8155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7329914451822278e-05,
+      "loss": 1.1299,
+      "step": 8156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7329209401155954e-05,
+      "loss": 1.1582,
+      "step": 8157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73285042717619e-05,
+      "loss": 1.3047,
+      "step": 8158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7327799063647703e-05,
+      "loss": 0.6802,
+      "step": 8159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732709377682093e-05,
+      "loss": 1.1406,
+      "step": 8160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7326388411289158e-05,
+      "loss": 0.9072,
+      "step": 8161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732568296705997e-05,
+      "loss": 1.3018,
+      "step": 8162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7324977444140934e-05,
+      "loss": 1.5107,
+      "step": 8163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7324271842539634e-05,
+      "loss": 1.0293,
+      "step": 8164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732356616226365e-05,
+      "loss": 1.25,
+      "step": 8165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732286040332056e-05,
+      "loss": 1.1426,
+      "step": 8166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7322154565717943e-05,
+      "loss": 0.7822,
+      "step": 8167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7321448649463388e-05,
+      "loss": 1.3379,
+      "step": 8168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732074265456447e-05,
+      "loss": 0.9512,
+      "step": 8169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7320036581028778e-05,
+      "loss": 1.0537,
+      "step": 8170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731933042886389e-05,
+      "loss": 1.0547,
+      "step": 8171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73186241980774e-05,
+      "loss": 1.4941,
+      "step": 8172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731791788867689e-05,
+      "loss": 1.4434,
+      "step": 8173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7317211500669944e-05,
+      "loss": 1.4121,
+      "step": 8174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731650503406415e-05,
+      "loss": 1.1016,
+      "step": 8175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7315798488867104e-05,
+      "loss": 1.5859,
+      "step": 8176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7315091865086386e-05,
+      "loss": 1.0684,
+      "step": 8177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7314385162729595e-05,
+      "loss": 1.1514,
+      "step": 8178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7313678381804312e-05,
+      "loss": 1.1895,
+      "step": 8179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731297152231814e-05,
+      "loss": 1.3105,
+      "step": 8180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7312264584278663e-05,
+      "loss": 1.0664,
+      "step": 8181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7311557567693476e-05,
+      "loss": 0.7505,
+      "step": 8182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731085047257018e-05,
+      "loss": 1.2236,
+      "step": 8183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7310143298916365e-05,
+      "loss": 1.4863,
+      "step": 8184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730943604673963e-05,
+      "loss": 1.0283,
+      "step": 8185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308728716047568e-05,
+      "loss": 1.0723,
+      "step": 8186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730802130684778e-05,
+      "loss": 1.1562,
+      "step": 8187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7307313819147862e-05,
+      "loss": 0.9395,
+      "step": 8188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7306606252955417e-05,
+      "loss": 1.4209,
+      "step": 8189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305898608278043e-05,
+      "loss": 1.0991,
+      "step": 8190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305190885123346e-05,
+      "loss": 1.3799,
+      "step": 8191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730448308349892e-05,
+      "loss": 1.5684,
+      "step": 8192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7303775203412374e-05,
+      "loss": 0.7588,
+      "step": 8193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730306724487131e-05,
+      "loss": 1.252,
+      "step": 8194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7302359207883332e-05,
+      "loss": 0.7524,
+      "step": 8195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7301651092456045e-05,
+      "loss": 0.9873,
+      "step": 8196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730094289859706e-05,
+      "loss": 1.1504,
+      "step": 8197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730023462631398e-05,
+      "loss": 1.2598,
+      "step": 8198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729952627561441e-05,
+      "loss": 1.0449,
+      "step": 8199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7298817846505965e-05,
+      "loss": 0.9336,
+      "step": 8200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729810933899625e-05,
+      "loss": 0.9619,
+      "step": 8201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729740075309288e-05,
+      "loss": 1.187,
+      "step": 8202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7296692088803464e-05,
+      "loss": 1.1123,
+      "step": 8203
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7295983346135615e-05,
+      "loss": 1.0396,
+      "step": 8204
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7295274525096943e-05,
+      "loss": 1.2188,
+      "step": 8205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7294565625695066e-05,
+      "loss": 1.248,
+      "step": 8206
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7293856647937596e-05,
+      "loss": 1.0376,
+      "step": 8207
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729314759183215e-05,
+      "loss": 1.2046,
+      "step": 8208
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7292438457386344e-05,
+      "loss": 0.9316,
+      "step": 8209
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72917292446078e-05,
+      "loss": 0.8223,
+      "step": 8210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729101995350412e-05,
+      "loss": 0.873,
+      "step": 8211
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7290310584082943e-05,
+      "loss": 1.3164,
+      "step": 8212
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7289601136351878e-05,
+      "loss": 1.2256,
+      "step": 8213
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7288891610318547e-05,
+      "loss": 1.3691,
+      "step": 8214
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7288182005990575e-05,
+      "loss": 1.0186,
+      "step": 8215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728747232337558e-05,
+      "loss": 1.2646,
+      "step": 8216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7286762562481185e-05,
+      "loss": 1.1162,
+      "step": 8217
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728605272331502e-05,
+      "loss": 0.8545,
+      "step": 8218
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72853428058847e-05,
+      "loss": 1.3887,
+      "step": 8219
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7284632810197857e-05,
+      "loss": 1.0205,
+      "step": 8220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728392273626212e-05,
+      "loss": 1.4434,
+      "step": 8221
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728321258408511e-05,
+      "loss": 0.9502,
+      "step": 8222
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728250235367446e-05,
+      "loss": 1.1143,
+      "step": 8223
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7281792045037798e-05,
+      "loss": 0.874,
+      "step": 8224
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728108165818275e-05,
+      "loss": 1.2344,
+      "step": 8225
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7280371193116955e-05,
+      "loss": 0.8608,
+      "step": 8226
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7279660649848038e-05,
+      "loss": 0.9424,
+      "step": 8227
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7278950028383632e-05,
+      "loss": 0.9824,
+      "step": 8228
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727823932873137e-05,
+      "loss": 1.1279,
+      "step": 8229
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7277528550898884e-05,
+      "loss": 1.2314,
+      "step": 8230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7276817694893816e-05,
+      "loss": 1.0732,
+      "step": 8231
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72761067607238e-05,
+      "loss": 1.1191,
+      "step": 8232
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727539574839647e-05,
+      "loss": 1.1538,
+      "step": 8233
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727468465791946e-05,
+      "loss": 1.2295,
+      "step": 8234
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7273973489300414e-05,
+      "loss": 1.2012,
+      "step": 8235
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727326224254697e-05,
+      "loss": 0.8237,
+      "step": 8236
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7272550917666765e-05,
+      "loss": 1.4395,
+      "step": 8237
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271839514667447e-05,
+      "loss": 1.5,
+      "step": 8238
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271128033556648e-05,
+      "loss": 1.2891,
+      "step": 8239
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727041647434202e-05,
+      "loss": 1.2832,
+      "step": 8240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7269704837031195e-05,
+      "loss": 1.0742,
+      "step": 8241
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268993121631825e-05,
+      "loss": 1.0596,
+      "step": 8242
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268281328151558e-05,
+      "loss": 0.8862,
+      "step": 8243
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726756945659803e-05,
+      "loss": 1.2168,
+      "step": 8244
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7266857506978897e-05,
+      "loss": 0.6689,
+      "step": 8245
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72661454793018e-05,
+      "loss": 1.209,
+      "step": 8246
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726543337357439e-05,
+      "loss": 1.7031,
+      "step": 8247
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726472118980432e-05,
+      "loss": 1.1094,
+      "step": 8248
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7264008927999236e-05,
+      "loss": 0.9033,
+      "step": 8249
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7263296588166784e-05,
+      "loss": 0.9609,
+      "step": 8250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7262584170314625e-05,
+      "loss": 1.2295,
+      "step": 8251
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7261871674450407e-05,
+      "loss": 0.8086,
+      "step": 8252
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726115910058178e-05,
+      "loss": 1.3984,
+      "step": 8253
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7260446448716403e-05,
+      "loss": 0.9473,
+      "step": 8254
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725973371886193e-05,
+      "loss": 1.1738,
+      "step": 8255
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725902091102602e-05,
+      "loss": 1.0703,
+      "step": 8256
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7258308025216322e-05,
+      "loss": 1.0327,
+      "step": 8257
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257595061440502e-05,
+      "loss": 1.1143,
+      "step": 8258
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7256882019706213e-05,
+      "loss": 1.3936,
+      "step": 8259
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7256168900021114e-05,
+      "loss": 1.2451,
+      "step": 8260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725545570239287e-05,
+      "loss": 1.2041,
+      "step": 8261
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7254742426829136e-05,
+      "loss": 1.123,
+      "step": 8262
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7254029073337577e-05,
+      "loss": 1.1294,
+      "step": 8263
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7253315641925854e-05,
+      "loss": 1.0811,
+      "step": 8264
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7252602132601636e-05,
+      "loss": 1.374,
+      "step": 8265
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725188854537258e-05,
+      "loss": 0.9346,
+      "step": 8266
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725117488024635e-05,
+      "loss": 1.2217,
+      "step": 8267
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725046113723062e-05,
+      "loss": 1.0361,
+      "step": 8268
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724974731633305e-05,
+      "loss": 1.2969,
+      "step": 8269
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724903341756131e-05,
+      "loss": 1.0938,
+      "step": 8270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7248319440923073e-05,
+      "loss": 1.3164,
+      "step": 8271
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7247605386426e-05,
+      "loss": 0.9253,
+      "step": 8272
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7246891254077766e-05,
+      "loss": 0.8198,
+      "step": 8273
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724617704388604e-05,
+      "loss": 1.1572,
+      "step": 8274
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7245462755858496e-05,
+      "loss": 0.9707,
+      "step": 8275
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7244748390002804e-05,
+      "loss": 1.1885,
+      "step": 8276
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724403394632664e-05,
+      "loss": 1.2568,
+      "step": 8277
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7243319424837676e-05,
+      "loss": 1.3828,
+      "step": 8278
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7242604825543587e-05,
+      "loss": 1.3633,
+      "step": 8279
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7241890148452052e-05,
+      "loss": 1.3613,
+      "step": 8280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7241175393570747e-05,
+      "loss": 0.9844,
+      "step": 8281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7240460560907345e-05,
+      "loss": 1.3438,
+      "step": 8282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723974565046953e-05,
+      "loss": 1.1309,
+      "step": 8283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7239030662264982e-05,
+      "loss": 1.0356,
+      "step": 8284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7238315596301374e-05,
+      "loss": 0.8467,
+      "step": 8285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7237600452586396e-05,
+      "loss": 1.3193,
+      "step": 8286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7236885231127725e-05,
+      "loss": 1.3926,
+      "step": 8287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723616993193304e-05,
+      "loss": 0.9795,
+      "step": 8288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723545455501003e-05,
+      "loss": 1.0586,
+      "step": 8289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723473910036638e-05,
+      "loss": 0.9355,
+      "step": 8290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723402356800977e-05,
+      "loss": 1.2305,
+      "step": 8291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7233307957947892e-05,
+      "loss": 0.9971,
+      "step": 8292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7232592270188426e-05,
+      "loss": 1.0742,
+      "step": 8293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231876504739067e-05,
+      "loss": 1.2686,
+      "step": 8294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231160661607497e-05,
+      "loss": 1.1899,
+      "step": 8295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723044474080141e-05,
+      "loss": 1.3545,
+      "step": 8296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7229728742328493e-05,
+      "loss": 0.8135,
+      "step": 8297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722901266619644e-05,
+      "loss": 1.4648,
+      "step": 8298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722829651241294e-05,
+      "loss": 0.8857,
+      "step": 8299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7227580280985687e-05,
+      "loss": 1.0684,
+      "step": 8300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7226863971922373e-05,
+      "loss": 1.3613,
+      "step": 8301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72261475852307e-05,
+      "loss": 1.2109,
+      "step": 8302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722543112091835e-05,
+      "loss": 0.7363,
+      "step": 8303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722471457899303e-05,
+      "loss": 1.165,
+      "step": 8304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722399795946243e-05,
+      "loss": 1.2246,
+      "step": 8305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7223281262334252e-05,
+      "loss": 1.6211,
+      "step": 8306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222564487616195e-05,
+      "loss": 1.5625,
+      "step": 8307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722184763531595e-05,
+      "loss": 1.0625,
+      "step": 8308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7221130705441228e-05,
+      "loss": 1.124,
+      "step": 8309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7220413697999724e-05,
+      "loss": 0.9824,
+      "step": 8310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7219696612999143e-05,
+      "loss": 1.084,
+      "step": 8311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7218979450447183e-05,
+      "loss": 1.3174,
+      "step": 8312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7218262210351552e-05,
+      "loss": 1.0947,
+      "step": 8313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7217544892719952e-05,
+      "loss": 1.2598,
+      "step": 8314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721682749756009e-05,
+      "loss": 1.1533,
+      "step": 8315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7216110024879672e-05,
+      "loss": 0.9912,
+      "step": 8316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7215392474686404e-05,
+      "loss": 1.1104,
+      "step": 8317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7214674846987992e-05,
+      "loss": 0.9946,
+      "step": 8318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7213957141792144e-05,
+      "loss": 0.8125,
+      "step": 8319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721323935910657e-05,
+      "loss": 1.021,
+      "step": 8320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7212521498938987e-05,
+      "loss": 0.7773,
+      "step": 8321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211803561297097e-05,
+      "loss": 0.75,
+      "step": 8322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211085546188617e-05,
+      "loss": 1.5273,
+      "step": 8323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7210367453621258e-05,
+      "loss": 1.0635,
+      "step": 8324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720964928360273e-05,
+      "loss": 1.0781,
+      "step": 8325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7208931036140753e-05,
+      "loss": 1.5137,
+      "step": 8326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720821271124304e-05,
+      "loss": 1.1768,
+      "step": 8327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7207494308917305e-05,
+      "loss": 0.8633,
+      "step": 8328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720677582917127e-05,
+      "loss": 1.7207,
+      "step": 8329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7206057272012647e-05,
+      "loss": 0.8848,
+      "step": 8330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720533863744916e-05,
+      "loss": 1.2988,
+      "step": 8331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720461992548852e-05,
+      "loss": 0.9553,
+      "step": 8332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7203901136138455e-05,
+      "loss": 0.9922,
+      "step": 8333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720318226940668e-05,
+      "loss": 1.1187,
+      "step": 8334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7202463325300925e-05,
+      "loss": 1.2617,
+      "step": 8335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7201744303828904e-05,
+      "loss": 1.0195,
+      "step": 8336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7201025204998347e-05,
+      "loss": 1.2363,
+      "step": 8337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7200306028816974e-05,
+      "loss": 1.0088,
+      "step": 8338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719958677529251e-05,
+      "loss": 0.7095,
+      "step": 8339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7198867444432687e-05,
+      "loss": 1.0967,
+      "step": 8340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7198148036245223e-05,
+      "loss": 1.1758,
+      "step": 8341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7197428550737853e-05,
+      "loss": 1.0234,
+      "step": 8342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71967089879183e-05,
+      "loss": 0.9531,
+      "step": 8343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7195989347794296e-05,
+      "loss": 1.1729,
+      "step": 8344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719526963037357e-05,
+      "loss": 1.1777,
+      "step": 8345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7194549835663858e-05,
+      "loss": 0.9961,
+      "step": 8346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193829963672885e-05,
+      "loss": 1.29,
+      "step": 8347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193110014408387e-05,
+      "loss": 1.3701,
+      "step": 8348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7192389987878092e-05,
+      "loss": 1.21,
+      "step": 8349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7191669884089743e-05,
+      "loss": 0.9619,
+      "step": 8350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719094970305107e-05,
+      "loss": 1.4746,
+      "step": 8351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719022944476981e-05,
+      "loss": 1.1387,
+      "step": 8352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7189509109253698e-05,
+      "loss": 1.2822,
+      "step": 8353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7188788696510477e-05,
+      "loss": 0.9575,
+      "step": 8354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718806820654788e-05,
+      "loss": 1.0117,
+      "step": 8355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187347639373648e-05,
+      "loss": 0.959,
+      "step": 8356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718662699499552e-05,
+      "loss": 0.96,
+      "step": 8357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7185906273421237e-05,
+      "loss": 0.9512,
+      "step": 8358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7185185474658543e-05,
+      "loss": 0.9141,
+      "step": 8359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718446459871518e-05,
+      "loss": 0.7339,
+      "step": 8360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7183743645598887e-05,
+      "loss": 0.7715,
+      "step": 8361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7183022615317413e-05,
+      "loss": 1.0996,
+      "step": 8362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71823015078785e-05,
+      "loss": 1.1279,
+      "step": 8363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71815803232899e-05,
+      "loss": 1.0513,
+      "step": 8364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7180859061559354e-05,
+      "loss": 1.3672,
+      "step": 8365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718013772269461e-05,
+      "loss": 1.4375,
+      "step": 8366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7179416306703418e-05,
+      "loss": 0.9316,
+      "step": 8367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7178694813593525e-05,
+      "loss": 1.1729,
+      "step": 8368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7177973243372684e-05,
+      "loss": 0.8765,
+      "step": 8369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7177251596048647e-05,
+      "loss": 1.0381,
+      "step": 8370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7176529871629158e-05,
+      "loss": 1.0342,
+      "step": 8371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7175808070121976e-05,
+      "loss": 0.9639,
+      "step": 8372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7175086191534854e-05,
+      "loss": 1.25,
+      "step": 8373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7174364235875543e-05,
+      "loss": 1.3535,
+      "step": 8374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7173642203151802e-05,
+      "loss": 1.0723,
+      "step": 8375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7172920093371384e-05,
+      "loss": 1.0801,
+      "step": 8376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7172197906542046e-05,
+      "loss": 1.4199,
+      "step": 8377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7171475642671546e-05,
+      "loss": 1.1104,
+      "step": 8378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717075330176764e-05,
+      "loss": 1.2354,
+      "step": 8379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7170030883838093e-05,
+      "loss": 0.8396,
+      "step": 8380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7169308388890658e-05,
+      "loss": 1.2559,
+      "step": 8381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7168585816933102e-05,
+      "loss": 1.2832,
+      "step": 8382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7167863167973184e-05,
+      "loss": 1.2998,
+      "step": 8383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7167140442018665e-05,
+      "loss": 1.0791,
+      "step": 8384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7166417639077308e-05,
+      "loss": 1.293,
+      "step": 8385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7165694759156878e-05,
+      "loss": 1.0459,
+      "step": 8386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716497180226514e-05,
+      "loss": 1.2832,
+      "step": 8387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716424876840986e-05,
+      "loss": 1.0293,
+      "step": 8388
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716352565759881e-05,
+      "loss": 1.2461,
+      "step": 8389
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7162802469839747e-05,
+      "loss": 1.0537,
+      "step": 8390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7162079205140445e-05,
+      "loss": 1.0479,
+      "step": 8391
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716135586350867e-05,
+      "loss": 1.3145,
+      "step": 8392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7160632444952198e-05,
+      "loss": 1.1235,
+      "step": 8393
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7159908949478797e-05,
+      "loss": 0.8568,
+      "step": 8394
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7159185377096238e-05,
+      "loss": 1.6758,
+      "step": 8395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715846172781229e-05,
+      "loss": 1.0352,
+      "step": 8396
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157738001634727e-05,
+      "loss": 0.9785,
+      "step": 8397
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715701419857133e-05,
+      "loss": 1.1621,
+      "step": 8398
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7156290318629864e-05,
+      "loss": 1.2598,
+      "step": 8399
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715556636181811e-05,
+      "loss": 1.2822,
+      "step": 8400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7154842328143846e-05,
+      "loss": 0.7778,
+      "step": 8401
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7154118217614844e-05,
+      "loss": 1.1504,
+      "step": 8402
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715339403023889e-05,
+      "loss": 0.9653,
+      "step": 8403
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7152669766023758e-05,
+      "loss": 1.2676,
+      "step": 8404
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7151945424977227e-05,
+      "loss": 0.9707,
+      "step": 8405
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7151221007107074e-05,
+      "loss": 1.1025,
+      "step": 8406
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715049651242109e-05,
+      "loss": 1.4922,
+      "step": 8407
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7149771940927055e-05,
+      "loss": 1.2441,
+      "step": 8408
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7149047292632746e-05,
+      "loss": 1.1074,
+      "step": 8409
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714832256754595e-05,
+      "loss": 1.5527,
+      "step": 8410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714759776567445e-05,
+      "loss": 1.4121,
+      "step": 8411
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714687288702604e-05,
+      "loss": 1.2207,
+      "step": 8412
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7146147931608493e-05,
+      "loss": 1.0928,
+      "step": 8413
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7145422899429606e-05,
+      "loss": 0.9736,
+      "step": 8414
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7144697790497168e-05,
+      "loss": 1.4492,
+      "step": 8415
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714397260481896e-05,
+      "loss": 1.1943,
+      "step": 8416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7143247342402776e-05,
+      "loss": 1.2666,
+      "step": 8417
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714252200325641e-05,
+      "loss": 1.2725,
+      "step": 8418
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7141796587387647e-05,
+      "loss": 1.0146,
+      "step": 8419
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7141071094804283e-05,
+      "loss": 1.3535,
+      "step": 8420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714034552551411e-05,
+      "loss": 1.3281,
+      "step": 8421
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7139619879524917e-05,
+      "loss": 1.2617,
+      "step": 8422
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713889415684451e-05,
+      "loss": 1.1729,
+      "step": 8423
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7138168357480674e-05,
+      "loss": 0.9131,
+      "step": 8424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713744248144121e-05,
+      "loss": 1.0654,
+      "step": 8425
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7136716528733916e-05,
+      "loss": 1.2383,
+      "step": 8426
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7135990499366583e-05,
+      "loss": 1.2607,
+      "step": 8427
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713526439334702e-05,
+      "loss": 1.332,
+      "step": 8428
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7134538210683017e-05,
+      "loss": 0.918,
+      "step": 8429
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713381195138238e-05,
+      "loss": 1.1475,
+      "step": 8430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7133085615452912e-05,
+      "loss": 0.6846,
+      "step": 8431
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713235920290241e-05,
+      "loss": 1.1064,
+      "step": 8432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7131632713738677e-05,
+      "loss": 1.356,
+      "step": 8433
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713090614796952e-05,
+      "loss": 1.3574,
+      "step": 8434
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7130179505602745e-05,
+      "loss": 1.2354,
+      "step": 8435
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712945278664615e-05,
+      "loss": 1.3164,
+      "step": 8436
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7128725991107545e-05,
+      "loss": 0.9648,
+      "step": 8437
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7127999118994743e-05,
+      "loss": 0.9443,
+      "step": 8438
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7127272170315542e-05,
+      "loss": 1.002,
+      "step": 8439
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7126545145077756e-05,
+      "loss": 0.792,
+      "step": 8440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712581804328919e-05,
+      "loss": 1.1279,
+      "step": 8441
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7125090864957664e-05,
+      "loss": 1.0479,
+      "step": 8442
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7124363610090978e-05,
+      "loss": 1.4902,
+      "step": 8443
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7123636278696952e-05,
+      "loss": 1.2656,
+      "step": 8444
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7122908870783392e-05,
+      "loss": 1.0156,
+      "step": 8445
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712218138635812e-05,
+      "loss": 1.1904,
+      "step": 8446
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712145382542894e-05,
+      "loss": 1.2432,
+      "step": 8447
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7120726188003674e-05,
+      "loss": 1.1963,
+      "step": 8448
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7119998474090138e-05,
+      "loss": 1.248,
+      "step": 8449
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7119270683696144e-05,
+      "loss": 1.0083,
+      "step": 8450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7118542816829515e-05,
+      "loss": 1.0693,
+      "step": 8451
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7117814873498067e-05,
+      "loss": 0.8779,
+      "step": 8452
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711708685370962e-05,
+      "loss": 1.0112,
+      "step": 8453
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116358757471996e-05,
+      "loss": 1.2695,
+      "step": 8454
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7115630584793012e-05,
+      "loss": 1.0566,
+      "step": 8455
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7114902335680492e-05,
+      "loss": 0.9492,
+      "step": 8456
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711417401014226e-05,
+      "loss": 1.1357,
+      "step": 8457
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7113445608186134e-05,
+      "loss": 1.0693,
+      "step": 8458
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7112717129819945e-05,
+      "loss": 1.2412,
+      "step": 8459
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7111988575051514e-05,
+      "loss": 0.834,
+      "step": 8460
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7111259943888665e-05,
+      "loss": 1.2773,
+      "step": 8461
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711053123633923e-05,
+      "loss": 1.1133,
+      "step": 8462
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7109802452411033e-05,
+      "loss": 0.9209,
+      "step": 8463
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7109073592111905e-05,
+      "loss": 1.3232,
+      "step": 8464
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7108344655449674e-05,
+      "loss": 0.917,
+      "step": 8465
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7107615642432166e-05,
+      "loss": 0.9932,
+      "step": 8466
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7106886553067217e-05,
+      "loss": 0.8384,
+      "step": 8467
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710615738736266e-05,
+      "loss": 1.4922,
+      "step": 8468
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710542814532632e-05,
+      "loss": 1.4258,
+      "step": 8469
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710469882696604e-05,
+      "loss": 1.5786,
+      "step": 8470
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710396943228964e-05,
+      "loss": 1.1372,
+      "step": 8471
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710323996130497e-05,
+      "loss": 0.9443,
+      "step": 8472
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7102510414019857e-05,
+      "loss": 1.502,
+      "step": 8473
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710178079044214e-05,
+      "loss": 1.3018,
+      "step": 8474
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101051090579658e-05,
+      "loss": 0.8574,
+      "step": 8475
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7100321314440243e-05,
+      "loss": 1.1084,
+      "step": 8476
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7099591462031743e-05,
+      "loss": 1.0195,
+      "step": 8477
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709886153336199e-05,
+      "loss": 1.2461,
+      "step": 8478
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709813152843883e-05,
+      "loss": 1.1904,
+      "step": 8479
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7097401447270103e-05,
+      "loss": 1.6719,
+      "step": 8480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7096671289863645e-05,
+      "loss": 1.2441,
+      "step": 8481
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7095941056227306e-05,
+      "loss": 1.3418,
+      "step": 8482
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7095210746368932e-05,
+      "loss": 0.9531,
+      "step": 8483
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7094480360296363e-05,
+      "loss": 1.186,
+      "step": 8484
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093749898017445e-05,
+      "loss": 1.0859,
+      "step": 8485
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093019359540027e-05,
+      "loss": 1.457,
+      "step": 8486
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709228874487195e-05,
+      "loss": 0.9082,
+      "step": 8487
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709155805402107e-05,
+      "loss": 1.0352,
+      "step": 8488
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709082728699523e-05,
+      "loss": 1.2402,
+      "step": 8489
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7090096443802286e-05,
+      "loss": 0.894,
+      "step": 8490
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7089365524450082e-05,
+      "loss": 1.4219,
+      "step": 8491
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708863452894647e-05,
+      "loss": 1.1641,
+      "step": 8492
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7087903457299303e-05,
+      "loss": 0.8955,
+      "step": 8493
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708717230951643e-05,
+      "loss": 1.1357,
+      "step": 8494
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708644108560572e-05,
+      "loss": 0.8594,
+      "step": 8495
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7085709785575007e-05,
+      "loss": 1.291,
+      "step": 8496
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7084978409432165e-05,
+      "loss": 1.0645,
+      "step": 8497
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7084246957185036e-05,
+      "loss": 1.6328,
+      "step": 8498
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7083515428841483e-05,
+      "loss": 1.1152,
+      "step": 8499
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708278382440936e-05,
+      "loss": 1.0264,
+      "step": 8500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708205214389653e-05,
+      "loss": 1.4238,
+      "step": 8501
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7081320387310857e-05,
+      "loss": 1.3086,
+      "step": 8502
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708058855466019e-05,
+      "loss": 1.5215,
+      "step": 8503
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7079856645952396e-05,
+      "loss": 1.0117,
+      "step": 8504
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7079124661195338e-05,
+      "loss": 1.2666,
+      "step": 8505
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7078392600396876e-05,
+      "loss": 1.2412,
+      "step": 8506
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7077660463564874e-05,
+      "loss": 1.0811,
+      "step": 8507
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7076928250707196e-05,
+      "loss": 1.3457,
+      "step": 8508
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7076195961831713e-05,
+      "loss": 1.6113,
+      "step": 8509
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707546359694628e-05,
+      "loss": 0.9102,
+      "step": 8510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7074731156058772e-05,
+      "loss": 1.1357,
+      "step": 8511
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7073998639177056e-05,
+      "loss": 1.5371,
+      "step": 8512
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7073266046308997e-05,
+      "loss": 0.8857,
+      "step": 8513
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707253337746247e-05,
+      "loss": 1.3516,
+      "step": 8514
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707180063264534e-05,
+      "loss": 1.1611,
+      "step": 8515
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7071067811865477e-05,
+      "loss": 1.2207,
+      "step": 8516
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707033491513076e-05,
+      "loss": 1.6816,
+      "step": 8517
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7069601942449047e-05,
+      "loss": 1.0332,
+      "step": 8518
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706886889382823e-05,
+      "loss": 1.292,
+      "step": 8519
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7068135769276166e-05,
+      "loss": 1.5723,
+      "step": 8520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706740256880074e-05,
+      "loss": 1.2529,
+      "step": 8521
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706666929240983e-05,
+      "loss": 1.1348,
+      "step": 8522
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7065935940111305e-05,
+      "loss": 1.1846,
+      "step": 8523
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7065202511913045e-05,
+      "loss": 0.9707,
+      "step": 8524
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706446900782293e-05,
+      "loss": 1.7119,
+      "step": 8525
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7063735427848837e-05,
+      "loss": 1.0259,
+      "step": 8526
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7063001771998647e-05,
+      "loss": 1.3164,
+      "step": 8527
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7062268040280242e-05,
+      "loss": 1.1094,
+      "step": 8528
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70615342327015e-05,
+      "loss": 1.375,
+      "step": 8529
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7060800349270307e-05,
+      "loss": 0.9629,
+      "step": 8530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706006638999454e-05,
+      "loss": 1.3359,
+      "step": 8531
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705933235488209e-05,
+      "loss": 0.8145,
+      "step": 8532
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705859824394084e-05,
+      "loss": 1.0957,
+      "step": 8533
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7057864057178674e-05,
+      "loss": 0.9507,
+      "step": 8534
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7057129794603477e-05,
+      "loss": 1.1509,
+      "step": 8535
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705639545622314e-05,
+      "loss": 1.1299,
+      "step": 8536
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705566104204555e-05,
+      "loss": 0.9414,
+      "step": 8537
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705492655207859e-05,
+      "loss": 1.249,
+      "step": 8538
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705419198633016e-05,
+      "loss": 1.0908,
+      "step": 8539
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7053457344808144e-05,
+      "loss": 1.0312,
+      "step": 8540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705272262752043e-05,
+      "loss": 0.9502,
+      "step": 8541
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7051987834474918e-05,
+      "loss": 1.4297,
+      "step": 8542
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7051252965679498e-05,
+      "loss": 1.3359,
+      "step": 8543
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705051802114206e-05,
+      "loss": 1.1211,
+      "step": 8544
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70497830008705e-05,
+      "loss": 1.1904,
+      "step": 8545
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7049047904872715e-05,
+      "loss": 1.1504,
+      "step": 8546
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7048312733156605e-05,
+      "loss": 1.041,
+      "step": 8547
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7047577485730057e-05,
+      "loss": 1.1738,
+      "step": 8548
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7046842162600977e-05,
+      "loss": 1.3066,
+      "step": 8549
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7046106763777265e-05,
+      "loss": 1.3232,
+      "step": 8550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7045371289266812e-05,
+      "loss": 0.7661,
+      "step": 8551
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7044635739077523e-05,
+      "loss": 1.0156,
+      "step": 8552
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70439001132173e-05,
+      "loss": 0.8867,
+      "step": 8553
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7043164411694043e-05,
+      "loss": 1.2617,
+      "step": 8554
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7042428634515655e-05,
+      "loss": 1.1494,
+      "step": 8555
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704169278169004e-05,
+      "loss": 1.2744,
+      "step": 8556
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70409568532251e-05,
+      "loss": 1.1152,
+      "step": 8557
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7040220849128742e-05,
+      "loss": 1.2793,
+      "step": 8558
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7039484769408874e-05,
+      "loss": 1.4238,
+      "step": 8559
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7038748614073403e-05,
+      "loss": 0.9185,
+      "step": 8560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703801238313023e-05,
+      "loss": 1.2461,
+      "step": 8561
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7037276076587267e-05,
+      "loss": 1.167,
+      "step": 8562
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7036539694452425e-05,
+      "loss": 0.9194,
+      "step": 8563
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7035803236733612e-05,
+      "loss": 1.2646,
+      "step": 8564
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7035066703438737e-05,
+      "loss": 1.3047,
+      "step": 8565
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7034330094575717e-05,
+      "loss": 0.7988,
+      "step": 8566
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703359341015246e-05,
+      "loss": 1.2988,
+      "step": 8567
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703285665017688e-05,
+      "loss": 1.0303,
+      "step": 8568
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7032119814656893e-05,
+      "loss": 0.9023,
+      "step": 8569
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703138290360041e-05,
+      "loss": 0.8936,
+      "step": 8570
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703064591701535e-05,
+      "loss": 0.9409,
+      "step": 8571
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7029908854909628e-05,
+      "loss": 0.7686,
+      "step": 8572
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7029171717291163e-05,
+      "loss": 1.1055,
+      "step": 8573
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702843450416787e-05,
+      "loss": 1.0693,
+      "step": 8574
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702769721554767e-05,
+      "loss": 0.9766,
+      "step": 8575
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7026959851438483e-05,
+      "loss": 0.8662,
+      "step": 8576
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7026222411848225e-05,
+      "loss": 1.3535,
+      "step": 8577
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7025484896784824e-05,
+      "loss": 0.9922,
+      "step": 8578
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70247473062562e-05,
+      "loss": 1.127,
+      "step": 8579
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702400964027027e-05,
+      "loss": 1.04,
+      "step": 8580
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702327189883497e-05,
+      "loss": 1.042,
+      "step": 8581
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702253408195821e-05,
+      "loss": 1.0454,
+      "step": 8582
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702179618964793e-05,
+      "loss": 1.3379,
+      "step": 8583
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7021058221912045e-05,
+      "loss": 1.1143,
+      "step": 8584
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7020320178758485e-05,
+      "loss": 0.6938,
+      "step": 8585
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701958206019518e-05,
+      "loss": 1.0977,
+      "step": 8586
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7018843866230056e-05,
+      "loss": 1.2324,
+      "step": 8587
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7018105596871046e-05,
+      "loss": 1.1006,
+      "step": 8588
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7017367252126076e-05,
+      "loss": 1.0967,
+      "step": 8589
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701662883200308e-05,
+      "loss": 1.2119,
+      "step": 8590
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7015890336509988e-05,
+      "loss": 0.6807,
+      "step": 8591
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7015151765654733e-05,
+      "loss": 1.4551,
+      "step": 8592
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701441311944525e-05,
+      "loss": 0.8193,
+      "step": 8593
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701367439788947e-05,
+      "loss": 0.9912,
+      "step": 8594
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701293560099533e-05,
+      "loss": 0.9854,
+      "step": 8595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701219672877077e-05,
+      "loss": 1.1953,
+      "step": 8596
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7011457781223722e-05,
+      "loss": 1.3867,
+      "step": 8597
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7010718758362122e-05,
+      "loss": 1.3359,
+      "step": 8598
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7009979660193912e-05,
+      "loss": 1.3145,
+      "step": 8599
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700924048672703e-05,
+      "loss": 1.4512,
+      "step": 8600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700850123796941e-05,
+      "loss": 1.4648,
+      "step": 8601
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7007761913929007e-05,
+      "loss": 1.1426,
+      "step": 8602
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700702251461375e-05,
+      "loss": 0.979,
+      "step": 8603
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7006283040031587e-05,
+      "loss": 1.1475,
+      "step": 8604
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7005543490190458e-05,
+      "loss": 1.0381,
+      "step": 8605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700480386509831e-05,
+      "loss": 0.8528,
+      "step": 8606
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7004064164763084e-05,
+      "loss": 1.2227,
+      "step": 8607
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700332438919273e-05,
+      "loss": 1.2979,
+      "step": 8608
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700258453839519e-05,
+      "loss": 0.916,
+      "step": 8609
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7001844612378415e-05,
+      "loss": 1.2012,
+      "step": 8610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700110461115035e-05,
+      "loss": 0.936,
+      "step": 8611
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7000364534718946e-05,
+      "loss": 0.6973,
+      "step": 8612
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6999624383092153e-05,
+      "loss": 1.2334,
+      "step": 8613
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6998884156277918e-05,
+      "loss": 0.9336,
+      "step": 8614
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69981438542842e-05,
+      "loss": 1.2178,
+      "step": 8615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699740347711894e-05,
+      "loss": 1.0161,
+      "step": 8616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69966630247901e-05,
+      "loss": 1.125,
+      "step": 8617
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699592249730563e-05,
+      "loss": 1.1729,
+      "step": 8618
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6995181894673483e-05,
+      "loss": 1.25,
+      "step": 8619
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6994441216901617e-05,
+      "loss": 1.2178,
+      "step": 8620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6993700463997984e-05,
+      "loss": 1.5273,
+      "step": 8621
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699295963597055e-05,
+      "loss": 1.3652,
+      "step": 8622
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6992218732827264e-05,
+      "loss": 0.917,
+      "step": 8623
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6991477754576084e-05,
+      "loss": 1.2803,
+      "step": 8624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6990736701224975e-05,
+      "loss": 1.1367,
+      "step": 8625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6989995572781898e-05,
+      "loss": 0.9102,
+      "step": 8626
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6989254369254805e-05,
+      "loss": 0.9497,
+      "step": 8627
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698851309065167e-05,
+      "loss": 0.9668,
+      "step": 8628
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698777173698044e-05,
+      "loss": 1.4414,
+      "step": 8629
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6987030308249097e-05,
+      "loss": 1.2246,
+      "step": 8630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698628880446559e-05,
+      "loss": 1.002,
+      "step": 8631
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6985547225637885e-05,
+      "loss": 1.1846,
+      "step": 8632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698480557177396e-05,
+      "loss": 1.1045,
+      "step": 8633
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698406384288177e-05,
+      "loss": 1.0596,
+      "step": 8634
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6983322038969286e-05,
+      "loss": 1.0054,
+      "step": 8635
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6982580160044478e-05,
+      "loss": 1.1602,
+      "step": 8636
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698183820611531e-05,
+      "loss": 1.084,
+      "step": 8637
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698109617718976e-05,
+      "loss": 1.0898,
+      "step": 8638
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698035407327579e-05,
+      "loss": 1.3799,
+      "step": 8639
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6979611894381377e-05,
+      "loss": 0.9707,
+      "step": 8640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697886964051449e-05,
+      "loss": 1.1191,
+      "step": 8641
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6978127311683103e-05,
+      "loss": 1.2139,
+      "step": 8642
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6977384907895193e-05,
+      "loss": 1.5117,
+      "step": 8643
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697664242915873e-05,
+      "loss": 1.3047,
+      "step": 8644
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697589987548169e-05,
+      "loss": 1.2627,
+      "step": 8645
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697515724687205e-05,
+      "loss": 1.1934,
+      "step": 8646
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697441454333779e-05,
+      "loss": 0.9248,
+      "step": 8647
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6973671764886886e-05,
+      "loss": 0.7734,
+      "step": 8648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6972928911527315e-05,
+      "loss": 1.0508,
+      "step": 8649
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697218598326706e-05,
+      "loss": 1.3008,
+      "step": 8650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6971442980114095e-05,
+      "loss": 0.7705,
+      "step": 8651
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6970699902076405e-05,
+      "loss": 1.1211,
+      "step": 8652
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696995674916197e-05,
+      "loss": 0.9336,
+      "step": 8653
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6969213521378777e-05,
+      "loss": 0.7231,
+      "step": 8654
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696847021873481e-05,
+      "loss": 1.4629,
+      "step": 8655
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6967726841238047e-05,
+      "loss": 1.0957,
+      "step": 8656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6966983388896477e-05,
+      "loss": 1.2534,
+      "step": 8657
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6966239861718083e-05,
+      "loss": 1.2002,
+      "step": 8658
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965496259710857e-05,
+      "loss": 0.9707,
+      "step": 8659
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696475258288278e-05,
+      "loss": 1.4473,
+      "step": 8660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696400883124185e-05,
+      "loss": 0.8022,
+      "step": 8661
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6963265004796044e-05,
+      "loss": 1.5078,
+      "step": 8662
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696252110355336e-05,
+      "loss": 1.417,
+      "step": 8663
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6961777127521784e-05,
+      "loss": 0.9531,
+      "step": 8664
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696103307670931e-05,
+      "loss": 1.0586,
+      "step": 8665
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6960288951123933e-05,
+      "loss": 1.2725,
+      "step": 8666
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6959544750773643e-05,
+      "loss": 0.6548,
+      "step": 8667
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695880047566643e-05,
+      "loss": 1.2334,
+      "step": 8668
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69580561258103e-05,
+      "loss": 1.4375,
+      "step": 8669
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6957311701213238e-05,
+      "loss": 1.1514,
+      "step": 8670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695656720188324e-05,
+      "loss": 0.6177,
+      "step": 8671
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6955822627828316e-05,
+      "loss": 0.7256,
+      "step": 8672
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695507797905645e-05,
+      "loss": 1.0547,
+      "step": 8673
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6954333255575646e-05,
+      "loss": 1.4551,
+      "step": 8674
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6953588457393902e-05,
+      "loss": 1.0938,
+      "step": 8675
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695284358451922e-05,
+      "loss": 1.3779,
+      "step": 8676
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6952098636959606e-05,
+      "loss": 1.1055,
+      "step": 8677
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695135361472305e-05,
+      "loss": 1.1797,
+      "step": 8678
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6950608517817565e-05,
+      "loss": 1.124,
+      "step": 8679
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6949863346251155e-05,
+      "loss": 0.645,
+      "step": 8680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6949118100031815e-05,
+      "loss": 1.3086,
+      "step": 8681
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6948372779167558e-05,
+      "loss": 1.1924,
+      "step": 8682
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6947627383666383e-05,
+      "loss": 1.0635,
+      "step": 8683
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6946881913536306e-05,
+      "loss": 0.7725,
+      "step": 8684
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694613636878533e-05,
+      "loss": 0.8657,
+      "step": 8685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694539074942146e-05,
+      "loss": 0.9238,
+      "step": 8686
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694464505545271e-05,
+      "loss": 1.3086,
+      "step": 8687
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694389928688709e-05,
+      "loss": 1.0469,
+      "step": 8688
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694315344373261e-05,
+      "loss": 1.126,
+      "step": 8689
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6942407525997277e-05,
+      "loss": 1.5137,
+      "step": 8690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694166153368911e-05,
+      "loss": 1.1104,
+      "step": 8691
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694091546681612e-05,
+      "loss": 0.8857,
+      "step": 8692
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694016932538632e-05,
+      "loss": 1.3027,
+      "step": 8693
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6939423109407724e-05,
+      "loss": 1.1738,
+      "step": 8694
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6938676818888352e-05,
+      "loss": 1.3237,
+      "step": 8695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6937930453836212e-05,
+      "loss": 1.252,
+      "step": 8696
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6937184014259333e-05,
+      "loss": 1.1895,
+      "step": 8697
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693643750016572e-05,
+      "loss": 1.3975,
+      "step": 8698
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69356909115634e-05,
+      "loss": 0.9746,
+      "step": 8699
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693494424846039e-05,
+      "loss": 0.5425,
+      "step": 8700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6934197510864713e-05,
+      "loss": 0.9053,
+      "step": 8701
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6933450698784387e-05,
+      "loss": 1.2734,
+      "step": 8702
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6932703812227435e-05,
+      "loss": 1.1841,
+      "step": 8703
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693195685120188e-05,
+      "loss": 1.0322,
+      "step": 8704
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6931209815715747e-05,
+      "loss": 0.873,
+      "step": 8705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6930462705777057e-05,
+      "loss": 1.4062,
+      "step": 8706
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692971552139384e-05,
+      "loss": 1.084,
+      "step": 8707
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6928968262574118e-05,
+      "loss": 1.1787,
+      "step": 8708
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6928220929325915e-05,
+      "loss": 0.9531,
+      "step": 8709
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6927473521657268e-05,
+      "loss": 1.0254,
+      "step": 8710
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6926726039576198e-05,
+      "loss": 1.3311,
+      "step": 8711
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6925978483090734e-05,
+      "loss": 1.4609,
+      "step": 8712
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6925230852208912e-05,
+      "loss": 1.5605,
+      "step": 8713
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6924483146938756e-05,
+      "loss": 0.9326,
+      "step": 8714
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6923735367288303e-05,
+      "loss": 1.5957,
+      "step": 8715
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6922987513265582e-05,
+      "loss": 1.042,
+      "step": 8716
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6922239584878624e-05,
+      "loss": 1.0557,
+      "step": 8717
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692149158213547e-05,
+      "loss": 1.0605,
+      "step": 8718
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6920743505044147e-05,
+      "loss": 1.2832,
+      "step": 8719
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6919995353612698e-05,
+      "loss": 1.0576,
+      "step": 8720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6919247127849155e-05,
+      "loss": 1.4141,
+      "step": 8721
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6918498827761554e-05,
+      "loss": 1.207,
+      "step": 8722
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6917750453357935e-05,
+      "loss": 1.0312,
+      "step": 8723
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6917002004646338e-05,
+      "loss": 0.9697,
+      "step": 8724
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69162534816348e-05,
+      "loss": 1.2383,
+      "step": 8725
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6915504884331364e-05,
+      "loss": 1.0928,
+      "step": 8726
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6914756212744067e-05,
+      "loss": 1.1572,
+      "step": 8727
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6914007466880956e-05,
+      "loss": 0.6426,
+      "step": 8728
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6913258646750074e-05,
+      "loss": 0.835,
+      "step": 8729
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6912509752359457e-05,
+      "loss": 1.2959,
+      "step": 8730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6911760783717153e-05,
+      "loss": 0.7904,
+      "step": 8731
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6911011740831212e-05,
+      "loss": 1.417,
+      "step": 8732
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6910262623709677e-05,
+      "loss": 1.2988,
+      "step": 8733
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6909513432360592e-05,
+      "loss": 0.9854,
+      "step": 8734
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690876416679201e-05,
+      "loss": 1.0361,
+      "step": 8735
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6908014827011975e-05,
+      "loss": 1.0557,
+      "step": 8736
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6907265413028536e-05,
+      "loss": 0.9639,
+      "step": 8737
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6906515924849745e-05,
+      "loss": 1.3662,
+      "step": 8738
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690576636248365e-05,
+      "loss": 1.0947,
+      "step": 8739
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6905016725938307e-05,
+      "loss": 1.1953,
+      "step": 8740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6904267015221765e-05,
+      "loss": 1.3652,
+      "step": 8741
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6903517230342077e-05,
+      "loss": 1.0312,
+      "step": 8742
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6902767371307297e-05,
+      "loss": 1.4766,
+      "step": 8743
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6902017438125483e-05,
+      "loss": 0.9736,
+      "step": 8744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6901267430804684e-05,
+      "loss": 1.1172,
+      "step": 8745
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6900517349352964e-05,
+      "loss": 1.3545,
+      "step": 8746
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899767193778377e-05,
+      "loss": 1.1714,
+      "step": 8747
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899016964088976e-05,
+      "loss": 1.3525,
+      "step": 8748
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6898266660292825e-05,
+      "loss": 1.2373,
+      "step": 8749
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6897516282397982e-05,
+      "loss": 1.2764,
+      "step": 8750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6896765830412512e-05,
+      "loss": 1.0586,
+      "step": 8751
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689601530434447e-05,
+      "loss": 1.0791,
+      "step": 8752
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689526470420192e-05,
+      "loss": 1.1338,
+      "step": 8753
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6894514029992922e-05,
+      "loss": 1.2148,
+      "step": 8754
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6893763281725543e-05,
+      "loss": 1.2314,
+      "step": 8755
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6893012459407847e-05,
+      "loss": 1.251,
+      "step": 8756
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6892261563047897e-05,
+      "loss": 0.9194,
+      "step": 8757
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6891510592653764e-05,
+      "loss": 1.1982,
+      "step": 8758
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6890759548233505e-05,
+      "loss": 1.1113,
+      "step": 8759
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6890008429795198e-05,
+      "loss": 0.8779,
+      "step": 8760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6889257237346902e-05,
+      "loss": 1.1299,
+      "step": 8761
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6888505970896696e-05,
+      "loss": 1.2188,
+      "step": 8762
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688775463045264e-05,
+      "loss": 1.1582,
+      "step": 8763
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688700321602281e-05,
+      "loss": 1.0303,
+      "step": 8764
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6886251727615278e-05,
+      "loss": 1.0811,
+      "step": 8765
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6885500165238113e-05,
+      "loss": 1.2227,
+      "step": 8766
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6884748528899386e-05,
+      "loss": 0.8535,
+      "step": 8767
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688399681860718e-05,
+      "loss": 0.8506,
+      "step": 8768
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688324503436956e-05,
+      "loss": 1.5234,
+      "step": 8769
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688249317619461e-05,
+      "loss": 1.3975,
+      "step": 8770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6881741244090396e-05,
+      "loss": 1.1133,
+      "step": 8771
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6880989238065003e-05,
+      "loss": 1.0913,
+      "step": 8772
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688023715812651e-05,
+      "loss": 1.2695,
+      "step": 8773
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6879485004282987e-05,
+      "loss": 1.0752,
+      "step": 8774
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687873277654252e-05,
+      "loss": 0.8877,
+      "step": 8775
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6877980474913188e-05,
+      "loss": 0.9229,
+      "step": 8776
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687722809940307e-05,
+      "loss": 1.4375,
+      "step": 8777
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687647565002025e-05,
+      "loss": 1.1211,
+      "step": 8778
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687572312677281e-05,
+      "loss": 1.001,
+      "step": 8779
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687497052966883e-05,
+      "loss": 1.209,
+      "step": 8780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68742178587164e-05,
+      "loss": 0.9106,
+      "step": 8781
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68734651139236e-05,
+      "loss": 0.8467,
+      "step": 8782
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6872712295298523e-05,
+      "loss": 1.1934,
+      "step": 8783
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6871959402849244e-05,
+      "loss": 0.8938,
+      "step": 8784
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687120643658386e-05,
+      "loss": 1.1074,
+      "step": 8785
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6870453396510456e-05,
+      "loss": 1.3926,
+      "step": 8786
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686970028263712e-05,
+      "loss": 1.0264,
+      "step": 8787
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6868947094971942e-05,
+      "loss": 1.1924,
+      "step": 8788
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686819383352301e-05,
+      "loss": 1.1396,
+      "step": 8789
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686744049829842e-05,
+      "loss": 1.0049,
+      "step": 8790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686668708930626e-05,
+      "loss": 1.2617,
+      "step": 8791
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6865933606554626e-05,
+      "loss": 1.0059,
+      "step": 8792
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686518005005161e-05,
+      "loss": 1.2383,
+      "step": 8793
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6864426419805312e-05,
+      "loss": 1.1084,
+      "step": 8794
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686367271582382e-05,
+      "loss": 0.7485,
+      "step": 8795
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686291893811523e-05,
+      "loss": 1.0684,
+      "step": 8796
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6862165086687642e-05,
+      "loss": 1.2373,
+      "step": 8797
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6861411161549154e-05,
+      "loss": 1.3965,
+      "step": 8798
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6860657162707858e-05,
+      "loss": 1.2217,
+      "step": 8799
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685990309017186e-05,
+      "loss": 1.3135,
+      "step": 8800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685914894394926e-05,
+      "loss": 1.4375,
+      "step": 8801
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6858394724048156e-05,
+      "loss": 1.1143,
+      "step": 8802
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685764043047665e-05,
+      "loss": 1.0176,
+      "step": 8803
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6856886063242847e-05,
+      "loss": 1.3789,
+      "step": 8804
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6856131622354847e-05,
+      "loss": 1.2715,
+      "step": 8805
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6855377107820757e-05,
+      "loss": 0.8672,
+      "step": 8806
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6854622519648672e-05,
+      "loss": 0.8623,
+      "step": 8807
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685386785784671e-05,
+      "loss": 0.8779,
+      "step": 8808
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6853113122422975e-05,
+      "loss": 1.3086,
+      "step": 8809
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6852358313385568e-05,
+      "loss": 0.8574,
+      "step": 8810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68516034307426e-05,
+      "loss": 1.0752,
+      "step": 8811
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6850848474502185e-05,
+      "loss": 1.3789,
+      "step": 8812
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685009344467242e-05,
+      "loss": 1.0723,
+      "step": 8813
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6849338341261428e-05,
+      "loss": 1.0518,
+      "step": 8814
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6848583164277314e-05,
+      "loss": 1.251,
+      "step": 8815
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6847827913728188e-05,
+      "loss": 1.6719,
+      "step": 8816
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6847072589622172e-05,
+      "loss": 1.2383,
+      "step": 8817
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6846317191967364e-05,
+      "loss": 1.6328,
+      "step": 8818
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6845561720771892e-05,
+      "loss": 1.0518,
+      "step": 8819
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6844806176043867e-05,
+      "loss": 1.4395,
+      "step": 8820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68440505577914e-05,
+      "loss": 0.9297,
+      "step": 8821
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684329486602261e-05,
+      "loss": 1.2549,
+      "step": 8822
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684253910074562e-05,
+      "loss": 0.9189,
+      "step": 8823
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6841783261968546e-05,
+      "loss": 1.0156,
+      "step": 8824
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68410273496995e-05,
+      "loss": 1.0898,
+      "step": 8825
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6840271363946608e-05,
+      "loss": 1.0947,
+      "step": 8826
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683951530471799e-05,
+      "loss": 1.0576,
+      "step": 8827
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6838759172021763e-05,
+      "loss": 1.1182,
+      "step": 8828
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6838002965866055e-05,
+      "loss": 1.4385,
+      "step": 8829
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6837246686258987e-05,
+      "loss": 0.9688,
+      "step": 8830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683649033320868e-05,
+      "loss": 1.2461,
+      "step": 8831
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6835733906723263e-05,
+      "loss": 1.5615,
+      "step": 8832
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6834977406810854e-05,
+      "loss": 0.9248,
+      "step": 8833
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683422083347959e-05,
+      "loss": 1.1445,
+      "step": 8834
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683346418673759e-05,
+      "loss": 1.1436,
+      "step": 8835
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683270746659298e-05,
+      "loss": 1.1562,
+      "step": 8836
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6831950673053895e-05,
+      "loss": 1.1836,
+      "step": 8837
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683119380612846e-05,
+      "loss": 1.1558,
+      "step": 8838
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6830436865824803e-05,
+      "loss": 1.0605,
+      "step": 8839
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682967985215106e-05,
+      "loss": 1.2969,
+      "step": 8840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6828922765115362e-05,
+      "loss": 0.9619,
+      "step": 8841
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682816560472584e-05,
+      "loss": 1.2139,
+      "step": 8842
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682740837099062e-05,
+      "loss": 1.2607,
+      "step": 8843
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682665106391785e-05,
+      "loss": 0.9678,
+      "step": 8844
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6825893683515654e-05,
+      "loss": 1.0693,
+      "step": 8845
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6825136229792172e-05,
+      "loss": 0.7881,
+      "step": 8846
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682437870275554e-05,
+      "loss": 1.3086,
+      "step": 8847
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6823621102413893e-05,
+      "loss": 0.8652,
+      "step": 8848
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6822863428775372e-05,
+      "loss": 1.2305,
+      "step": 8849
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6822105681848112e-05,
+      "loss": 0.8477,
+      "step": 8850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682134786164025e-05,
+      "loss": 0.8936,
+      "step": 8851
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682058996815994e-05,
+      "loss": 0.9395,
+      "step": 8852
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6819832001415306e-05,
+      "loss": 1.2236,
+      "step": 8853
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6819073961414497e-05,
+      "loss": 0.9336,
+      "step": 8854
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6818315848165663e-05,
+      "loss": 1.0933,
+      "step": 8855
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6817557661676933e-05,
+      "loss": 1.2119,
+      "step": 8856
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681679940195646e-05,
+      "loss": 1.0996,
+      "step": 8857
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681604106901239e-05,
+      "loss": 0.876,
+      "step": 8858
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6815282662852865e-05,
+      "loss": 0.7041,
+      "step": 8859
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681452418348603e-05,
+      "loss": 1.3301,
+      "step": 8860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6813765630920042e-05,
+      "loss": 1.1357,
+      "step": 8861
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6813007005163034e-05,
+      "loss": 1.1919,
+      "step": 8862
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6812248306223165e-05,
+      "loss": 0.9495,
+      "step": 8863
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6811489534108585e-05,
+      "loss": 1.4502,
+      "step": 8864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681073068882744e-05,
+      "loss": 1.0723,
+      "step": 8865
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6809971770387888e-05,
+      "loss": 1.0303,
+      "step": 8866
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6809212778798068e-05,
+      "loss": 1.6602,
+      "step": 8867
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6808453714066147e-05,
+      "loss": 1.1914,
+      "step": 8868
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6807694576200273e-05,
+      "loss": 1.4355,
+      "step": 8869
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6806935365208598e-05,
+      "loss": 0.7266,
+      "step": 8870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680617608109928e-05,
+      "loss": 0.7397,
+      "step": 8871
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6805416723880473e-05,
+      "loss": 1.1816,
+      "step": 8872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680465729356034e-05,
+      "loss": 0.9648,
+      "step": 8873
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6803897790147026e-05,
+      "loss": 1.165,
+      "step": 8874
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68031382136487e-05,
+      "loss": 0.8237,
+      "step": 8875
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6802378564073524e-05,
+      "loss": 0.8054,
+      "step": 8876
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6801618841429642e-05,
+      "loss": 1.0674,
+      "step": 8877
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680085904572523e-05,
+      "loss": 1.3447,
+      "step": 8878
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680009917696844e-05,
+      "loss": 1.0107,
+      "step": 8879
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6799339235167443e-05,
+      "loss": 0.6924,
+      "step": 8880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6798579220330395e-05,
+      "loss": 0.8516,
+      "step": 8881
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6797819132465462e-05,
+      "loss": 0.8213,
+      "step": 8882
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679705897158081e-05,
+      "loss": 1.0259,
+      "step": 8883
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.67962987376846e-05,
+      "loss": 1.0244,
+      "step": 8884
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6795538430785005e-05,
+      "loss": 0.959,
+      "step": 8885
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794778050890187e-05,
+      "loss": 0.9727,
+      "step": 8886
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794017598008312e-05,
+      "loss": 0.9658,
+      "step": 8887
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6793257072147552e-05,
+      "loss": 1.2734,
+      "step": 8888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6792496473316076e-05,
+      "loss": 1.043,
+      "step": 8889
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6791735801522056e-05,
+      "loss": 0.7683,
+      "step": 8890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6790975056773658e-05,
+      "loss": 1.0576,
+      "step": 8891
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6790214239079056e-05,
+      "loss": 1.0908,
+      "step": 8892
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6789453348446424e-05,
+      "loss": 1.3154,
+      "step": 8893
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6788692384883933e-05,
+      "loss": 0.9521,
+      "step": 8894
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678793134839976e-05,
+      "loss": 1.2822,
+      "step": 8895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6787170239002075e-05,
+      "loss": 1.1514,
+      "step": 8896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6786409056699054e-05,
+      "loss": 1.0098,
+      "step": 8897
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678564780149888e-05,
+      "loss": 1.0674,
+      "step": 8898
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6784886473409724e-05,
+      "loss": 0.8848,
+      "step": 8899
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6784125072439765e-05,
+      "loss": 1.0566,
+      "step": 8900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6783363598597186e-05,
+      "loss": 1.0957,
+      "step": 8901
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6782602051890156e-05,
+      "loss": 1.4766,
+      "step": 8902
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678184043232687e-05,
+      "loss": 1.583,
+      "step": 8903
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678107873991549e-05,
+      "loss": 0.9297,
+      "step": 8904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678031697466422e-05,
+      "loss": 1.3301,
+      "step": 8905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6779555136581228e-05,
+      "loss": 1.2109,
+      "step": 8906
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67787932256747e-05,
+      "loss": 0.9502,
+      "step": 8907
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6778031241952816e-05,
+      "loss": 1.2871,
+      "step": 8908
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6777269185423774e-05,
+      "loss": 0.8267,
+      "step": 8909
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6776507056095746e-05,
+      "loss": 0.9902,
+      "step": 8910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6775744853976926e-05,
+      "loss": 1.4033,
+      "step": 8911
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6774982579075503e-05,
+      "loss": 1.4316,
+      "step": 8912
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6774220231399658e-05,
+      "loss": 1.248,
+      "step": 8913
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6773457810957584e-05,
+      "loss": 1.1709,
+      "step": 8914
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677269531775747e-05,
+      "loss": 1.0498,
+      "step": 8915
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6771932751807506e-05,
+      "loss": 1.166,
+      "step": 8916
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6771170113115888e-05,
+      "loss": 1.4844,
+      "step": 8917
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770407401690796e-05,
+      "loss": 1.0674,
+      "step": 8918
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676964461754044e-05,
+      "loss": 1.0,
+      "step": 8919
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6768881760672998e-05,
+      "loss": 1.0752,
+      "step": 8920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6768118831096672e-05,
+      "loss": 1.2227,
+      "step": 8921
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6767355828819653e-05,
+      "loss": 1.1787,
+      "step": 8922
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6766592753850142e-05,
+      "loss": 1.1963,
+      "step": 8923
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676582960619633e-05,
+      "loss": 1.4707,
+      "step": 8924
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676506638586642e-05,
+      "loss": 0.9297,
+      "step": 8925
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676430309286861e-05,
+      "loss": 1.2363,
+      "step": 8926
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6763539727211094e-05,
+      "loss": 1.3809,
+      "step": 8927
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6762776288902073e-05,
+      "loss": 1.0454,
+      "step": 8928
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6762012777949756e-05,
+      "loss": 1.3008,
+      "step": 8929
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676124919436233e-05,
+      "loss": 1.085,
+      "step": 8930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760485538148006e-05,
+      "loss": 1.0215,
+      "step": 8931
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675972180931499e-05,
+      "loss": 1.2383,
+      "step": 8932
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675895800787148e-05,
+      "loss": 1.249,
+      "step": 8933
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6758194133825675e-05,
+      "loss": 1.167,
+      "step": 8934
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6757430187185796e-05,
+      "loss": 1.4688,
+      "step": 8935
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6756666167960035e-05,
+      "loss": 1.0674,
+      "step": 8936
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6755902076156606e-05,
+      "loss": 1.207,
+      "step": 8937
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6755137911783713e-05,
+      "loss": 0.9004,
+      "step": 8938
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6754373674849563e-05,
+      "loss": 1.2588,
+      "step": 8939
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6753609365362373e-05,
+      "loss": 1.3477,
+      "step": 8940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6752844983330347e-05,
+      "loss": 1.9102,
+      "step": 8941
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6752080528761697e-05,
+      "loss": 0.832,
+      "step": 8942
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6751316001664634e-05,
+      "loss": 1.1182,
+      "step": 8943
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6750551402047368e-05,
+      "loss": 1.1465,
+      "step": 8944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6749786729918115e-05,
+      "loss": 1.2129,
+      "step": 8945
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674902198528509e-05,
+      "loss": 0.7739,
+      "step": 8946
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6748257168156507e-05,
+      "loss": 1.8398,
+      "step": 8947
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6747492278540578e-05,
+      "loss": 1.1592,
+      "step": 8948
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674672731644552e-05,
+      "loss": 1.2021,
+      "step": 8949
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6745962281879556e-05,
+      "loss": 1.2568,
+      "step": 8950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67451971748509e-05,
+      "loss": 1.2686,
+      "step": 8951
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6744431995367762e-05,
+      "loss": 1.1055,
+      "step": 8952
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6743666743438372e-05,
+      "loss": 1.0322,
+      "step": 8953
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674290141907095e-05,
+      "loss": 0.853,
+      "step": 8954
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6742136022273713e-05,
+      "loss": 1.7109,
+      "step": 8955
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6741370553054882e-05,
+      "loss": 1.0391,
+      "step": 8956
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674060501142268e-05,
+      "loss": 1.1289,
+      "step": 8957
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6739839397385336e-05,
+      "loss": 1.2568,
+      "step": 8958
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6739073710951066e-05,
+      "loss": 0.5986,
+      "step": 8959
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6738307952128097e-05,
+      "loss": 1.2744,
+      "step": 8960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6737542120924655e-05,
+      "loss": 1.3516,
+      "step": 8961
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736776217348966e-05,
+      "loss": 1.3506,
+      "step": 8962
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673601024140926e-05,
+      "loss": 0.9165,
+      "step": 8963
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673524419311376e-05,
+      "loss": 1.1885,
+      "step": 8964
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6734478072470696e-05,
+      "loss": 1.1084,
+      "step": 8965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6733711879488304e-05,
+      "loss": 0.7524,
+      "step": 8966
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6732945614174805e-05,
+      "loss": 0.8232,
+      "step": 8967
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6732179276538434e-05,
+      "loss": 1.1831,
+      "step": 8968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673141286658742e-05,
+      "loss": 0.8242,
+      "step": 8969
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673064638433e-05,
+      "loss": 1.3193,
+      "step": 8970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6729879829774403e-05,
+      "loss": 1.2676,
+      "step": 8971
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6729113202928868e-05,
+      "loss": 0.7861,
+      "step": 8972
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6728346503801627e-05,
+      "loss": 1.4707,
+      "step": 8973
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672757973240091e-05,
+      "loss": 1.1543,
+      "step": 8974
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726812888734964e-05,
+      "loss": 1.251,
+      "step": 8975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726045972812022e-05,
+      "loss": 0.5146,
+      "step": 8976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6725278984640318e-05,
+      "loss": 0.9316,
+      "step": 8977
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6724511924228093e-05,
+      "loss": 0.7451,
+      "step": 8978
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6723744791583593e-05,
+      "loss": 0.9385,
+      "step": 8979
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6722977586715046e-05,
+      "loss": 1.0742,
+      "step": 8980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67222103096307e-05,
+      "loss": 1.2734,
+      "step": 8981
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67214429603388e-05,
+      "loss": 1.2109,
+      "step": 8982
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672067553884758e-05,
+      "loss": 1.2715,
+      "step": 8983
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6719908045165295e-05,
+      "loss": 0.9004,
+      "step": 8984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6719140479300176e-05,
+      "loss": 0.8086,
+      "step": 8985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6718372841260478e-05,
+      "loss": 0.9336,
+      "step": 8986
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6717605131054443e-05,
+      "loss": 0.8955,
+      "step": 8987
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6716837348690315e-05,
+      "loss": 1.1396,
+      "step": 8988
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6716069494176344e-05,
+      "loss": 0.8076,
+      "step": 8989
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6715301567520777e-05,
+      "loss": 1.3662,
+      "step": 8990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6714533568731867e-05,
+      "loss": 0.8403,
+      "step": 8991
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6713765497817858e-05,
+      "loss": 0.7871,
+      "step": 8992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6712997354787003e-05,
+      "loss": 0.876,
+      "step": 8993
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6712229139647552e-05,
+      "loss": 1.334,
+      "step": 8994
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6711460852407756e-05,
+      "loss": 1.3594,
+      "step": 8995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671069249307587e-05,
+      "loss": 1.3271,
+      "step": 8996
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670992406166015e-05,
+      "loss": 0.9019,
+      "step": 8997
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6709155558168843e-05,
+      "loss": 1.0,
+      "step": 8998
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670838698261021e-05,
+      "loss": 0.9082,
+      "step": 8999
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6707618334992503e-05,
+      "loss": 1.2373,
+      "step": 9000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670684961532398e-05,
+      "loss": 1.5098,
+      "step": 9001
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6706080823612897e-05,
+      "loss": 0.9336,
+      "step": 9002
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6705311959867515e-05,
+      "loss": 1.0781,
+      "step": 9003
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670454302409609e-05,
+      "loss": 0.9736,
+      "step": 9004
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6703774016306885e-05,
+      "loss": 1.1479,
+      "step": 9005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6703004936508155e-05,
+      "loss": 1.3467,
+      "step": 9006
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6702235784708168e-05,
+      "loss": 0.8057,
+      "step": 9007
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6701466560915182e-05,
+      "loss": 1.0088,
+      "step": 9008
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6700697265137456e-05,
+      "loss": 1.0635,
+      "step": 9009
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6699927897383262e-05,
+      "loss": 1.4443,
+      "step": 9010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669915845766086e-05,
+      "loss": 1.3223,
+      "step": 9011
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6698388945978515e-05,
+      "loss": 1.1924,
+      "step": 9012
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669761936234449e-05,
+      "loss": 1.418,
+      "step": 9013
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6696849706767056e-05,
+      "loss": 1.4727,
+      "step": 9014
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669607997925448e-05,
+      "loss": 1.3857,
+      "step": 9015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669531017981503e-05,
+      "loss": 1.3379,
+      "step": 9016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6694540308456965e-05,
+      "loss": 1.1689,
+      "step": 9017
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6693770365188573e-05,
+      "loss": 0.6445,
+      "step": 9018
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6693000350018108e-05,
+      "loss": 1.2441,
+      "step": 9019
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669223026295385e-05,
+      "loss": 1.5,
+      "step": 9020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6691460104004074e-05,
+      "loss": 1.1055,
+      "step": 9021
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6690689873177045e-05,
+      "loss": 1.3428,
+      "step": 9022
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668991957048104e-05,
+      "loss": 0.748,
+      "step": 9023
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668914919592433e-05,
+      "loss": 0.8408,
+      "step": 9024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6688378749515193e-05,
+      "loss": 0.874,
+      "step": 9025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668760823126191e-05,
+      "loss": 1.0176,
+      "step": 9026
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6686837641172745e-05,
+      "loss": 1.6367,
+      "step": 9027
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668606697925599e-05,
+      "loss": 1.4307,
+      "step": 9028
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6685296245519913e-05,
+      "loss": 1.21,
+      "step": 9029
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6684525439972793e-05,
+      "loss": 1.2812,
+      "step": 9030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6683754562622918e-05,
+      "loss": 1.3486,
+      "step": 9031
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668298361347856e-05,
+      "loss": 0.7715,
+      "step": 9032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6682212592548e-05,
+      "loss": 0.9873,
+      "step": 9033
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6681441499839527e-05,
+      "loss": 1.582,
+      "step": 9034
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668067033536142e-05,
+      "loss": 0.9795,
+      "step": 9035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6679899099121964e-05,
+      "loss": 0.9038,
+      "step": 9036
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6679127791129438e-05,
+      "loss": 1.1357,
+      "step": 9037
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6678356411392135e-05,
+      "loss": 0.9673,
+      "step": 9038
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6677584959918337e-05,
+      "loss": 1.1211,
+      "step": 9039
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6676813436716325e-05,
+      "loss": 1.2979,
+      "step": 9040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6676041841794397e-05,
+      "loss": 1.3965,
+      "step": 9041
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6675270175160837e-05,
+      "loss": 1.3105,
+      "step": 9042
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6674498436823933e-05,
+      "loss": 1.0527,
+      "step": 9043
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6673726626791974e-05,
+      "loss": 1.1904,
+      "step": 9044
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667295474507325e-05,
+      "loss": 1.3936,
+      "step": 9045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6672182791676058e-05,
+      "loss": 0.9473,
+      "step": 9046
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667141076660868e-05,
+      "loss": 1.0986,
+      "step": 9047
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6670638669879418e-05,
+      "loss": 1.0488,
+      "step": 9048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666986650149656e-05,
+      "loss": 1.0811,
+      "step": 9049
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6669094261468405e-05,
+      "loss": 1.0654,
+      "step": 9050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6668321949803246e-05,
+      "loss": 1.1475,
+      "step": 9051
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6667549566509377e-05,
+      "loss": 1.2754,
+      "step": 9052
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6666777111595098e-05,
+      "loss": 0.9404,
+      "step": 9053
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6666004585068703e-05,
+      "loss": 1.252,
+      "step": 9054
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6665231986938493e-05,
+      "loss": 1.1533,
+      "step": 9055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6664459317212765e-05,
+      "loss": 1.2002,
+      "step": 9056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666368657589982e-05,
+      "loss": 0.9541,
+      "step": 9057
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666291376300796e-05,
+      "loss": 1.1104,
+      "step": 9058
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6662140878545484e-05,
+      "loss": 1.1777,
+      "step": 9059
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6661367922520692e-05,
+      "loss": 1.0898,
+      "step": 9060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666059489494189e-05,
+      "loss": 1.1699,
+      "step": 9061
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665982179581738e-05,
+      "loss": 1.4824,
+      "step": 9062
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665904862515547e-05,
+      "loss": 1.248,
+      "step": 9063
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665827538296446e-05,
+      "loss": 1.1455,
+      "step": 9064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665750206925266e-05,
+      "loss": 1.1719,
+      "step": 9065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665672868402837e-05,
+      "loss": 1.4219,
+      "step": 9066
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665595522729991e-05,
+      "loss": 0.8779,
+      "step": 9067
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6655181699075574e-05,
+      "loss": 0.5886,
+      "step": 9068
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6654408099363683e-05,
+      "loss": 1.293,
+      "step": 9069
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6653634428172538e-05,
+      "loss": 1.1729,
+      "step": 9070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665286068551045e-05,
+      "loss": 1.0977,
+      "step": 9071
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665208687138574e-05,
+      "loss": 0.9443,
+      "step": 9072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6651312985806706e-05,
+      "loss": 0.9932,
+      "step": 9073
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665053902878167e-05,
+      "loss": 0.7646,
+      "step": 9074
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6649765000318944e-05,
+      "loss": 0.8916,
+      "step": 9075
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6648990900426843e-05,
+      "loss": 1.1211,
+      "step": 9076
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6648216729113677e-05,
+      "loss": 0.959,
+      "step": 9077
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664744248638777e-05,
+      "loss": 1.2002,
+      "step": 9078
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664666817225743e-05,
+      "loss": 1.1592,
+      "step": 9079
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664589378673098e-05,
+      "loss": 1.2432,
+      "step": 9080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664511932981674e-05,
+      "loss": 0.8555,
+      "step": 9081
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6644344801523022e-05,
+      "loss": 0.9971,
+      "step": 9082
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6643570201858152e-05,
+      "loss": 0.8232,
+      "step": 9083
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6642795530830444e-05,
+      "loss": 1.1895,
+      "step": 9084
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6642020788448227e-05,
+      "loss": 1.2139,
+      "step": 9085
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664124597471982e-05,
+      "loss": 1.2451,
+      "step": 9086
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6640471089653543e-05,
+      "loss": 0.6958,
+      "step": 9087
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663969613325772e-05,
+      "loss": 1.4766,
+      "step": 9088
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6638921105540676e-05,
+      "loss": 0.9902,
+      "step": 9089
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6638146006510744e-05,
+      "loss": 1.0488,
+      "step": 9090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6637370836176233e-05,
+      "loss": 1.2148,
+      "step": 9091
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6636595594545486e-05,
+      "loss": 0.9966,
+      "step": 9092
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6635820281626822e-05,
+      "loss": 1.1309,
+      "step": 9093
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663504489742857e-05,
+      "loss": 1.0684,
+      "step": 9094
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663426944195906e-05,
+      "loss": 0.9229,
+      "step": 9095
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6633493915226625e-05,
+      "loss": 1.2373,
+      "step": 9096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6632718317239585e-05,
+      "loss": 0.9434,
+      "step": 9097
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631942648006284e-05,
+      "loss": 1.666,
+      "step": 9098
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631166907535045e-05,
+      "loss": 1.0391,
+      "step": 9099
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6630391095834203e-05,
+      "loss": 1.0488,
+      "step": 9100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6629615212912097e-05,
+      "loss": 0.8154,
+      "step": 9101
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662883925877705e-05,
+      "loss": 0.8589,
+      "step": 9102
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6628063233437404e-05,
+      "loss": 1.4473,
+      "step": 9103
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6627287136901497e-05,
+      "loss": 1.2246,
+      "step": 9104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662651096917766e-05,
+      "loss": 1.749,
+      "step": 9105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6625734730274238e-05,
+      "loss": 1.2871,
+      "step": 9106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662495842019956e-05,
+      "loss": 0.9023,
+      "step": 9107
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662418203896197e-05,
+      "loss": 1.0166,
+      "step": 9108
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662340558656981e-05,
+      "loss": 1.0444,
+      "step": 9109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6622629063031413e-05,
+      "loss": 1.0811,
+      "step": 9110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6621852468355124e-05,
+      "loss": 1.6328,
+      "step": 9111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6621075802549285e-05,
+      "loss": 1.1353,
+      "step": 9112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662029906562224e-05,
+      "loss": 1.1113,
+      "step": 9113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661952225758233e-05,
+      "loss": 1.2568,
+      "step": 9114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6618745378437902e-05,
+      "loss": 1.1846,
+      "step": 9115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6617968428197298e-05,
+      "loss": 1.1592,
+      "step": 9116
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6617191406868864e-05,
+      "loss": 1.4551,
+      "step": 9117
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661641431446095e-05,
+      "loss": 1.0996,
+      "step": 9118
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6615637150981897e-05,
+      "loss": 1.4395,
+      "step": 9119
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661485991644006e-05,
+      "loss": 0.7539,
+      "step": 9120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6614082610843785e-05,
+      "loss": 1.1392,
+      "step": 9121
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661330523420142e-05,
+      "loss": 1.1064,
+      "step": 9122
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6612527786521314e-05,
+      "loss": 1.0645,
+      "step": 9123
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6611750267811824e-05,
+      "loss": 1.0781,
+      "step": 9124
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6610972678081293e-05,
+      "loss": 0.8291,
+      "step": 9125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6610195017338085e-05,
+      "loss": 1.1797,
+      "step": 9126
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6609417285590542e-05,
+      "loss": 1.2881,
+      "step": 9127
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6608639482847027e-05,
+      "loss": 1.5029,
+      "step": 9128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660786160911589e-05,
+      "loss": 1.0957,
+      "step": 9129
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6607083664405487e-05,
+      "loss": 1.0146,
+      "step": 9130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6606305648724175e-05,
+      "loss": 0.8564,
+      "step": 9131
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6605527562080313e-05,
+      "loss": 1.1455,
+      "step": 9132
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6604749404482255e-05,
+      "loss": 1.1826,
+      "step": 9133
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660397117593836e-05,
+      "loss": 1.3281,
+      "step": 9134
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6603192876456997e-05,
+      "loss": 1.1426,
+      "step": 9135
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6602414506046513e-05,
+      "loss": 0.9648,
+      "step": 9136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6601636064715278e-05,
+      "loss": 1.1045,
+      "step": 9137
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6600857552471644e-05,
+      "loss": 0.8379,
+      "step": 9138
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6600078969323987e-05,
+      "loss": 1.2998,
+      "step": 9139
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6599300315280658e-05,
+      "loss": 1.1182,
+      "step": 9140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659852159035003e-05,
+      "loss": 0.9824,
+      "step": 9141
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6597742794540462e-05,
+      "loss": 1.168,
+      "step": 9142
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659696392786032e-05,
+      "loss": 0.9209,
+      "step": 9143
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6596184990317974e-05,
+      "loss": 0.9004,
+      "step": 9144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659540598192179e-05,
+      "loss": 0.9902,
+      "step": 9145
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659462690268013e-05,
+      "loss": 0.9355,
+      "step": 9146
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659384775260137e-05,
+      "loss": 1.0801,
+      "step": 9147
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6593068531693875e-05,
+      "loss": 0.999,
+      "step": 9148
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6592289239966016e-05,
+      "loss": 1.3828,
+      "step": 9149
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659150987742617e-05,
+      "loss": 1.2441,
+      "step": 9150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6590730444082697e-05,
+      "loss": 0.7378,
+      "step": 9151
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6589950939943976e-05,
+      "loss": 1.1221,
+      "step": 9152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6589171365018387e-05,
+      "loss": 1.0957,
+      "step": 9153
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6588391719314288e-05,
+      "loss": 0.8301,
+      "step": 9154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6587612002840067e-05,
+      "loss": 1.25,
+      "step": 9155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658683221560409e-05,
+      "loss": 1.3926,
+      "step": 9156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6586052357614745e-05,
+      "loss": 1.2852,
+      "step": 9157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65852724288804e-05,
+      "loss": 0.6724,
+      "step": 9158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6584492429409433e-05,
+      "loss": 1.2363,
+      "step": 9159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658371235921022e-05,
+      "loss": 1.1934,
+      "step": 9160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658293221829115e-05,
+      "loss": 0.939,
+      "step": 9161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6582152006660597e-05,
+      "loss": 1.1465,
+      "step": 9162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658137172432694e-05,
+      "loss": 1.3379,
+      "step": 9163
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6580591371298563e-05,
+      "loss": 0.9141,
+      "step": 9164
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6579810947583846e-05,
+      "loss": 1.5498,
+      "step": 9165
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657903045319118e-05,
+      "loss": 0.8428,
+      "step": 9166
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6578249888128937e-05,
+      "loss": 1.041,
+      "step": 9167
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657746925240551e-05,
+      "loss": 1.2383,
+      "step": 9168
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6576688546029277e-05,
+      "loss": 0.9529,
+      "step": 9169
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6575907769008633e-05,
+      "loss": 1.168,
+      "step": 9170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6575126921351958e-05,
+      "loss": 0.9126,
+      "step": 9171
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6574346003067642e-05,
+      "loss": 0.7754,
+      "step": 9172
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573565014164076e-05,
+      "loss": 1.1455,
+      "step": 9173
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6572783954649644e-05,
+      "loss": 1.1113,
+      "step": 9174
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657200282453274e-05,
+      "loss": 1.293,
+      "step": 9175
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6571221623821753e-05,
+      "loss": 1.1582,
+      "step": 9176
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6570440352525073e-05,
+      "loss": 0.9365,
+      "step": 9177
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6569659010651093e-05,
+      "loss": 1.1348,
+      "step": 9178
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6568877598208207e-05,
+      "loss": 1.0996,
+      "step": 9179
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656809611520481e-05,
+      "loss": 0.959,
+      "step": 9180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6567314561649287e-05,
+      "loss": 1.0371,
+      "step": 9181
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6566532937550048e-05,
+      "loss": 0.6997,
+      "step": 9182
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656575124291548e-05,
+      "loss": 1.2031,
+      "step": 9183
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6564969477753982e-05,
+      "loss": 0.9189,
+      "step": 9184
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6564187642073946e-05,
+      "loss": 0.8271,
+      "step": 9185
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656340573588378e-05,
+      "loss": 1.6465,
+      "step": 9186
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6562623759191872e-05,
+      "loss": 0.6455,
+      "step": 9187
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6561841712006632e-05,
+      "loss": 1.2578,
+      "step": 9188
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6561059594336457e-05,
+      "loss": 1.0811,
+      "step": 9189
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6560277406189742e-05,
+      "loss": 1.1611,
+      "step": 9190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6559495147574897e-05,
+      "loss": 1.3496,
+      "step": 9191
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655871281850032e-05,
+      "loss": 0.8765,
+      "step": 9192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655793041897442e-05,
+      "loss": 1.1748,
+      "step": 9193
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6557147949005597e-05,
+      "loss": 1.2256,
+      "step": 9194
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6556365408602257e-05,
+      "loss": 0.877,
+      "step": 9195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6555582797772803e-05,
+      "loss": 1.2266,
+      "step": 9196
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6554800116525644e-05,
+      "loss": 1.0967,
+      "step": 9197
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6554017364869187e-05,
+      "loss": 1.0039,
+      "step": 9198
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655323454281184e-05,
+      "loss": 1.2119,
+      "step": 9199
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6552451650362014e-05,
+      "loss": 1.2803,
+      "step": 9200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6551668687528114e-05,
+      "loss": 1.2832,
+      "step": 9201
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6550885654318556e-05,
+      "loss": 0.7244,
+      "step": 9202
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6550102550741745e-05,
+      "loss": 0.8848,
+      "step": 9203
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6549319376806096e-05,
+      "loss": 1.1194,
+      "step": 9204
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6548536132520023e-05,
+      "loss": 1.1611,
+      "step": 9205
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6547752817891937e-05,
+      "loss": 1.1426,
+      "step": 9206
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6546969432930257e-05,
+      "loss": 1.123,
+      "step": 9207
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6546185977643385e-05,
+      "loss": 1.2402,
+      "step": 9208
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654540245203975e-05,
+      "loss": 0.8428,
+      "step": 9209
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6544618856127764e-05,
+      "loss": 1.0718,
+      "step": 9210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6543835189915844e-05,
+      "loss": 1.3311,
+      "step": 9211
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6543051453412403e-05,
+      "loss": 0.9448,
+      "step": 9212
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6542267646625866e-05,
+      "loss": 1.2832,
+      "step": 9213
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6541483769564653e-05,
+      "loss": 1.4414,
+      "step": 9214
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6540699822237182e-05,
+      "loss": 1.1504,
+      "step": 9215
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.653991580465187e-05,
+      "loss": 1.3457,
+      "step": 9216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6539131716817146e-05,
+      "loss": 0.917,
+      "step": 9217
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6538347558741424e-05,
+      "loss": 1.29,
+      "step": 9218
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6537563330433134e-05,
+      "loss": 0.9551,
+      "step": 9219
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6536779031900697e-05,
+      "loss": 1.1406,
+      "step": 9220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6535994663152536e-05,
+      "loss": 0.9502,
+      "step": 9221
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6535210224197083e-05,
+      "loss": 0.9512,
+      "step": 9222
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6534425715042758e-05,
+      "loss": 1.2832,
+      "step": 9223
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6533641135697988e-05,
+      "loss": 0.9404,
+      "step": 9224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6532856486171204e-05,
+      "loss": 1.3242,
+      "step": 9225
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653207176647083e-05,
+      "loss": 1.291,
+      "step": 9226
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6531286976605302e-05,
+      "loss": 0.9922,
+      "step": 9227
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6530502116583046e-05,
+      "loss": 1.7285,
+      "step": 9228
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652971718641249e-05,
+      "loss": 0.8682,
+      "step": 9229
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6528932186102067e-05,
+      "loss": 1.3984,
+      "step": 9230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652814711566021e-05,
+      "loss": 1.0371,
+      "step": 9231
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6527361975095354e-05,
+      "loss": 1.0,
+      "step": 9232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6526576764415935e-05,
+      "loss": 1.208,
+      "step": 9233
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6525791483630377e-05,
+      "loss": 0.873,
+      "step": 9234
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6525006132747124e-05,
+      "loss": 1.4531,
+      "step": 9235
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6524220711774604e-05,
+      "loss": 1.074,
+      "step": 9236
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6523435220721266e-05,
+      "loss": 1.4004,
+      "step": 9237
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652264965959554e-05,
+      "loss": 1.1484,
+      "step": 9238
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6521864028405863e-05,
+      "loss": 1.1904,
+      "step": 9239
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6521078327160675e-05,
+      "loss": 1.1455,
+      "step": 9240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6520292555868417e-05,
+      "loss": 1.0186,
+      "step": 9241
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651950671453753e-05,
+      "loss": 0.8945,
+      "step": 9242
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6518720803176454e-05,
+      "loss": 1.0645,
+      "step": 9243
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6517934821793628e-05,
+      "loss": 1.2881,
+      "step": 9244
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6517148770397505e-05,
+      "loss": 1.0703,
+      "step": 9245
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6516362648996515e-05,
+      "loss": 1.1675,
+      "step": 9246
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651557645759911e-05,
+      "loss": 1.0107,
+      "step": 9247
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651479019621373e-05,
+      "loss": 1.123,
+      "step": 9248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651400386484883e-05,
+      "loss": 1.1396,
+      "step": 9249
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651321746351285e-05,
+      "loss": 1.1621,
+      "step": 9250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6512430992214235e-05,
+      "loss": 1.2598,
+      "step": 9251
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6511644450961442e-05,
+      "loss": 1.0063,
+      "step": 9252
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6510857839762908e-05,
+      "loss": 1.0967,
+      "step": 9253
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651007115862709e-05,
+      "loss": 0.8125,
+      "step": 9254
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650928440756244e-05,
+      "loss": 1.1045,
+      "step": 9255
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.65084975865774e-05,
+      "loss": 1.2715,
+      "step": 9256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6507710695680427e-05,
+      "loss": 0.7646,
+      "step": 9257
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6506923734879977e-05,
+      "loss": 1.4316,
+      "step": 9258
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.65061367041845e-05,
+      "loss": 1.0576,
+      "step": 9259
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6505349603602448e-05,
+      "loss": 1.2871,
+      "step": 9260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6504562433142282e-05,
+      "loss": 0.9023,
+      "step": 9261
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650377519281245e-05,
+      "loss": 0.7603,
+      "step": 9262
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6502987882621408e-05,
+      "loss": 1.3076,
+      "step": 9263
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650220050257762e-05,
+      "loss": 1.0576,
+      "step": 9264
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6501413052689542e-05,
+      "loss": 0.9316,
+      "step": 9265
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650062553296563e-05,
+      "loss": 1.002,
+      "step": 9266
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649983794341434e-05,
+      "loss": 1.375,
+      "step": 9267
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649905028404414e-05,
+      "loss": 0.8682,
+      "step": 9268
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6498262554863488e-05,
+      "loss": 1.707,
+      "step": 9269
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6497474755880843e-05,
+      "loss": 1.2168,
+      "step": 9270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6496686887104666e-05,
+      "loss": 1.1162,
+      "step": 9271
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6495898948543426e-05,
+      "loss": 1.0693,
+      "step": 9272
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649511094020558e-05,
+      "loss": 1.1865,
+      "step": 9273
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64943228620996e-05,
+      "loss": 1.2344,
+      "step": 9274
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649353471423394e-05,
+      "loss": 1.1963,
+      "step": 9275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649274649661708e-05,
+      "loss": 1.1338,
+      "step": 9276
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6491958209257476e-05,
+      "loss": 1.4531,
+      "step": 9277
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64911698521636e-05,
+      "loss": 1.2266,
+      "step": 9278
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649038142534392e-05,
+      "loss": 1.0039,
+      "step": 9279
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6489592928806904e-05,
+      "loss": 1.207,
+      "step": 9280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6488804362561023e-05,
+      "loss": 1.3599,
+      "step": 9281
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6488015726614745e-05,
+      "loss": 1.1133,
+      "step": 9282
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6487227020976545e-05,
+      "loss": 1.1357,
+      "step": 9283
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6486438245654892e-05,
+      "loss": 1.1748,
+      "step": 9284
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648564940065826e-05,
+      "loss": 1.1377,
+      "step": 9285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648486048599512e-05,
+      "loss": 1.2139,
+      "step": 9286
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6484071501673953e-05,
+      "loss": 1.5371,
+      "step": 9287
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6483282447703228e-05,
+      "loss": 0.9053,
+      "step": 9288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648249332409142e-05,
+      "loss": 1.2119,
+      "step": 9289
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6481704130847013e-05,
+      "loss": 1.1553,
+      "step": 9290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648091486797847e-05,
+      "loss": 1.1758,
+      "step": 9291
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6480125535494284e-05,
+      "loss": 0.9922,
+      "step": 9292
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6479336133402927e-05,
+      "loss": 1.3594,
+      "step": 9293
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6478546661712876e-05,
+      "loss": 1.0029,
+      "step": 9294
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6477757120432617e-05,
+      "loss": 0.918,
+      "step": 9295
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6476967509570628e-05,
+      "loss": 1.1504,
+      "step": 9296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6476177829135392e-05,
+      "loss": 0.9707,
+      "step": 9297
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6475388079135387e-05,
+      "loss": 1.0928,
+      "step": 9298
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6474598259579102e-05,
+      "loss": 1.1865,
+      "step": 9299
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6473808370475023e-05,
+      "loss": 1.1328,
+      "step": 9300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6473018411831623e-05,
+      "loss": 1.4307,
+      "step": 9301
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6472228383657398e-05,
+      "loss": 1.375,
+      "step": 9302
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647143828596083e-05,
+      "loss": 0.8096,
+      "step": 9303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6470648118750407e-05,
+      "loss": 0.9561,
+      "step": 9304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6469857882034617e-05,
+      "loss": 1.2041,
+      "step": 9305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6469067575821947e-05,
+      "loss": 1.0713,
+      "step": 9306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646827720012089e-05,
+      "loss": 1.1021,
+      "step": 9307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646748675493993e-05,
+      "loss": 1.0361,
+      "step": 9308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6466696240287562e-05,
+      "loss": 1.041,
+      "step": 9309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6465905656172275e-05,
+      "loss": 1.2988,
+      "step": 9310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6465115002602563e-05,
+      "loss": 1.375,
+      "step": 9311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646432427958692e-05,
+      "loss": 1.4668,
+      "step": 9312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6463533487133838e-05,
+      "loss": 1.0879,
+      "step": 9313
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646274262525181e-05,
+      "loss": 1.2305,
+      "step": 9314
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6461951693949332e-05,
+      "loss": 1.0566,
+      "step": 9315
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64611606932349e-05,
+      "loss": 1.3135,
+      "step": 9316
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646036962311701e-05,
+      "loss": 1.4277,
+      "step": 9317
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6459578483604168e-05,
+      "loss": 1.1909,
+      "step": 9318
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6458787274704857e-05,
+      "loss": 1.2617,
+      "step": 9319
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6457995996427587e-05,
+      "loss": 0.8535,
+      "step": 9320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6457204648780854e-05,
+      "loss": 1.3818,
+      "step": 9321
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6456413231773157e-05,
+      "loss": 1.1924,
+      "step": 9322
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6455621745412998e-05,
+      "loss": 0.96,
+      "step": 9323
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645483018970888e-05,
+      "loss": 0.7974,
+      "step": 9324
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6454038564669306e-05,
+      "loss": 1.2793,
+      "step": 9325
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6453246870302778e-05,
+      "loss": 1.2178,
+      "step": 9326
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64524551066178e-05,
+      "loss": 0.7559,
+      "step": 9327
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6451663273622876e-05,
+      "loss": 1.0107,
+      "step": 9328
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6450871371326514e-05,
+      "loss": 1.3906,
+      "step": 9329
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6450079399737223e-05,
+      "loss": 1.4531,
+      "step": 9330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6449287358863505e-05,
+      "loss": 1.3936,
+      "step": 9331
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6448495248713864e-05,
+      "loss": 1.293,
+      "step": 9332
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644770306929682e-05,
+      "loss": 1.3789,
+      "step": 9333
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6446910820620875e-05,
+      "loss": 1.0889,
+      "step": 9334
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6446118502694536e-05,
+      "loss": 1.1572,
+      "step": 9335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644532611552632e-05,
+      "loss": 1.0117,
+      "step": 9336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6444533659124734e-05,
+      "loss": 1.502,
+      "step": 9337
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6443741133498297e-05,
+      "loss": 1.1074,
+      "step": 9338
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6442948538655518e-05,
+      "loss": 0.9902,
+      "step": 9339
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6442155874604907e-05,
+      "loss": 1.3574,
+      "step": 9340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6441363141354983e-05,
+      "loss": 1.1807,
+      "step": 9341
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6440570338914258e-05,
+      "loss": 0.9014,
+      "step": 9342
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6439777467291253e-05,
+      "loss": 1.1807,
+      "step": 9343
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6438984526494483e-05,
+      "loss": 1.1182,
+      "step": 9344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643819151653246e-05,
+      "loss": 1.3301,
+      "step": 9345
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6437398437413713e-05,
+      "loss": 1.1592,
+      "step": 9346
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643660528914675e-05,
+      "loss": 1.0791,
+      "step": 9347
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6435812071740098e-05,
+      "loss": 1.3623,
+      "step": 9348
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6435018785202273e-05,
+      "loss": 0.9727,
+      "step": 9349
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6434225429541803e-05,
+      "loss": 1.1309,
+      "step": 9350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6433432004767195e-05,
+      "loss": 1.2012,
+      "step": 9351
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643263851088699e-05,
+      "loss": 0.9951,
+      "step": 9352
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64318449479097e-05,
+      "loss": 1.2305,
+      "step": 9353
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6431051315843853e-05,
+      "loss": 1.1465,
+      "step": 9354
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6430257614697973e-05,
+      "loss": 1.0186,
+      "step": 9355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6429463844480588e-05,
+      "loss": 1.2695,
+      "step": 9356
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642867000520022e-05,
+      "loss": 1.1982,
+      "step": 9357
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6427876096865394e-05,
+      "loss": 0.7852,
+      "step": 9358
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6427082119484648e-05,
+      "loss": 1.4434,
+      "step": 9359
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64262880730665e-05,
+      "loss": 1.5078,
+      "step": 9360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642549395761949e-05,
+      "loss": 1.583,
+      "step": 9361
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642469977315214e-05,
+      "loss": 0.9141,
+      "step": 9362
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642390551967298e-05,
+      "loss": 1.1387,
+      "step": 9363
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642311119719055e-05,
+      "loss": 0.917,
+      "step": 9364
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6422316805713373e-05,
+      "loss": 0.9365,
+      "step": 9365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6421522345249988e-05,
+      "loss": 1.1934,
+      "step": 9366
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6420727815808928e-05,
+      "loss": 1.3311,
+      "step": 9367
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6419933217398723e-05,
+      "loss": 1.2695,
+      "step": 9368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6419138550027916e-05,
+      "loss": 0.9644,
+      "step": 9369
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6418343813705036e-05,
+      "loss": 0.9307,
+      "step": 9370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641754900843862e-05,
+      "loss": 1.5361,
+      "step": 9371
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641675413423721e-05,
+      "loss": 0.9912,
+      "step": 9372
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6415959191109347e-05,
+      "loss": 1.2822,
+      "step": 9373
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641516417906356e-05,
+      "loss": 1.4619,
+      "step": 9374
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6414369098108396e-05,
+      "loss": 1.3242,
+      "step": 9375
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6413573948252393e-05,
+      "loss": 1.0234,
+      "step": 9376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6412778729504093e-05,
+      "loss": 1.0117,
+      "step": 9377
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6411983441872035e-05,
+      "loss": 1.3643,
+      "step": 9378
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641118808536477e-05,
+      "loss": 1.249,
+      "step": 9379
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6410392659990833e-05,
+      "loss": 0.7705,
+      "step": 9380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640959716575877e-05,
+      "loss": 1.0107,
+      "step": 9381
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6408801602677126e-05,
+      "loss": 1.0918,
+      "step": 9382
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6408005970754452e-05,
+      "loss": 1.2031,
+      "step": 9383
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6407210269999282e-05,
+      "loss": 1.1597,
+      "step": 9384
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640641450042018e-05,
+      "loss": 0.9932,
+      "step": 9385
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6405618662025678e-05,
+      "loss": 0.7959,
+      "step": 9386
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6404822754824337e-05,
+      "loss": 1.2031,
+      "step": 9387
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6404026778824696e-05,
+      "loss": 1.1113,
+      "step": 9388
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6403230734035314e-05,
+      "loss": 1.0349,
+      "step": 9389
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6402434620464736e-05,
+      "loss": 1.2471,
+      "step": 9390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6401638438121517e-05,
+      "loss": 1.1797,
+      "step": 9391
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6400842187014204e-05,
+      "loss": 1.2998,
+      "step": 9392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6400045867151357e-05,
+      "loss": 1.4014,
+      "step": 9393
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6399249478541527e-05,
+      "loss": 1.124,
+      "step": 9394
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6398453021193263e-05,
+      "loss": 0.8711,
+      "step": 9395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6397656495115132e-05,
+      "loss": 1.1758,
+      "step": 9396
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6396859900315677e-05,
+      "loss": 1.0576,
+      "step": 9397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6396063236803465e-05,
+      "loss": 1.2324,
+      "step": 9398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639526650458705e-05,
+      "loss": 1.1348,
+      "step": 9399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6394469703674988e-05,
+      "loss": 0.9629,
+      "step": 9400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639367283407584e-05,
+      "loss": 1.1201,
+      "step": 9401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639287589579816e-05,
+      "loss": 0.9189,
+      "step": 9402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639207888885052e-05,
+      "loss": 1.0693,
+      "step": 9403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6391281813241474e-05,
+      "loss": 1.3076,
+      "step": 9404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6390484668979586e-05,
+      "loss": 0.8906,
+      "step": 9405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6389687456073416e-05,
+      "loss": 1.0391,
+      "step": 9406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6388890174531526e-05,
+      "loss": 1.3945,
+      "step": 9407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6388092824362488e-05,
+      "loss": 1.1348,
+      "step": 9408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6387295405574857e-05,
+      "loss": 1.2617,
+      "step": 9409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6386497918177208e-05,
+      "loss": 1.085,
+      "step": 9410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63857003621781e-05,
+      "loss": 0.8682,
+      "step": 9411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63849027375861e-05,
+      "loss": 1.2803,
+      "step": 9412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6384105044409783e-05,
+      "loss": 0.8984,
+      "step": 9413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6383307282657712e-05,
+      "loss": 1.0371,
+      "step": 9414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6382509452338456e-05,
+      "loss": 1.0352,
+      "step": 9415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6381711553460587e-05,
+      "loss": 1.5996,
+      "step": 9416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6380913586032677e-05,
+      "loss": 1.2842,
+      "step": 9417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6380115550063294e-05,
+      "loss": 1.3384,
+      "step": 9418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637931744556101e-05,
+      "loss": 0.9463,
+      "step": 9419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6378519272534404e-05,
+      "loss": 1.4375,
+      "step": 9420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637772103099204e-05,
+      "loss": 1.4277,
+      "step": 9421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63769227209425e-05,
+      "loss": 1.2275,
+      "step": 9422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637612434239436e-05,
+      "loss": 1.2715,
+      "step": 9423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6375325895356188e-05,
+      "loss": 1.0762,
+      "step": 9424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6374527379836568e-05,
+      "loss": 1.2285,
+      "step": 9425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637372879584408e-05,
+      "loss": 1.1094,
+      "step": 9426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372930143387288e-05,
+      "loss": 1.0254,
+      "step": 9427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372131422474782e-05,
+      "loss": 1.3086,
+      "step": 9428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6371332633115144e-05,
+      "loss": 1.082,
+      "step": 9429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6370533775316943e-05,
+      "loss": 1.29,
+      "step": 9430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636973484908877e-05,
+      "loss": 1.1689,
+      "step": 9431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6368935854439208e-05,
+      "loss": 0.9653,
+      "step": 9432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6368136791376827e-05,
+      "loss": 1.1025,
+      "step": 9433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6367337659910223e-05,
+      "loss": 0.9229,
+      "step": 9434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6366538460047975e-05,
+      "loss": 1.1357,
+      "step": 9435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636573919179867e-05,
+      "loss": 0.7305,
+      "step": 9436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364939855170887e-05,
+      "loss": 1.2656,
+      "step": 9437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364140450173218e-05,
+      "loss": 1.4375,
+      "step": 9438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6363340976814247e-05,
+      "loss": 0.9883,
+      "step": 9439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6362541435102566e-05,
+      "loss": 1.2031,
+      "step": 9440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636174182504676e-05,
+      "loss": 1.2617,
+      "step": 9441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636094214665542e-05,
+      "loss": 1.2637,
+      "step": 9442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636014239993713e-05,
+      "loss": 0.9434,
+      "step": 9443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6359342584900488e-05,
+      "loss": 1.1138,
+      "step": 9444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6358542701554084e-05,
+      "loss": 0.7881,
+      "step": 9445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6357742749906504e-05,
+      "loss": 1.1504,
+      "step": 9446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635694272996635e-05,
+      "loss": 0.9648,
+      "step": 9447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635614264174221e-05,
+      "loss": 0.6958,
+      "step": 9448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635534248524268e-05,
+      "loss": 1.0732,
+      "step": 9449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635454226047635e-05,
+      "loss": 1.4531,
+      "step": 9450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635374196745182e-05,
+      "loss": 0.7842,
+      "step": 9451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635294160617769e-05,
+      "loss": 0.875,
+      "step": 9452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635214117666255e-05,
+      "loss": 1.2822,
+      "step": 9453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6351340678915003e-05,
+      "loss": 1.168,
+      "step": 9454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635054011294365e-05,
+      "loss": 1.1367,
+      "step": 9455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634973947875708e-05,
+      "loss": 1.6162,
+      "step": 9456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6348938776363903e-05,
+      "loss": 1.4609,
+      "step": 9457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6348138005772717e-05,
+      "loss": 1.625,
+      "step": 9458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634733716699212e-05,
+      "loss": 1.2832,
+      "step": 9459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6346536260030718e-05,
+      "loss": 1.6719,
+      "step": 9460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6345735284897113e-05,
+      "loss": 1.29,
+      "step": 9461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6344934241599912e-05,
+      "loss": 0.8682,
+      "step": 9462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6344133130147715e-05,
+      "loss": 1.1079,
+      "step": 9463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634333195054913e-05,
+      "loss": 1.2725,
+      "step": 9464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634253070281276e-05,
+      "loss": 0.8535,
+      "step": 9465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6341729386947214e-05,
+      "loss": 1.1738,
+      "step": 9466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63409280029611e-05,
+      "loss": 0.9238,
+      "step": 9467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6340126550863026e-05,
+      "loss": 1.0225,
+      "step": 9468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6339325030661596e-05,
+      "loss": 1.5098,
+      "step": 9469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633852344236543e-05,
+      "loss": 1.0635,
+      "step": 9470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633772178598313e-05,
+      "loss": 1.2891,
+      "step": 9471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6336920061523307e-05,
+      "loss": 1.0498,
+      "step": 9472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6336118268994578e-05,
+      "loss": 0.9946,
+      "step": 9473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6335316408405548e-05,
+      "loss": 0.9912,
+      "step": 9474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633451447976484e-05,
+      "loss": 1.2236,
+      "step": 9475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6333712483081062e-05,
+      "loss": 1.3418,
+      "step": 9476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633291041836283e-05,
+      "loss": 1.2705,
+      "step": 9477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633210828561876e-05,
+      "loss": 1.1348,
+      "step": 9478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6331306084857462e-05,
+      "loss": 1.1895,
+      "step": 9479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633050381608756e-05,
+      "loss": 1.0791,
+      "step": 9480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6329701479317676e-05,
+      "loss": 0.9912,
+      "step": 9481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6328899074556415e-05,
+      "loss": 1.2266,
+      "step": 9482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632809660181241e-05,
+      "loss": 1.126,
+      "step": 9483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632729406109427e-05,
+      "loss": 0.9092,
+      "step": 9484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6326491452410623e-05,
+      "loss": 1.0254,
+      "step": 9485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6325688775770086e-05,
+      "loss": 1.0146,
+      "step": 9486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632488603118128e-05,
+      "loss": 1.0474,
+      "step": 9487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6324083218652836e-05,
+      "loss": 1.1416,
+      "step": 9488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632328033819337e-05,
+      "loss": 0.8374,
+      "step": 9489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6322477389811507e-05,
+      "loss": 0.916,
+      "step": 9490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6321674373515874e-05,
+      "loss": 0.7124,
+      "step": 9491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6320871289315093e-05,
+      "loss": 0.8506,
+      "step": 9492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6320068137217798e-05,
+      "loss": 1.0342,
+      "step": 9493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6319264917232608e-05,
+      "loss": 1.2969,
+      "step": 9494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6318461629368153e-05,
+      "loss": 1.4287,
+      "step": 9495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6317658273633065e-05,
+      "loss": 0.7456,
+      "step": 9496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631685485003597e-05,
+      "loss": 0.9863,
+      "step": 9497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6316051358585503e-05,
+      "loss": 1.3652,
+      "step": 9498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631524779929029e-05,
+      "loss": 1.0723,
+      "step": 9499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631444417215896e-05,
+      "loss": 1.0845,
+      "step": 9500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6313640477200153e-05,
+      "loss": 1.2227,
+      "step": 9501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6312836714422498e-05,
+      "loss": 0.9175,
+      "step": 9502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6312032883834628e-05,
+      "loss": 1.2168,
+      "step": 9503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6311228985445175e-05,
+      "loss": 1.5879,
+      "step": 9504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631042501926278e-05,
+      "loss": 1.0273,
+      "step": 9505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6309620985296075e-05,
+      "loss": 1.5352,
+      "step": 9506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63088168835537e-05,
+      "loss": 1.0547,
+      "step": 9507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630801271404429e-05,
+      "loss": 1.0605,
+      "step": 9508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6307208476776483e-05,
+      "loss": 1.0713,
+      "step": 9509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630640417175892e-05,
+      "loss": 1.2686,
+      "step": 9510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6305599799000235e-05,
+      "loss": 1.1367,
+      "step": 9511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6304795358509077e-05,
+      "loss": 1.25,
+      "step": 9512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630399085029408e-05,
+      "loss": 1.3438,
+      "step": 9513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6303186274363884e-05,
+      "loss": 1.3057,
+      "step": 9514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630238163072714e-05,
+      "loss": 1.0903,
+      "step": 9515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6301576919392483e-05,
+      "loss": 1.2715,
+      "step": 9516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630077214036856e-05,
+      "loss": 0.8779,
+      "step": 9517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629996729366402e-05,
+      "loss": 0.9419,
+      "step": 9518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6299162379287504e-05,
+      "loss": 1.3525,
+      "step": 9519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6298357397247656e-05,
+      "loss": 1.2656,
+      "step": 9520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6297552347553126e-05,
+      "loss": 1.1299,
+      "step": 9521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629674723021256e-05,
+      "loss": 1.1582,
+      "step": 9522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6295942045234608e-05,
+      "loss": 1.0791,
+      "step": 9523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629513679262792e-05,
+      "loss": 1.0391,
+      "step": 9524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6294331472401143e-05,
+      "loss": 1.3691,
+      "step": 9525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6293526084562928e-05,
+      "loss": 1.3906,
+      "step": 9526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6292720629121926e-05,
+      "loss": 1.0703,
+      "step": 9527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6291915106086787e-05,
+      "loss": 1.2402,
+      "step": 9528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629110951546617e-05,
+      "loss": 1.2227,
+      "step": 9529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6290303857268724e-05,
+      "loss": 0.917,
+      "step": 9530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6289498131503103e-05,
+      "loss": 1.0645,
+      "step": 9531
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6288692338177962e-05,
+      "loss": 1.0771,
+      "step": 9532
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628788647730196e-05,
+      "loss": 1.377,
+      "step": 9533
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6287080548883744e-05,
+      "loss": 1.0625,
+      "step": 9534
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628627455293198e-05,
+      "loss": 1.0605,
+      "step": 9535
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6285468489455326e-05,
+      "loss": 1.3828,
+      "step": 9536
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6284662358462435e-05,
+      "loss": 1.166,
+      "step": 9537
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6283856159961967e-05,
+      "loss": 0.957,
+      "step": 9538
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6283049893962588e-05,
+      "loss": 1.2891,
+      "step": 9539
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628224356047295e-05,
+      "loss": 0.9365,
+      "step": 9540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628143715950172e-05,
+      "loss": 0.8926,
+      "step": 9541
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6280630691057556e-05,
+      "loss": 0.6421,
+      "step": 9542
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6279824155149126e-05,
+      "loss": 1.3105,
+      "step": 9543
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627901755178509e-05,
+      "loss": 0.9102,
+      "step": 9544
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6278210880974113e-05,
+      "loss": 1.2275,
+      "step": 9545
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6277404142724854e-05,
+      "loss": 0.6719,
+      "step": 9546
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6276597337045995e-05,
+      "loss": 0.9443,
+      "step": 9547
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6275790463946187e-05,
+      "loss": 1.6172,
+      "step": 9548
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6274983523434102e-05,
+      "loss": 1.4375,
+      "step": 9549
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6274176515518406e-05,
+      "loss": 0.9287,
+      "step": 9550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6273369440207774e-05,
+      "loss": 1.1689,
+      "step": 9551
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6272562297510867e-05,
+      "loss": 1.04,
+      "step": 9552
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6271755087436364e-05,
+      "loss": 0.9863,
+      "step": 9553
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6270947809992924e-05,
+      "loss": 0.9941,
+      "step": 9554
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627014046518923e-05,
+      "loss": 1.4277,
+      "step": 9555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626933305303395e-05,
+      "loss": 1.1641,
+      "step": 9556
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6268525573535753e-05,
+      "loss": 0.999,
+      "step": 9557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626771802670332e-05,
+      "loss": 1.2334,
+      "step": 9558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626691041254532e-05,
+      "loss": 1.0947,
+      "step": 9559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626610273107043e-05,
+      "loss": 1.2014,
+      "step": 9560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6265294982287323e-05,
+      "loss": 0.8604,
+      "step": 9561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6264487166204677e-05,
+      "loss": 1.1318,
+      "step": 9562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263679282831173e-05,
+      "loss": 0.8496,
+      "step": 9563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6262871332175487e-05,
+      "loss": 0.8691,
+      "step": 9564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6262063314246293e-05,
+      "loss": 0.936,
+      "step": 9565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6261255229052278e-05,
+      "loss": 0.9985,
+      "step": 9566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6260447076602115e-05,
+      "loss": 1.0166,
+      "step": 9567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6259638856904493e-05,
+      "loss": 0.979,
+      "step": 9568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6258830569968084e-05,
+      "loss": 1.1582,
+      "step": 9569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6258022215801575e-05,
+      "loss": 0.8169,
+      "step": 9570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6257213794413654e-05,
+      "loss": 1.2061,
+      "step": 9571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6256405305812998e-05,
+      "loss": 1.0186,
+      "step": 9572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625559675000829e-05,
+      "loss": 1.3477,
+      "step": 9573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6254788127008226e-05,
+      "loss": 1.0693,
+      "step": 9574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6253979436821477e-05,
+      "loss": 0.9375,
+      "step": 9575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6253170679456745e-05,
+      "loss": 1.2354,
+      "step": 9576
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6252361854922704e-05,
+      "loss": 1.0928,
+      "step": 9577
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625155296322805e-05,
+      "loss": 0.9883,
+      "step": 9578
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6250744004381468e-05,
+      "loss": 1.6855,
+      "step": 9579
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624993497839165e-05,
+      "loss": 1.1797,
+      "step": 9580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624912588526729e-05,
+      "loss": 1.0425,
+      "step": 9581
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6248316725017067e-05,
+      "loss": 0.9443,
+      "step": 9582
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6247507497649687e-05,
+      "loss": 1.3994,
+      "step": 9583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624669820317383e-05,
+      "loss": 1.3027,
+      "step": 9584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6245888841598198e-05,
+      "loss": 1.0264,
+      "step": 9585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6245079412931483e-05,
+      "loss": 0.9824,
+      "step": 9586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6244269917182374e-05,
+      "loss": 1.2803,
+      "step": 9587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6243460354359573e-05,
+      "loss": 1.2285,
+      "step": 9588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6242650724471772e-05,
+      "loss": 1.4609,
+      "step": 9589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624184102752767e-05,
+      "loss": 1.5293,
+      "step": 9590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6241031263535966e-05,
+      "loss": 1.2871,
+      "step": 9591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6240221432505355e-05,
+      "loss": 1.4707,
+      "step": 9592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6239411534444534e-05,
+      "loss": 1.0205,
+      "step": 9593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623860156936221e-05,
+      "loss": 0.8369,
+      "step": 9594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6237791537267077e-05,
+      "loss": 1.0205,
+      "step": 9595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6236981438167835e-05,
+      "loss": 0.8242,
+      "step": 9596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6236171272073193e-05,
+      "loss": 1.2637,
+      "step": 9597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6235361038991848e-05,
+      "loss": 1.0205,
+      "step": 9598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62345507389325e-05,
+      "loss": 0.9951,
+      "step": 9599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6233740371903866e-05,
+      "loss": 1.0156,
+      "step": 9600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6232929937914637e-05,
+      "loss": 1.127,
+      "step": 9601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6232119436973525e-05,
+      "loss": 1.4688,
+      "step": 9602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623130886908923e-05,
+      "loss": 0.918,
+      "step": 9603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623049823427047e-05,
+      "loss": 1.0713,
+      "step": 9604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622968753252594e-05,
+      "loss": 0.9814,
+      "step": 9605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622887676386436e-05,
+      "loss": 1.1045,
+      "step": 9606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622806592829443e-05,
+      "loss": 1.1113,
+      "step": 9607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6227255025824866e-05,
+      "loss": 0.7051,
+      "step": 9608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622644405646437e-05,
+      "loss": 1.3857,
+      "step": 9609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622563302022166e-05,
+      "loss": 1.1318,
+      "step": 9610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622482191710545e-05,
+      "loss": 1.5508,
+      "step": 9611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6224010747124445e-05,
+      "loss": 1.5078,
+      "step": 9612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6223199510287367e-05,
+      "loss": 0.6504,
+      "step": 9613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6222388206602923e-05,
+      "loss": 0.8887,
+      "step": 9614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6221576836079827e-05,
+      "loss": 1.125,
+      "step": 9615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6220765398726797e-05,
+      "loss": 0.8916,
+      "step": 9616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219953894552555e-05,
+      "loss": 1.022,
+      "step": 9617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621914232356581e-05,
+      "loss": 1.0244,
+      "step": 9618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6218330685775278e-05,
+      "loss": 1.1895,
+      "step": 9619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621751898118969e-05,
+      "loss": 1.1152,
+      "step": 9620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6216707209817748e-05,
+      "loss": 0.939,
+      "step": 9621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6215895371668186e-05,
+      "loss": 1.668,
+      "step": 9622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6215083466749716e-05,
+      "loss": 1.1895,
+      "step": 9623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621427149507106e-05,
+      "loss": 1.2617,
+      "step": 9624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6213459456640945e-05,
+      "loss": 0.9531,
+      "step": 9625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6212647351468088e-05,
+      "loss": 0.9941,
+      "step": 9626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621183517956122e-05,
+      "loss": 1.3232,
+      "step": 9627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6211022940929055e-05,
+      "loss": 1.1211,
+      "step": 9628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621021063558032e-05,
+      "loss": 1.3477,
+      "step": 9629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6209398263523746e-05,
+      "loss": 0.9346,
+      "step": 9630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620858582476806e-05,
+      "loss": 1.8359,
+      "step": 9631
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620777331932198e-05,
+      "loss": 1.209,
+      "step": 9632
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6206960747194244e-05,
+      "loss": 1.1172,
+      "step": 9633
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6206148108393568e-05,
+      "loss": 1.1533,
+      "step": 9634
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6205335402928692e-05,
+      "loss": 1.376,
+      "step": 9635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6204522630808344e-05,
+      "loss": 1.1406,
+      "step": 9636
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6203709792041247e-05,
+      "loss": 1.1709,
+      "step": 9637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6202896886636145e-05,
+      "loss": 0.8369,
+      "step": 9638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6202083914601757e-05,
+      "loss": 0.915,
+      "step": 9639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6201270875946825e-05,
+      "loss": 1.4854,
+      "step": 9640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6200457770680077e-05,
+      "loss": 1.1777,
+      "step": 9641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619964459881025e-05,
+      "loss": 1.2988,
+      "step": 9642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6198831360346077e-05,
+      "loss": 1.3145,
+      "step": 9643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6198018055296295e-05,
+      "loss": 1.3789,
+      "step": 9644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6197204683669635e-05,
+      "loss": 1.0083,
+      "step": 9645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6196391245474842e-05,
+      "loss": 0.748,
+      "step": 9646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619557774072065e-05,
+      "loss": 1.0898,
+      "step": 9647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6194764169415796e-05,
+      "loss": 0.7329,
+      "step": 9648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6193950531569022e-05,
+      "loss": 1.4453,
+      "step": 9649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6193136827189067e-05,
+      "loss": 1.2705,
+      "step": 9650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619232305628467e-05,
+      "loss": 1.6758,
+      "step": 9651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619150921886457e-05,
+      "loss": 1.1094,
+      "step": 9652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6190695314937516e-05,
+      "loss": 1.0361,
+      "step": 9653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6189881344512244e-05,
+      "loss": 1.0791,
+      "step": 9654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6189067307597498e-05,
+      "loss": 0.958,
+      "step": 9655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6188253204202026e-05,
+      "loss": 0.8516,
+      "step": 9656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6187439034334572e-05,
+      "loss": 1.2373,
+      "step": 9657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618662479800388e-05,
+      "loss": 1.1543,
+      "step": 9658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6185810495218698e-05,
+      "loss": 1.1094,
+      "step": 9659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618499612598777e-05,
+      "loss": 0.9336,
+      "step": 9660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6184181690319846e-05,
+      "loss": 1.0518,
+      "step": 9661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618336718822367e-05,
+      "loss": 1.0156,
+      "step": 9662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6182552619708e-05,
+      "loss": 1.3643,
+      "step": 9663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6181737984781574e-05,
+      "loss": 0.9692,
+      "step": 9664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6180923283453154e-05,
+      "loss": 1.3496,
+      "step": 9665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6180108515731487e-05,
+      "loss": 1.0664,
+      "step": 9666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6179293681625323e-05,
+      "loss": 1.3789,
+      "step": 9667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6178478781143414e-05,
+      "loss": 1.1926,
+      "step": 9668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617766381429452e-05,
+      "loss": 0.9189,
+      "step": 9669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6176848781087386e-05,
+      "loss": 0.9331,
+      "step": 9670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617603368153077e-05,
+      "loss": 0.8613,
+      "step": 9671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6175218515633427e-05,
+      "loss": 1.2578,
+      "step": 9672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617440328340412e-05,
+      "loss": 1.0474,
+      "step": 9673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6173587984851596e-05,
+      "loss": 0.8867,
+      "step": 9674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6172772619984623e-05,
+      "loss": 1.0679,
+      "step": 9675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6171957188811948e-05,
+      "loss": 1.1069,
+      "step": 9676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617114169134234e-05,
+      "loss": 1.5918,
+      "step": 9677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6170326127584552e-05,
+      "loss": 1.1992,
+      "step": 9678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616951049754735e-05,
+      "loss": 0.738,
+      "step": 9679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616869480123949e-05,
+      "loss": 0.9854,
+      "step": 9680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6167879038669735e-05,
+      "loss": 0.9751,
+      "step": 9681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616706320984685e-05,
+      "loss": 1.3291,
+      "step": 9682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6166247314779598e-05,
+      "loss": 1.082,
+      "step": 9683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6165431353476743e-05,
+      "loss": 1.0142,
+      "step": 9684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6164615325947045e-05,
+      "loss": 0.7043,
+      "step": 9685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6163799232199276e-05,
+      "loss": 1.0186,
+      "step": 9686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61629830722422e-05,
+      "loss": 1.0728,
+      "step": 9687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6162166846084585e-05,
+      "loss": 1.2266,
+      "step": 9688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6161350553735196e-05,
+      "loss": 1.0361,
+      "step": 9689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61605341952028e-05,
+      "loss": 1.1855,
+      "step": 9690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615971777049617e-05,
+      "loss": 1.1016,
+      "step": 9691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6158901279624073e-05,
+      "loss": 1.0293,
+      "step": 9692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6158084722595283e-05,
+      "loss": 0.979,
+      "step": 9693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615726809941857e-05,
+      "loss": 1.2939,
+      "step": 9694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61564514101027e-05,
+      "loss": 1.4238,
+      "step": 9695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6155634654656452e-05,
+      "loss": 0.8652,
+      "step": 9696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6154817833088596e-05,
+      "loss": 0.9043,
+      "step": 9697
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6154000945407914e-05,
+      "loss": 1.3525,
+      "step": 9698
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6153183991623165e-05,
+      "loss": 1.4971,
+      "step": 9699
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6152366971743145e-05,
+      "loss": 0.8027,
+      "step": 9700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615154988577661e-05,
+      "loss": 1.207,
+      "step": 9701
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6150732733732345e-05,
+      "loss": 0.998,
+      "step": 9702
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6149915515619133e-05,
+      "loss": 0.999,
+      "step": 9703
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6149098231445744e-05,
+      "loss": 0.7183,
+      "step": 9704
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148280881220967e-05,
+      "loss": 1.0068,
+      "step": 9705
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614746346495357e-05,
+      "loss": 0.6895,
+      "step": 9706
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146645982652336e-05,
+      "loss": 0.9004,
+      "step": 9707
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614582843432605e-05,
+      "loss": 1.0371,
+      "step": 9708
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614501081998349e-05,
+      "loss": 1.2891,
+      "step": 9709
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614419313963345e-05,
+      "loss": 1.0371,
+      "step": 9710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61433753932847e-05,
+      "loss": 1.0195,
+      "step": 9711
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6142557580946027e-05,
+      "loss": 1.1982,
+      "step": 9712
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6141739702626215e-05,
+      "loss": 1.252,
+      "step": 9713
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6140921758334053e-05,
+      "loss": 1.1807,
+      "step": 9714
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6140103748078323e-05,
+      "loss": 1.0928,
+      "step": 9715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613928567186782e-05,
+      "loss": 1.2051,
+      "step": 9716
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613846752971132e-05,
+      "loss": 1.2861,
+      "step": 9717
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6137649321617617e-05,
+      "loss": 1.2783,
+      "step": 9718
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61368310475955e-05,
+      "loss": 1.2939,
+      "step": 9719
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613601270765376e-05,
+      "loss": 1.0625,
+      "step": 9720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6135194301801184e-05,
+      "loss": 0.8945,
+      "step": 9721
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6134375830046566e-05,
+      "loss": 1.3223,
+      "step": 9722
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6133557292398694e-05,
+      "loss": 0.8887,
+      "step": 9723
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6132738688866363e-05,
+      "loss": 1.2188,
+      "step": 9724
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6131920019458368e-05,
+      "loss": 1.5889,
+      "step": 9725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6131101284183496e-05,
+      "loss": 1.1689,
+      "step": 9726
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6130282483050548e-05,
+      "loss": 1.1182,
+      "step": 9727
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612946361606832e-05,
+      "loss": 1.0928,
+      "step": 9728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6128644683245602e-05,
+      "loss": 1.1587,
+      "step": 9729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6127825684591195e-05,
+      "loss": 0.9902,
+      "step": 9730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6127006620113898e-05,
+      "loss": 1.1504,
+      "step": 9731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6126187489822503e-05,
+      "loss": 0.9551,
+      "step": 9732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6125368293725814e-05,
+      "loss": 0.9111,
+      "step": 9733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6124549031832632e-05,
+      "loss": 1.4199,
+      "step": 9734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612372970415175e-05,
+      "loss": 0.8203,
+      "step": 9735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6122910310691975e-05,
+      "loss": 0.7573,
+      "step": 9736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6122090851462108e-05,
+      "loss": 0.9395,
+      "step": 9737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6121271326470945e-05,
+      "loss": 1.0508,
+      "step": 9738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6120451735727294e-05,
+      "loss": 1.1025,
+      "step": 9739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611963207923996e-05,
+      "loss": 1.1475,
+      "step": 9740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611881235701775e-05,
+      "loss": 1.6094,
+      "step": 9741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117992569069465e-05,
+      "loss": 1.376,
+      "step": 9742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117172715403908e-05,
+      "loss": 0.9668,
+      "step": 9743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611635279602989e-05,
+      "loss": 1.2266,
+      "step": 9744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6115532810956218e-05,
+      "loss": 1.0205,
+      "step": 9745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6114712760191697e-05,
+      "loss": 0.9688,
+      "step": 9746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6113892643745138e-05,
+      "loss": 1.4961,
+      "step": 9747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6113072461625353e-05,
+      "loss": 1.3125,
+      "step": 9748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611225221384115e-05,
+      "loss": 0.8301,
+      "step": 9749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6111431900401333e-05,
+      "loss": 1.335,
+      "step": 9750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6110611521314723e-05,
+      "loss": 1.3867,
+      "step": 9751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610979107659013e-05,
+      "loss": 1.1611,
+      "step": 9752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6108970566236364e-05,
+      "loss": 0.7231,
+      "step": 9753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610814999026224e-05,
+      "loss": 0.7764,
+      "step": 9754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6107329348676576e-05,
+      "loss": 1.3682,
+      "step": 9755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6106508641488183e-05,
+      "loss": 1.0322,
+      "step": 9756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6105687868705873e-05,
+      "loss": 1.0488,
+      "step": 9757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610486703033847e-05,
+      "loss": 0.9629,
+      "step": 9758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6104046126394788e-05,
+      "loss": 0.8489,
+      "step": 9759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6103225156883645e-05,
+      "loss": 1.0947,
+      "step": 9760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6102404121813856e-05,
+      "loss": 1.4023,
+      "step": 9761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610158302119425e-05,
+      "loss": 1.0918,
+      "step": 9762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6100761855033638e-05,
+      "loss": 1.2412,
+      "step": 9763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609994062334084e-05,
+      "loss": 1.2329,
+      "step": 9764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6099119326124685e-05,
+      "loss": 0.5488,
+      "step": 9765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609829796339399e-05,
+      "loss": 1.0342,
+      "step": 9766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609747653515758e-05,
+      "loss": 1.1104,
+      "step": 9767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6096655041424273e-05,
+      "loss": 1.207,
+      "step": 9768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6095833482202897e-05,
+      "loss": 1.25,
+      "step": 9769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6095011857502282e-05,
+      "loss": 1.4814,
+      "step": 9770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609419016733125e-05,
+      "loss": 1.2197,
+      "step": 9771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609336841169862e-05,
+      "loss": 1.0967,
+      "step": 9772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6092546590613226e-05,
+      "loss": 1.3379,
+      "step": 9773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60917247040839e-05,
+      "loss": 1.0869,
+      "step": 9774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609090275211946e-05,
+      "loss": 1.2217,
+      "step": 9775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609008073472874e-05,
+      "loss": 1.375,
+      "step": 9776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608925865192057e-05,
+      "loss": 1.1396,
+      "step": 9777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6088436503703784e-05,
+      "loss": 1.2529,
+      "step": 9778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608761429008721e-05,
+      "loss": 1.1104,
+      "step": 9779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6086792011079677e-05,
+      "loss": 0.9521,
+      "step": 9780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608596966669002e-05,
+      "loss": 1.1089,
+      "step": 9781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6085147256927073e-05,
+      "loss": 0.8682,
+      "step": 9782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6084324781799674e-05,
+      "loss": 1.3311,
+      "step": 9783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608350224131665e-05,
+      "loss": 0.918,
+      "step": 9784
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6082679635486838e-05,
+      "loss": 1.1895,
+      "step": 9785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608185696431908e-05,
+      "loss": 1.0156,
+      "step": 9786
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6081034227822207e-05,
+      "loss": 1.293,
+      "step": 9787
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6080211426005055e-05,
+      "loss": 1.3047,
+      "step": 9788
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607938855887647e-05,
+      "loss": 1.2754,
+      "step": 9789
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607856562644529e-05,
+      "loss": 1.042,
+      "step": 9790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6077742628720344e-05,
+      "loss": 1.1147,
+      "step": 9791
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6076919565710484e-05,
+      "loss": 1.4785,
+      "step": 9792
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6076096437424544e-05,
+      "loss": 1.0781,
+      "step": 9793
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607527324387137e-05,
+      "loss": 1.5264,
+      "step": 9794
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6074449985059802e-05,
+      "loss": 0.958,
+      "step": 9795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6073626660998683e-05,
+      "loss": 1.0342,
+      "step": 9796
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607280327169686e-05,
+      "loss": 1.2354,
+      "step": 9797
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6071979817163175e-05,
+      "loss": 1.1846,
+      "step": 9798
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6071156297406474e-05,
+      "loss": 0.8682,
+      "step": 9799
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070332712435602e-05,
+      "loss": 1.1445,
+      "step": 9800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6069509062259406e-05,
+      "loss": 1.4648,
+      "step": 9801
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606868534688673e-05,
+      "loss": 1.0947,
+      "step": 9802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606786156632643e-05,
+      "loss": 0.8545,
+      "step": 9803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6067037720587352e-05,
+      "loss": 1.0415,
+      "step": 9804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6066213809678338e-05,
+      "loss": 0.6045,
+      "step": 9805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606538983360825e-05,
+      "loss": 0.9976,
+      "step": 9806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606456579238593e-05,
+      "loss": 1.1016,
+      "step": 9807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6063741686020233e-05,
+      "loss": 1.123,
+      "step": 9808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606291751452001e-05,
+      "loss": 1.2617,
+      "step": 9809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6062093277894116e-05,
+      "loss": 1.2773,
+      "step": 9810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6061268976151405e-05,
+      "loss": 1.0752,
+      "step": 9811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6060444609300726e-05,
+      "loss": 1.2021,
+      "step": 9812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6059620177350938e-05,
+      "loss": 0.8047,
+      "step": 9813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6058795680310896e-05,
+      "loss": 1.0215,
+      "step": 9814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6057971118189457e-05,
+      "loss": 0.9316,
+      "step": 9815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605714649099548e-05,
+      "loss": 0.9795,
+      "step": 9816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6056321798737817e-05,
+      "loss": 1.1191,
+      "step": 9817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055497041425333e-05,
+      "loss": 0.9951,
+      "step": 9818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605467221906689e-05,
+      "loss": 1.3359,
+      "step": 9819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6053847331671333e-05,
+      "loss": 0.7944,
+      "step": 9820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6053022379247537e-05,
+      "loss": 1.0049,
+      "step": 9821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6052197361804357e-05,
+      "loss": 1.1943,
+      "step": 9822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605137227935066e-05,
+      "loss": 1.3975,
+      "step": 9823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6050547131895302e-05,
+      "loss": 0.9395,
+      "step": 9824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6049721919447153e-05,
+      "loss": 1.147,
+      "step": 9825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604889664201507e-05,
+      "loss": 0.895,
+      "step": 9826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6048071299607927e-05,
+      "loss": 1.04,
+      "step": 9827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6047245892234577e-05,
+      "loss": 1.3525,
+      "step": 9828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60464204199039e-05,
+      "loss": 1.2266,
+      "step": 9829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604559488262475e-05,
+      "loss": 0.9902,
+      "step": 9830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6044769280406006e-05,
+      "loss": 1.1611,
+      "step": 9831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604394361325653e-05,
+      "loss": 1.2061,
+      "step": 9832
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6043117881185192e-05,
+      "loss": 0.5601,
+      "step": 9833
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6042292084200864e-05,
+      "loss": 1.2207,
+      "step": 9834
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6041466222312416e-05,
+      "loss": 1.1865,
+      "step": 9835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6040640295528714e-05,
+      "loss": 1.0327,
+      "step": 9836
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6039814303858636e-05,
+      "loss": 1.5029,
+      "step": 9837
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603898824731105e-05,
+      "loss": 1.0859,
+      "step": 9838
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6038162125894833e-05,
+      "loss": 1.293,
+      "step": 9839
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603733593961885e-05,
+      "loss": 1.3213,
+      "step": 9840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603650968849199e-05,
+      "loss": 1.0938,
+      "step": 9841
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603568337252312e-05,
+      "loss": 1.2695,
+      "step": 9842
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603485699172112e-05,
+      "loss": 0.9824,
+      "step": 9843
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6034030546094855e-05,
+      "loss": 1.334,
+      "step": 9844
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6033204035653218e-05,
+      "loss": 1.2744,
+      "step": 9845
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6032377460405077e-05,
+      "loss": 0.9424,
+      "step": 9846
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6031550820359315e-05,
+      "loss": 1.2861,
+      "step": 9847
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6030724115524812e-05,
+      "loss": 1.2754,
+      "step": 9848
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6029897345910443e-05,
+      "loss": 1.0684,
+      "step": 9849
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6029070511525094e-05,
+      "loss": 1.3076,
+      "step": 9850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6028243612377646e-05,
+      "loss": 1.0596,
+      "step": 9851
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6027416648476978e-05,
+      "loss": 1.1445,
+      "step": 9852
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6026589619831975e-05,
+      "loss": 1.0405,
+      "step": 9853
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6025762526451523e-05,
+      "loss": 1.2598,
+      "step": 9854
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6024935368344504e-05,
+      "loss": 1.1172,
+      "step": 9855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6024108145519803e-05,
+      "loss": 1.0811,
+      "step": 9856
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6023280857986305e-05,
+      "loss": 1.0078,
+      "step": 9857
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6022453505752896e-05,
+      "loss": 1.2637,
+      "step": 9858
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6021626088828467e-05,
+      "loss": 1.0645,
+      "step": 9859
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6020798607221903e-05,
+      "loss": 1.1299,
+      "step": 9860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601997106094209e-05,
+      "loss": 1.2275,
+      "step": 9861
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601914344999792e-05,
+      "loss": 1.0205,
+      "step": 9862
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6018315774398286e-05,
+      "loss": 1.1567,
+      "step": 9863
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601748803415207e-05,
+      "loss": 1.0996,
+      "step": 9864
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6016660229268176e-05,
+      "loss": 1.332,
+      "step": 9865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6015832359755483e-05,
+      "loss": 1.3027,
+      "step": 9866
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6015004425622895e-05,
+      "loss": 0.9058,
+      "step": 9867
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6014176426879294e-05,
+      "loss": 0.8818,
+      "step": 9868
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6013348363533584e-05,
+      "loss": 1.3564,
+      "step": 9869
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6012520235594653e-05,
+      "loss": 0.8618,
+      "step": 9870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6011692043071398e-05,
+      "loss": 1.1982,
+      "step": 9871
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601086378597272e-05,
+      "loss": 1.2344,
+      "step": 9872
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6010035464307507e-05,
+      "loss": 0.8848,
+      "step": 9873
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6009207078084666e-05,
+      "loss": 0.9492,
+      "step": 9874
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6008378627313086e-05,
+      "loss": 0.8857,
+      "step": 9875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6007550112001675e-05,
+      "loss": 1.0459,
+      "step": 9876
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6006721532159326e-05,
+      "loss": 1.5176,
+      "step": 9877
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6005892887794944e-05,
+      "loss": 1.2021,
+      "step": 9878
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6005064178917423e-05,
+      "loss": 1.0439,
+      "step": 9879
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6004235405535676e-05,
+      "loss": 0.9971,
+      "step": 9880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6003406567658594e-05,
+      "loss": 0.811,
+      "step": 9881
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6002577665295085e-05,
+      "loss": 1.3232,
+      "step": 9882
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6001748698454052e-05,
+      "loss": 0.9541,
+      "step": 9883
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.60009196671444e-05,
+      "loss": 0.8311,
+      "step": 9884
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600009057137504e-05,
+      "loss": 1.5371,
+      "step": 9885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5999261411154866e-05,
+      "loss": 1.2207,
+      "step": 9886
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5998432186492792e-05,
+      "loss": 1.1748,
+      "step": 9887
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5997602897397722e-05,
+      "loss": 1.0459,
+      "step": 9888
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5996773543878566e-05,
+      "loss": 1.3574,
+      "step": 9889
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5995944125944235e-05,
+      "loss": 0.79,
+      "step": 9890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5995114643603633e-05,
+      "loss": 0.7671,
+      "step": 9891
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5994285096865674e-05,
+      "loss": 1.3379,
+      "step": 9892
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5993455485739265e-05,
+      "loss": 1.1934,
+      "step": 9893
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599262581023332e-05,
+      "loss": 1.457,
+      "step": 9894
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5991796070356756e-05,
+      "loss": 1.1787,
+      "step": 9895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5990966266118474e-05,
+      "loss": 1.3438,
+      "step": 9896
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59901363975274e-05,
+      "loss": 1.0859,
+      "step": 9897
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5989306464592434e-05,
+      "loss": 1.2637,
+      "step": 9898
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5988476467322506e-05,
+      "loss": 0.9316,
+      "step": 9899
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5987646405726524e-05,
+      "loss": 1.0605,
+      "step": 9900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59868162798134e-05,
+      "loss": 1.0186,
+      "step": 9901
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598598608959206e-05,
+      "loss": 1.0273,
+      "step": 9902
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5985155835071417e-05,
+      "loss": 0.9814,
+      "step": 9903
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5984325516260385e-05,
+      "loss": 0.9795,
+      "step": 9904
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5983495133167894e-05,
+      "loss": 0.8276,
+      "step": 9905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982664685802852e-05,
+      "loss": 0.8584,
+      "step": 9906
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5981834174174185e-05,
+      "loss": 1.543,
+      "step": 9907
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5981003598290816e-05,
+      "loss": 1.1055,
+      "step": 9908
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5980172958161664e-05,
+      "loss": 1.0557,
+      "step": 9909
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5979342253795655e-05,
+      "loss": 0.9497,
+      "step": 9910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59785114852017e-05,
+      "loss": 1.334,
+      "step": 9911
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597768065238874e-05,
+      "loss": 1.0732,
+      "step": 9912
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5976849755365686e-05,
+      "loss": 1.0625,
+      "step": 9913
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597601879414147e-05,
+      "loss": 1.0605,
+      "step": 9914
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597518776872502e-05,
+      "loss": 1.1899,
+      "step": 9915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5974356679125254e-05,
+      "loss": 1.0117,
+      "step": 9916
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5973525525351105e-05,
+      "loss": 1.3613,
+      "step": 9917
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5972694307411503e-05,
+      "loss": 0.8125,
+      "step": 9918
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597186302531537e-05,
+      "loss": 1.5488,
+      "step": 9919
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597103167907164e-05,
+      "loss": 1.2451,
+      "step": 9920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597020026868924e-05,
+      "loss": 1.1377,
+      "step": 9921
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5969368794177108e-05,
+      "loss": 0.9492,
+      "step": 9922
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596853725554417e-05,
+      "loss": 1.3301,
+      "step": 9923
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5967705652799355e-05,
+      "loss": 1.1523,
+      "step": 9924
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59668739859516e-05,
+      "loss": 1.0195,
+      "step": 9925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5966042255009837e-05,
+      "loss": 0.9409,
+      "step": 9926
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5965210459983002e-05,
+      "loss": 0.9302,
+      "step": 9927
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5964378600880032e-05,
+      "loss": 1.5488,
+      "step": 9928
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596354667770985e-05,
+      "loss": 1.3574,
+      "step": 9929
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5962714690481408e-05,
+      "loss": 1.0908,
+      "step": 9930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5961882639203637e-05,
+      "loss": 1.1299,
+      "step": 9931
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596105052388547e-05,
+      "loss": 1.2168,
+      "step": 9932
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5960218344535848e-05,
+      "loss": 1.0332,
+      "step": 9933
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5959386101163713e-05,
+      "loss": 1.4297,
+      "step": 9934
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5958553793778005e-05,
+      "loss": 0.9541,
+      "step": 9935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595772142238766e-05,
+      "loss": 1.002,
+      "step": 9936
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5956888987001618e-05,
+      "loss": 0.9893,
+      "step": 9937
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5956056487628832e-05,
+      "loss": 0.9497,
+      "step": 9938
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5955223924278228e-05,
+      "loss": 1.0742,
+      "step": 9939
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5954391296958763e-05,
+      "loss": 0.8379,
+      "step": 9940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595355860567937e-05,
+      "loss": 1.1777,
+      "step": 9941
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5952725850449003e-05,
+      "loss": 1.2363,
+      "step": 9942
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59518930312766e-05,
+      "loss": 1.3398,
+      "step": 9943
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5951060148171112e-05,
+      "loss": 1.4531,
+      "step": 9944
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595022720114148e-05,
+      "loss": 0.9673,
+      "step": 9945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5949394190196658e-05,
+      "loss": 0.9072,
+      "step": 9946
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5948561115345587e-05,
+      "loss": 1.5078,
+      "step": 9947
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594772797659722e-05,
+      "loss": 1.3633,
+      "step": 9948
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5946894773960508e-05,
+      "loss": 0.6675,
+      "step": 9949
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594606150744439e-05,
+      "loss": 1.1133,
+      "step": 9950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5945228177057832e-05,
+      "loss": 0.8975,
+      "step": 9951
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5944394782809777e-05,
+      "loss": 0.8398,
+      "step": 9952
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5943561324709173e-05,
+      "loss": 0.9375,
+      "step": 9953
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5942727802764982e-05,
+      "loss": 0.6134,
+      "step": 9954
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5941894216986152e-05,
+      "loss": 1.2617,
+      "step": 9955
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594106056738164e-05,
+      "loss": 1.2012,
+      "step": 9956
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5940226853960397e-05,
+      "loss": 0.9741,
+      "step": 9957
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5939393076731383e-05,
+      "loss": 1.1348,
+      "step": 9958
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593855923570355e-05,
+      "loss": 1.0742,
+      "step": 9959
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5937725330885855e-05,
+      "loss": 1.1016,
+      "step": 9960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5936891362287256e-05,
+      "loss": 0.9004,
+      "step": 9961
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5936057329916718e-05,
+      "loss": 0.9541,
+      "step": 9962
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5935223233783186e-05,
+      "loss": 1.0486,
+      "step": 9963
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5934389073895633e-05,
+      "loss": 1.0918,
+      "step": 9964
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5933554850263013e-05,
+      "loss": 1.3496,
+      "step": 9965
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5932720562894287e-05,
+      "loss": 1.1914,
+      "step": 9966
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5931886211798414e-05,
+      "loss": 1.1602,
+      "step": 9967
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5931051796984362e-05,
+      "loss": 1.1602,
+      "step": 9968
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930217318461093e-05,
+      "loss": 1.3906,
+      "step": 9969
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592938277623757e-05,
+      "loss": 0.8457,
+      "step": 9970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5928548170322754e-05,
+      "loss": 1.0723,
+      "step": 9971
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5927713500725616e-05,
+      "loss": 1.082,
+      "step": 9972
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5926878767455115e-05,
+      "loss": 1.2822,
+      "step": 9973
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592604397052022e-05,
+      "loss": 0.8965,
+      "step": 9974
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5925209109929904e-05,
+      "loss": 0.8857,
+      "step": 9975
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5924374185693124e-05,
+      "loss": 1.1953,
+      "step": 9976
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5923539197818857e-05,
+      "loss": 1.0571,
+      "step": 9977
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5922704146316068e-05,
+      "loss": 0.8809,
+      "step": 9978
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592186903119373e-05,
+      "loss": 1.2109,
+      "step": 9979
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592103385246081e-05,
+      "loss": 0.7979,
+      "step": 9980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592019861012628e-05,
+      "loss": 1.1748,
+      "step": 9981
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5919363304199114e-05,
+      "loss": 1.3438,
+      "step": 9982
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591852793468828e-05,
+      "loss": 0.9365,
+      "step": 9983
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591769250160276e-05,
+      "loss": 1.0186,
+      "step": 9984
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5916857004951516e-05,
+      "loss": 0.9253,
+      "step": 9985
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5916021444743535e-05,
+      "loss": 0.9956,
+      "step": 9986
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591518582098778e-05,
+      "loss": 1.1123,
+      "step": 9987
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591435013369324e-05,
+      "loss": 0.9629,
+      "step": 9988
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591351438286888e-05,
+      "loss": 1.3486,
+      "step": 9989
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5912678568523684e-05,
+      "loss": 1.1426,
+      "step": 9990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5911842690666627e-05,
+      "loss": 1.1099,
+      "step": 9991
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591100674930669e-05,
+      "loss": 0.7637,
+      "step": 9992
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591017074445285e-05,
+      "loss": 1.1895,
+      "step": 9993
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590933467611409e-05,
+      "loss": 0.8037,
+      "step": 9994
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5908498544299387e-05,
+      "loss": 1.2656,
+      "step": 9995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5907662349017727e-05,
+      "loss": 1.1562,
+      "step": 9996
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5906826090278085e-05,
+      "loss": 1.1113,
+      "step": 9997
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5905989768089453e-05,
+      "loss": 0.9814,
+      "step": 9998
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5905153382460807e-05,
+      "loss": 1.29,
+      "step": 9999
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5904316933401136e-05,
+      "loss": 1.2773,
+      "step": 10000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5903480420919423e-05,
+      "loss": 1.2061,
+      "step": 10001
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5902643845024653e-05,
+      "loss": 1.0703,
+      "step": 10002
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5901807205725814e-05,
+      "loss": 1.1064,
+      "step": 10003
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5900970503031885e-05,
+      "loss": 1.0898,
+      "step": 10004
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5900133736951865e-05,
+      "loss": 1.0845,
+      "step": 10005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589929690749474e-05,
+      "loss": 0.9941,
+      "step": 10006
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589846001466949e-05,
+      "loss": 0.9678,
+      "step": 10007
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5897623058485116e-05,
+      "loss": 1.2383,
+      "step": 10008
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58967860389506e-05,
+      "loss": 1.6953,
+      "step": 10009
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895948956074937e-05,
+      "loss": 0.9229,
+      "step": 10010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895111809867113e-05,
+      "loss": 0.7749,
+      "step": 10011
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589427460033613e-05,
+      "loss": 1.0879,
+      "step": 10012
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5893437327490975e-05,
+      "loss": 1.2969,
+      "step": 10013
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5892599991340644e-05,
+      "loss": 1.0596,
+      "step": 10014
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589176259189413e-05,
+      "loss": 0.9502,
+      "step": 10015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5890925129160424e-05,
+      "loss": 1.498,
+      "step": 10016
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5890087603148526e-05,
+      "loss": 1.3926,
+      "step": 10017
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5889250013867437e-05,
+      "loss": 1.1514,
+      "step": 10018
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5888412361326147e-05,
+      "loss": 1.1836,
+      "step": 10019
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5887574645533654e-05,
+      "loss": 1.2568,
+      "step": 10020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588673686649896e-05,
+      "loss": 1.1816,
+      "step": 10021
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5885899024231065e-05,
+      "loss": 1.3613,
+      "step": 10022
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5885061118738964e-05,
+      "loss": 1.2041,
+      "step": 10023
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5884223150031662e-05,
+      "loss": 1.02,
+      "step": 10024
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5883385118118155e-05,
+      "loss": 0.9512,
+      "step": 10025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588254702300745e-05,
+      "loss": 1.1836,
+      "step": 10026
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5881708864708545e-05,
+      "loss": 1.0312,
+      "step": 10027
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880870643230447e-05,
+      "loss": 1.1621,
+      "step": 10028
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880032358582157e-05,
+      "loss": 1.4062,
+      "step": 10029
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5879194010772683e-05,
+      "loss": 0.8677,
+      "step": 10030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5878355599811027e-05,
+      "loss": 1.4062,
+      "step": 10031
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5877517125706193e-05,
+      "loss": 1.0791,
+      "step": 10032
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58766785884672e-05,
+      "loss": 0.9507,
+      "step": 10033
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587583998810304e-05,
+      "loss": 1.5615,
+      "step": 10034
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5875001324622723e-05,
+      "loss": 1.375,
+      "step": 10035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5874162598035267e-05,
+      "loss": 1.2275,
+      "step": 10036
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5873323808349676e-05,
+      "loss": 1.3652,
+      "step": 10037
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5872484955574955e-05,
+      "loss": 0.8184,
+      "step": 10038
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5871646039720124e-05,
+      "loss": 1.0498,
+      "step": 10039
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587080706079419e-05,
+      "loss": 1.4902,
+      "step": 10040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5869968018806162e-05,
+      "loss": 0.6978,
+      "step": 10041
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5869128913765056e-05,
+      "loss": 1.2158,
+      "step": 10042
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5868289745679883e-05,
+      "loss": 0.957,
+      "step": 10043
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5867450514559664e-05,
+      "loss": 1.2148,
+      "step": 10044
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5866611220413407e-05,
+      "loss": 1.2891,
+      "step": 10045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5865771863250125e-05,
+      "loss": 1.2676,
+      "step": 10046
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586493244307884e-05,
+      "loss": 1.4424,
+      "step": 10047
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586409295990857e-05,
+      "loss": 1.3555,
+      "step": 10048
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5863253413748326e-05,
+      "loss": 1.0977,
+      "step": 10049
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586241380460713e-05,
+      "loss": 1.2334,
+      "step": 10050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5861574132494e-05,
+      "loss": 0.9561,
+      "step": 10051
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5860734397417957e-05,
+      "loss": 0.9282,
+      "step": 10052
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585989459938802e-05,
+      "loss": 1.4844,
+      "step": 10053
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585905473841321e-05,
+      "loss": 0.9102,
+      "step": 10054
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5858214814502547e-05,
+      "loss": 0.9717,
+      "step": 10055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585737482766505e-05,
+      "loss": 1.4424,
+      "step": 10056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5856534777909752e-05,
+      "loss": 0.9336,
+      "step": 10057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5855694665245665e-05,
+      "loss": 1.2764,
+      "step": 10058
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5854854489681823e-05,
+      "loss": 1.2119,
+      "step": 10059
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5854014251227244e-05,
+      "loss": 0.9873,
+      "step": 10060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585317394989096e-05,
+      "loss": 1.2422,
+      "step": 10061
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585233358568199e-05,
+      "loss": 1.3789,
+      "step": 10062
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5851493158609362e-05,
+      "loss": 1.1719,
+      "step": 10063
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585065266868211e-05,
+      "loss": 1.3809,
+      "step": 10064
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5849812115909257e-05,
+      "loss": 1.043,
+      "step": 10065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584897150029983e-05,
+      "loss": 1.2402,
+      "step": 10066
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5848130821862866e-05,
+      "loss": 0.9189,
+      "step": 10067
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5847290080607388e-05,
+      "loss": 1.1582,
+      "step": 10068
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584644927654243e-05,
+      "loss": 0.9697,
+      "step": 10069
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5845608409677024e-05,
+      "loss": 1.1221,
+      "step": 10070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5844767480020203e-05,
+      "loss": 1.333,
+      "step": 10071
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5843926487580998e-05,
+      "loss": 0.908,
+      "step": 10072
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5843085432368442e-05,
+      "loss": 1.1016,
+      "step": 10073
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5842244314391568e-05,
+      "loss": 1.1973,
+      "step": 10074
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5841403133659416e-05,
+      "loss": 1.3105,
+      "step": 10075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584056189018102e-05,
+      "loss": 0.9209,
+      "step": 10076
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5839720583965413e-05,
+      "loss": 1.2227,
+      "step": 10077
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583887921502164e-05,
+      "loss": 1.0391,
+      "step": 10078
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5838037783358727e-05,
+      "loss": 1.1489,
+      "step": 10079
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5837196288985724e-05,
+      "loss": 0.854,
+      "step": 10080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583635473191166e-05,
+      "loss": 1.1113,
+      "step": 10081
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5835513112145583e-05,
+      "loss": 1.1904,
+      "step": 10082
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5834671429696524e-05,
+      "loss": 0.875,
+      "step": 10083
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5833829684573532e-05,
+      "loss": 1.2705,
+      "step": 10084
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583298787678565e-05,
+      "loss": 1.1025,
+      "step": 10085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5832146006341915e-05,
+      "loss": 0.8594,
+      "step": 10086
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583130407325137e-05,
+      "loss": 0.8711,
+      "step": 10087
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5830462077523064e-05,
+      "loss": 0.7905,
+      "step": 10088
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5829620019166034e-05,
+      "loss": 0.8672,
+      "step": 10089
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5828777898189333e-05,
+      "loss": 1.0449,
+      "step": 10090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5827935714602e-05,
+      "loss": 1.4102,
+      "step": 10091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5827093468413086e-05,
+      "loss": 1.042,
+      "step": 10092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5826251159631637e-05,
+      "loss": 1.252,
+      "step": 10093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5825408788266697e-05,
+      "loss": 1.332,
+      "step": 10094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582456635432732e-05,
+      "loss": 1.5098,
+      "step": 10095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5823723857822553e-05,
+      "loss": 0.9463,
+      "step": 10096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822881298761445e-05,
+      "loss": 1.1201,
+      "step": 10097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822038677153045e-05,
+      "loss": 1.2002,
+      "step": 10098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5821195993006407e-05,
+      "loss": 0.9688,
+      "step": 10099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5820353246330588e-05,
+      "loss": 0.9531,
+      "step": 10100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581951043713463e-05,
+      "loss": 1.1436,
+      "step": 10101
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5818667565427585e-05,
+      "loss": 1.1738,
+      "step": 10102
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5817824631218522e-05,
+      "loss": 1.0986,
+      "step": 10103
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5816981634516483e-05,
+      "loss": 1.2578,
+      "step": 10104
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581613857533052e-05,
+      "loss": 0.6797,
+      "step": 10105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5815295453669702e-05,
+      "loss": 1.0244,
+      "step": 10106
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5814452269543073e-05,
+      "loss": 0.8633,
+      "step": 10107
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58136090229597e-05,
+      "loss": 1.4072,
+      "step": 10108
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5812765713928638e-05,
+      "loss": 0.9727,
+      "step": 10109
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5811922342458938e-05,
+      "loss": 1.0371,
+      "step": 10110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581107890855967e-05,
+      "loss": 1.001,
+      "step": 10111
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5810235412239886e-05,
+      "loss": 1.3584,
+      "step": 10112
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5809391853508653e-05,
+      "loss": 0.8169,
+      "step": 10113
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5808548232375026e-05,
+      "loss": 1.3271,
+      "step": 10114
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580770454884807e-05,
+      "loss": 0.9912,
+      "step": 10115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580686080293685e-05,
+      "loss": 0.9561,
+      "step": 10116
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5806016994650425e-05,
+      "loss": 1.1709,
+      "step": 10117
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5805173123997856e-05,
+      "loss": 1.3594,
+      "step": 10118
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5804329190988216e-05,
+      "loss": 1.0527,
+      "step": 10119
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5803485195630565e-05,
+      "loss": 1.0977,
+      "step": 10120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580264113793397e-05,
+      "loss": 1.2637,
+      "step": 10121
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58017970179075e-05,
+      "loss": 1.0361,
+      "step": 10122
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5800952835560213e-05,
+      "loss": 1.293,
+      "step": 10123
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5800108590901192e-05,
+      "loss": 1.7832,
+      "step": 10124
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5799264283939492e-05,
+      "loss": 1.126,
+      "step": 10125
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579841991468419e-05,
+      "loss": 1.2637,
+      "step": 10126
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579757548314435e-05,
+      "loss": 1.3955,
+      "step": 10127
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579673098932905e-05,
+      "loss": 1.0693,
+      "step": 10128
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5795886433247355e-05,
+      "loss": 1.0693,
+      "step": 10129
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579504181490834e-05,
+      "loss": 1.5371,
+      "step": 10130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579419713432108e-05,
+      "loss": 1.2676,
+      "step": 10131
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579335239149464e-05,
+      "loss": 1.0713,
+      "step": 10132
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.57925075864381e-05,
+      "loss": 1.1309,
+      "step": 10133
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5791662719160533e-05,
+      "loss": 1.0474,
+      "step": 10134
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5790817789671017e-05,
+      "loss": 0.9941,
+      "step": 10135
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5789972797978623e-05,
+      "loss": 0.9624,
+      "step": 10136
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578912774409243e-05,
+      "loss": 0.7739,
+      "step": 10137
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578828262802152e-05,
+      "loss": 0.9409,
+      "step": 10138
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578743744977496e-05,
+      "loss": 0.9873,
+      "step": 10139
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5786592209361842e-05,
+      "loss": 0.9941,
+      "step": 10140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5785746906791236e-05,
+      "loss": 1.0518,
+      "step": 10141
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578490154207222e-05,
+      "loss": 1.0938,
+      "step": 10142
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5784056115213886e-05,
+      "loss": 1.1211,
+      "step": 10143
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5783210626225304e-05,
+      "loss": 1.6094,
+      "step": 10144
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5782365075115557e-05,
+      "loss": 1.3467,
+      "step": 10145
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5781519461893734e-05,
+      "loss": 0.6338,
+      "step": 10146
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5780673786568916e-05,
+      "loss": 0.8784,
+      "step": 10147
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5779828049150185e-05,
+      "loss": 1.0967,
+      "step": 10148
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5778982249646627e-05,
+      "loss": 1.1611,
+      "step": 10149
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5778136388067324e-05,
+      "loss": 1.1475,
+      "step": 10150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5777290464421367e-05,
+      "loss": 1.1484,
+      "step": 10151
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5776444478717842e-05,
+      "loss": 1.0215,
+      "step": 10152
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5775598430965833e-05,
+      "loss": 0.7803,
+      "step": 10153
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5774752321174428e-05,
+      "loss": 1.1777,
+      "step": 10154
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577390614935272e-05,
+      "loss": 1.1689,
+      "step": 10155
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5773059915509792e-05,
+      "loss": 0.8945,
+      "step": 10156
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.577221361965474e-05,
+      "loss": 0.9336,
+      "step": 10157
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5771367261796653e-05,
+      "loss": 0.937,
+      "step": 10158
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.577052084194462e-05,
+      "loss": 1.1631,
+      "step": 10159
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576967436010773e-05,
+      "loss": 1.2168,
+      "step": 10160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5768827816295088e-05,
+      "loss": 1.2383,
+      "step": 10161
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5767981210515772e-05,
+      "loss": 0.9346,
+      "step": 10162
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5767134542778885e-05,
+      "loss": 1.4336,
+      "step": 10163
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576628781309352e-05,
+      "loss": 1.2275,
+      "step": 10164
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576544102146877e-05,
+      "loss": 1.4043,
+      "step": 10165
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576459416791374e-05,
+      "loss": 0.9414,
+      "step": 10166
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576374725243751e-05,
+      "loss": 0.8218,
+      "step": 10167
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576290027504919e-05,
+      "loss": 1.1846,
+      "step": 10168
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5762053235757876e-05,
+      "loss": 1.2236,
+      "step": 10169
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5761206134572665e-05,
+      "loss": 1.0938,
+      "step": 10170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5760358971502658e-05,
+      "loss": 1.0962,
+      "step": 10171
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5759511746556948e-05,
+      "loss": 1.0166,
+      "step": 10172
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5758664459744646e-05,
+      "loss": 1.2734,
+      "step": 10173
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5757817111074846e-05,
+      "loss": 0.9688,
+      "step": 10174
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575696970055665e-05,
+      "loss": 1.209,
+      "step": 10175
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5756122228199163e-05,
+      "loss": 1.2773,
+      "step": 10176
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5755274694011492e-05,
+      "loss": 1.1035,
+      "step": 10177
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5754427098002735e-05,
+      "loss": 0.7881,
+      "step": 10178
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5753579440181996e-05,
+      "loss": 0.8105,
+      "step": 10179
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5752731720558383e-05,
+      "loss": 1.1729,
+      "step": 10180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5751883939141005e-05,
+      "loss": 0.9819,
+      "step": 10181
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5751036095938964e-05,
+      "loss": 0.8701,
+      "step": 10182
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5750188190961366e-05,
+      "loss": 0.9141,
+      "step": 10183
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5749340224217324e-05,
+      "loss": 0.9692,
+      "step": 10184
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574849219571594e-05,
+      "loss": 0.9424,
+      "step": 10185
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5747644105466327e-05,
+      "loss": 1.4209,
+      "step": 10186
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57467959534776e-05,
+      "loss": 0.8877,
+      "step": 10187
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574594773975886e-05,
+      "loss": 1.8516,
+      "step": 10188
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5745099464319223e-05,
+      "loss": 0.9453,
+      "step": 10189
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57442511271678e-05,
+      "loss": 0.9067,
+      "step": 10190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574340272831371e-05,
+      "loss": 0.7314,
+      "step": 10191
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5742554267766052e-05,
+      "loss": 0.918,
+      "step": 10192
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5741705745533954e-05,
+      "loss": 1.25,
+      "step": 10193
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574085716162652e-05,
+      "loss": 0.9717,
+      "step": 10194
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5740008516052872e-05,
+      "loss": 0.9873,
+      "step": 10195
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5739159808822123e-05,
+      "loss": 0.7861,
+      "step": 10196
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573831103994339e-05,
+      "loss": 1.1748,
+      "step": 10197
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573746220942579e-05,
+      "loss": 0.8003,
+      "step": 10198
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573661331727844e-05,
+      "loss": 1.123,
+      "step": 10199
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573576436351046e-05,
+      "loss": 1.1797,
+      "step": 10200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5734915348130973e-05,
+      "loss": 0.9668,
+      "step": 10201
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573406627114909e-05,
+      "loss": 0.9619,
+      "step": 10202
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5733217132573938e-05,
+      "loss": 1.3115,
+      "step": 10203
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5732367932414636e-05,
+      "loss": 0.832,
+      "step": 10204
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5731518670680307e-05,
+      "loss": 1.2539,
+      "step": 10205
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573066934738007e-05,
+      "loss": 0.8633,
+      "step": 10206
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5729819962523052e-05,
+      "loss": 1.2822,
+      "step": 10207
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5728970516118374e-05,
+      "loss": 0.8984,
+      "step": 10208
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5728121008175162e-05,
+      "loss": 1.2471,
+      "step": 10209
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572727143870254e-05,
+      "loss": 0.9756,
+      "step": 10210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572642180770964e-05,
+      "loss": 0.9116,
+      "step": 10211
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5725572115205575e-05,
+      "loss": 1.0122,
+      "step": 10212
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5724722361199486e-05,
+      "loss": 1.2451,
+      "step": 10213
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5723872545700492e-05,
+      "loss": 1.1953,
+      "step": 10214
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5723022668717728e-05,
+      "loss": 1.0977,
+      "step": 10215
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5722172730260314e-05,
+      "loss": 1.415,
+      "step": 10216
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572132273033739e-05,
+      "loss": 1.3438,
+      "step": 10217
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5720472668958076e-05,
+      "loss": 0.876,
+      "step": 10218
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5719622546131518e-05,
+      "loss": 1.0859,
+      "step": 10219
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571877236186683e-05,
+      "loss": 1.3613,
+      "step": 10220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5717922116173158e-05,
+      "loss": 1.2412,
+      "step": 10221
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5717071809059623e-05,
+      "loss": 1.252,
+      "step": 10222
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571622144053537e-05,
+      "loss": 1.1914,
+      "step": 10223
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5715371010609527e-05,
+      "loss": 0.8672,
+      "step": 10224
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5714520519291232e-05,
+      "loss": 1.3594,
+      "step": 10225
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571366996658962e-05,
+      "loss": 0.9214,
+      "step": 10226
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5712819352513827e-05,
+      "loss": 1.0298,
+      "step": 10227
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571196867707299e-05,
+      "loss": 1.1807,
+      "step": 10228
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5711117940276244e-05,
+      "loss": 1.0034,
+      "step": 10229
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5710267142132732e-05,
+      "loss": 0.8799,
+      "step": 10230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5709416282651587e-05,
+      "loss": 0.7427,
+      "step": 10231
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5708565361841953e-05,
+      "loss": 1.1758,
+      "step": 10232
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570771437971297e-05,
+      "loss": 0.8789,
+      "step": 10233
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5706863336273783e-05,
+      "loss": 1.291,
+      "step": 10234
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5706012231533527e-05,
+      "loss": 1.0684,
+      "step": 10235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5705161065501346e-05,
+      "loss": 0.7583,
+      "step": 10236
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5704309838186382e-05,
+      "loss": 0.8447,
+      "step": 10237
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570345854959778e-05,
+      "loss": 0.9482,
+      "step": 10238
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5702607199744684e-05,
+      "loss": 1.1074,
+      "step": 10239
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5701755788636242e-05,
+      "loss": 0.874,
+      "step": 10240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5700904316281593e-05,
+      "loss": 0.9541,
+      "step": 10241
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570005278268989e-05,
+      "loss": 1.2178,
+      "step": 10242
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569920118787027e-05,
+      "loss": 0.9966,
+      "step": 10243
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5698349531831895e-05,
+      "loss": 1.0508,
+      "step": 10244
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5697497814583903e-05,
+      "loss": 1.1113,
+      "step": 10245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5696646036135446e-05,
+      "loss": 1.1709,
+      "step": 10246
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569579419649567e-05,
+      "loss": 1.1128,
+      "step": 10247
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569494229567373e-05,
+      "loss": 0.7324,
+      "step": 10248
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5694090333678772e-05,
+      "loss": 1.0859,
+      "step": 10249
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5693238310519952e-05,
+      "loss": 1.0029,
+      "step": 10250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569238622620642e-05,
+      "loss": 0.8369,
+      "step": 10251
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569153408074733e-05,
+      "loss": 1.4512,
+      "step": 10252
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5690681874151832e-05,
+      "loss": 0.7812,
+      "step": 10253
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5689829606429084e-05,
+      "loss": 1.3887,
+      "step": 10254
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5688977277588237e-05,
+      "loss": 1.0684,
+      "step": 10255
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5688124887638455e-05,
+      "loss": 0.9785,
+      "step": 10256
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5687272436588877e-05,
+      "loss": 1.0938,
+      "step": 10257
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568641992444868e-05,
+      "loss": 1.2393,
+      "step": 10258
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568556735122701e-05,
+      "loss": 0.8721,
+      "step": 10259
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5684714716933024e-05,
+      "loss": 1.1211,
+      "step": 10260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5683862021575888e-05,
+      "loss": 1.1709,
+      "step": 10261
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5683009265164755e-05,
+      "loss": 1.1768,
+      "step": 10262
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5682156447708785e-05,
+      "loss": 1.0801,
+      "step": 10263
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5681303569217142e-05,
+      "loss": 1.0078,
+      "step": 10264
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5680450629698987e-05,
+      "loss": 1.0527,
+      "step": 10265
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567959762916348e-05,
+      "loss": 1.0044,
+      "step": 10266
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5678744567619783e-05,
+      "loss": 1.21,
+      "step": 10267
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5677891445077064e-05,
+      "loss": 1.1514,
+      "step": 10268
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5677038261544485e-05,
+      "loss": 1.0928,
+      "step": 10269
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5676185017031203e-05,
+      "loss": 1.5176,
+      "step": 10270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5675331711546392e-05,
+      "loss": 0.9517,
+      "step": 10271
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5674478345099217e-05,
+      "loss": 1.3428,
+      "step": 10272
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5673624917698843e-05,
+      "loss": 1.6426,
+      "step": 10273
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5672771429354436e-05,
+      "loss": 0.4473,
+      "step": 10274
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5671917880075165e-05,
+      "loss": 0.9258,
+      "step": 10275
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56710642698702e-05,
+      "loss": 1.041,
+      "step": 10276
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567021059874871e-05,
+      "loss": 0.8887,
+      "step": 10277
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566935686671986e-05,
+      "loss": 1.0703,
+      "step": 10278
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566850307379283e-05,
+      "loss": 1.1006,
+      "step": 10279
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5667649219976773e-05,
+      "loss": 1.1465,
+      "step": 10280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5666795305280883e-05,
+      "loss": 1.0664,
+      "step": 10281
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566594132971432e-05,
+      "loss": 0.5918,
+      "step": 10282
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5665087293286264e-05,
+      "loss": 0.8584,
+      "step": 10283
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566423319600588e-05,
+      "loss": 1.0557,
+      "step": 10284
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5663379037882345e-05,
+      "loss": 1.0566,
+      "step": 10285
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5662524818924837e-05,
+      "loss": 1.4551,
+      "step": 10286
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5661670539142533e-05,
+      "loss": 0.792,
+      "step": 10287
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5660816198544605e-05,
+      "loss": 1.2578,
+      "step": 10288
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565996179714023e-05,
+      "loss": 1.4053,
+      "step": 10289
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5659107334938592e-05,
+      "loss": 1.1514,
+      "step": 10290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565825281194886e-05,
+      "loss": 1.2295,
+      "step": 10291
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565739822818022e-05,
+      "loss": 1.2529,
+      "step": 10292
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565654358364185e-05,
+      "loss": 0.9473,
+      "step": 10293
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565568887834293e-05,
+      "loss": 1.2832,
+      "step": 10294
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565483411229264e-05,
+      "loss": 1.1016,
+      "step": 10295
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565397928550016e-05,
+      "loss": 1.0605,
+      "step": 10296
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565312439797468e-05,
+      "loss": 1.0879,
+      "step": 10297
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5652269449725375e-05,
+      "loss": 1.3945,
+      "step": 10298
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565141444076143e-05,
+      "loss": 1.1631,
+      "step": 10299
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565055937109203e-05,
+      "loss": 1.0908,
+      "step": 10300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564970424072636e-05,
+      "loss": 1.0029,
+      "step": 10301
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5648849049673606e-05,
+      "loss": 1.29,
+      "step": 10302
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5647993797942954e-05,
+      "loss": 0.8174,
+      "step": 10303
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564713848554359e-05,
+      "loss": 1.207,
+      "step": 10304
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56462831124847e-05,
+      "loss": 1.2461,
+      "step": 10305
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5645427678775478e-05,
+      "loss": 1.1211,
+      "step": 10306
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5644572184425106e-05,
+      "loss": 0.855,
+      "step": 10307
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5643716629442778e-05,
+      "loss": 0.8696,
+      "step": 10308
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564286101383768e-05,
+      "loss": 1.1973,
+      "step": 10309
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5642005337619005e-05,
+      "loss": 0.9526,
+      "step": 10310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5641149600795945e-05,
+      "loss": 1.002,
+      "step": 10311
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5640293803377697e-05,
+      "loss": 1.2695,
+      "step": 10312
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563943794537344e-05,
+      "loss": 1.3926,
+      "step": 10313
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5638582026792378e-05,
+      "loss": 1.2871,
+      "step": 10314
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5637726047643705e-05,
+      "loss": 1.6484,
+      "step": 10315
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563687000793661e-05,
+      "loss": 1.1885,
+      "step": 10316
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563601390768029e-05,
+      "loss": 0.9033,
+      "step": 10317
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5635157746883947e-05,
+      "loss": 1.1826,
+      "step": 10318
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5634301525556767e-05,
+      "loss": 0.9062,
+      "step": 10319
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5633445243707955e-05,
+      "loss": 0.9873,
+      "step": 10320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563258890134671e-05,
+      "loss": 1.3594,
+      "step": 10321
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5631732498482222e-05,
+      "loss": 1.126,
+      "step": 10322
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56308760351237e-05,
+      "loss": 1.2529,
+      "step": 10323
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5630019511280335e-05,
+      "loss": 1.002,
+      "step": 10324
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5629162926961336e-05,
+      "loss": 1.3486,
+      "step": 10325
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56283062821759e-05,
+      "loss": 1.2539,
+      "step": 10326
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5627449576933225e-05,
+      "loss": 1.4141,
+      "step": 10327
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562659281124252e-05,
+      "loss": 0.9082,
+      "step": 10328
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5625735985112983e-05,
+      "loss": 1.0479,
+      "step": 10329
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562487909855382e-05,
+      "loss": 1.373,
+      "step": 10330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5624022151574237e-05,
+      "loss": 1.1143,
+      "step": 10331
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5623165144183436e-05,
+      "loss": 0.8623,
+      "step": 10332
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5622308076390625e-05,
+      "loss": 1.1206,
+      "step": 10333
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5621450948205008e-05,
+      "loss": 1.3408,
+      "step": 10334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562059375963579e-05,
+      "loss": 1.2275,
+      "step": 10335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5619736510692187e-05,
+      "loss": 1.1621,
+      "step": 10336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56188792013834e-05,
+      "loss": 1.1777,
+      "step": 10337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5618021831718642e-05,
+      "loss": 1.707,
+      "step": 10338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5617164401707122e-05,
+      "loss": 1.2461,
+      "step": 10339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5616306911358043e-05,
+      "loss": 1.2227,
+      "step": 10340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5615449360680622e-05,
+      "loss": 1.0186,
+      "step": 10341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5614591749684077e-05,
+      "loss": 1.2764,
+      "step": 10342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5613734078377608e-05,
+      "loss": 1.1357,
+      "step": 10343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5612876346770432e-05,
+      "loss": 1.1377,
+      "step": 10344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5612018554871763e-05,
+      "loss": 1.2715,
+      "step": 10345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561116070269082e-05,
+      "loss": 1.0293,
+      "step": 10346
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5610302790236807e-05,
+      "loss": 0.9253,
+      "step": 10347
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5609444817518952e-05,
+      "loss": 1.0952,
+      "step": 10348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5608586784546463e-05,
+      "loss": 1.1992,
+      "step": 10349
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5607728691328555e-05,
+      "loss": 0.7881,
+      "step": 10350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5606870537874452e-05,
+      "loss": 1.0166,
+      "step": 10351
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560601232419337e-05,
+      "loss": 1.2637,
+      "step": 10352
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5605154050294524e-05,
+      "loss": 1.4473,
+      "step": 10353
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5604295716187135e-05,
+      "loss": 0.9414,
+      "step": 10354
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5603437321880423e-05,
+      "loss": 0.9404,
+      "step": 10355
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560257886738361e-05,
+      "loss": 0.5796,
+      "step": 10356
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5601720352705917e-05,
+      "loss": 0.8623,
+      "step": 10357
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5600861777856568e-05,
+      "loss": 1.1846,
+      "step": 10358
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5600003142844776e-05,
+      "loss": 1.0918,
+      "step": 10359
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5599144447679777e-05,
+      "loss": 1.2559,
+      "step": 10360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5598285692370784e-05,
+      "loss": 0.9424,
+      "step": 10361
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559742687692703e-05,
+      "loss": 1.2422,
+      "step": 10362
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559656800135773e-05,
+      "loss": 1.4707,
+      "step": 10363
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559570906567212e-05,
+      "loss": 1.0547,
+      "step": 10364
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5594850069879424e-05,
+      "loss": 1.3047,
+      "step": 10365
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5593991013988863e-05,
+      "loss": 1.2588,
+      "step": 10366
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559313189800967e-05,
+      "loss": 1.127,
+      "step": 10367
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5592272721951073e-05,
+      "loss": 1.3115,
+      "step": 10368
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5591413485822303e-05,
+      "loss": 0.9512,
+      "step": 10369
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5590554189632585e-05,
+      "loss": 1.2061,
+      "step": 10370
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558969483339115e-05,
+      "loss": 0.8125,
+      "step": 10371
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5588835417107227e-05,
+      "loss": 1.0361,
+      "step": 10372
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5587975940790054e-05,
+      "loss": 0.9082,
+      "step": 10373
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558711640444886e-05,
+      "loss": 1.3164,
+      "step": 10374
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5586256808092877e-05,
+      "loss": 0.9575,
+      "step": 10375
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558539715173134e-05,
+      "loss": 0.7075,
+      "step": 10376
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558453743537348e-05,
+      "loss": 1.0742,
+      "step": 10377
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5583677659028535e-05,
+      "loss": 0.623,
+      "step": 10378
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5582817822705733e-05,
+      "loss": 1.2236,
+      "step": 10379
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5581957926414327e-05,
+      "loss": 0.8486,
+      "step": 10380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5581097970163536e-05,
+      "loss": 0.8628,
+      "step": 10381
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5580237953962605e-05,
+      "loss": 1.0508,
+      "step": 10382
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5579377877820774e-05,
+      "loss": 0.8994,
+      "step": 10383
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5578517741747274e-05,
+      "loss": 0.8979,
+      "step": 10384
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5577657545751355e-05,
+      "loss": 0.8779,
+      "step": 10385
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557679728984225e-05,
+      "loss": 0.7651,
+      "step": 10386
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5575936974029196e-05,
+      "loss": 1.0049,
+      "step": 10387
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5575076598321445e-05,
+      "loss": 1.0391,
+      "step": 10388
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5574216162728227e-05,
+      "loss": 1.2656,
+      "step": 10389
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5573355667258792e-05,
+      "loss": 0.7329,
+      "step": 10390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5572495111922387e-05,
+      "loss": 1.2256,
+      "step": 10391
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5571634496728243e-05,
+      "loss": 0.8975,
+      "step": 10392
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5570773821685615e-05,
+      "loss": 0.9082,
+      "step": 10393
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5569913086803744e-05,
+      "loss": 0.9863,
+      "step": 10394
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5569052292091877e-05,
+      "loss": 1.1943,
+      "step": 10395
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5568191437559257e-05,
+      "loss": 1.0439,
+      "step": 10396
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5567330523215133e-05,
+      "loss": 0.9785,
+      "step": 10397
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5566469549068758e-05,
+      "loss": 1.3096,
+      "step": 10398
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5565608515129373e-05,
+      "loss": 0.5923,
+      "step": 10399
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556474742140623e-05,
+      "loss": 1.0732,
+      "step": 10400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5563886267908576e-05,
+      "loss": 0.9644,
+      "step": 10401
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5563025054645665e-05,
+      "loss": 0.9175,
+      "step": 10402
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5562163781626747e-05,
+      "loss": 1.2275,
+      "step": 10403
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556130244886107e-05,
+      "loss": 1.3379,
+      "step": 10404
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556044105635789e-05,
+      "loss": 1.498,
+      "step": 10405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5559579604126458e-05,
+      "loss": 0.9746,
+      "step": 10406
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5558718092176026e-05,
+      "loss": 1.0542,
+      "step": 10407
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5557856520515856e-05,
+      "loss": 0.8892,
+      "step": 10408
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555699488915519e-05,
+      "loss": 1.04,
+      "step": 10409
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5556133198103292e-05,
+      "loss": 1.666,
+      "step": 10410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5555271447369417e-05,
+      "loss": 1.2666,
+      "step": 10411
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555440963696282e-05,
+      "loss": 0.917,
+      "step": 10412
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555354776689276e-05,
+      "loss": 1.0547,
+      "step": 10413
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555268583716849e-05,
+      "loss": 1.1509,
+      "step": 10414
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5551823847799274e-05,
+      "loss": 0.8066,
+      "step": 10415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555096179879437e-05,
+      "loss": 0.9033,
+      "step": 10416
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5550099690163038e-05,
+      "loss": 0.9834,
+      "step": 10417
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5549237521914537e-05,
+      "loss": 1.1123,
+      "step": 10418
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554837529405813e-05,
+      "loss": 1.2607,
+      "step": 10419
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5547513006603075e-05,
+      "loss": 1.2451,
+      "step": 10420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5546650659558638e-05,
+      "loss": 0.8745,
+      "step": 10421
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554578825293408e-05,
+      "loss": 1.0107,
+      "step": 10422
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5544925786738665e-05,
+      "loss": 1.0762,
+      "step": 10423
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554406326098166e-05,
+      "loss": 1.2842,
+      "step": 10424
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5543200675672324e-05,
+      "loss": 1.25,
+      "step": 10425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5542338030819933e-05,
+      "loss": 1.1709,
+      "step": 10426
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5541475326433743e-05,
+      "loss": 1.1035,
+      "step": 10427
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5540612562523025e-05,
+      "loss": 0.7603,
+      "step": 10428
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5539749739097047e-05,
+      "loss": 1.0166,
+      "step": 10429
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5538886856165073e-05,
+      "loss": 0.8125,
+      "step": 10430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5538023913736374e-05,
+      "loss": 0.6558,
+      "step": 10431
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5537160911820225e-05,
+      "loss": 1.3135,
+      "step": 10432
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5536297850425888e-05,
+      "loss": 0.9111,
+      "step": 10433
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5535434729562637e-05,
+      "loss": 1.0693,
+      "step": 10434
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5534571549239744e-05,
+      "loss": 1.3926,
+      "step": 10435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553370830946648e-05,
+      "loss": 1.4492,
+      "step": 10436
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553284501025212e-05,
+      "loss": 1.2275,
+      "step": 10437
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5531981651605934e-05,
+      "loss": 1.0273,
+      "step": 10438
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5531118233537196e-05,
+      "loss": 0.8975,
+      "step": 10439
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5530254756055185e-05,
+      "loss": 1.0518,
+      "step": 10440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552939121916917e-05,
+      "loss": 0.7524,
+      "step": 10441
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552852762288843e-05,
+      "loss": 0.8857,
+      "step": 10442
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5527663967222236e-05,
+      "loss": 0.9902,
+      "step": 10443
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5526800252179877e-05,
+      "loss": 0.8354,
+      "step": 10444
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5525936477770623e-05,
+      "loss": 1.3809,
+      "step": 10445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552507264400375e-05,
+      "loss": 0.8984,
+      "step": 10446
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5524208750888543e-05,
+      "loss": 0.9834,
+      "step": 10447
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5523344798434277e-05,
+      "loss": 1.248,
+      "step": 10448
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5522480786650236e-05,
+      "loss": 0.9111,
+      "step": 10449
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5521616715545696e-05,
+      "loss": 1.1465,
+      "step": 10450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5520752585129944e-05,
+      "loss": 0.9707,
+      "step": 10451
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551988839541226e-05,
+      "loss": 1.1143,
+      "step": 10452
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5519024146401927e-05,
+      "loss": 1.2676,
+      "step": 10453
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5518159838108226e-05,
+      "loss": 1.041,
+      "step": 10454
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5517295470540444e-05,
+      "loss": 1.4199,
+      "step": 10455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5516431043707863e-05,
+      "loss": 1.0264,
+      "step": 10456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551556655761977e-05,
+      "loss": 1.168,
+      "step": 10457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5514702012285455e-05,
+      "loss": 1.0381,
+      "step": 10458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5513837407714195e-05,
+      "loss": 1.5596,
+      "step": 10459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5512972743915287e-05,
+      "loss": 1.498,
+      "step": 10460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5512108020898012e-05,
+      "loss": 1.1064,
+      "step": 10461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5511243238671668e-05,
+      "loss": 1.124,
+      "step": 10462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5510378397245533e-05,
+      "loss": 1.2773,
+      "step": 10463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55095134966289e-05,
+      "loss": 0.9629,
+      "step": 10464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5508648536831065e-05,
+      "loss": 1.1953,
+      "step": 10465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5507783517861308e-05,
+      "loss": 1.1982,
+      "step": 10466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5506918439728933e-05,
+      "loss": 1.1035,
+      "step": 10467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5506053302443226e-05,
+      "loss": 0.7422,
+      "step": 10468
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550518810601348e-05,
+      "loss": 0.9756,
+      "step": 10469
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550432285044899e-05,
+      "loss": 1.7012,
+      "step": 10470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5503457535759053e-05,
+      "loss": 0.7002,
+      "step": 10471
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5502592161952958e-05,
+      "loss": 1.2432,
+      "step": 10472
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5501726729040006e-05,
+      "loss": 1.2559,
+      "step": 10473
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5500861237029488e-05,
+      "loss": 0.6123,
+      "step": 10474
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5499995685930703e-05,
+      "loss": 1.2725,
+      "step": 10475
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549913007575295e-05,
+      "loss": 0.8652,
+      "step": 10476
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5498264406505526e-05,
+      "loss": 1.4492,
+      "step": 10477
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549739867819773e-05,
+      "loss": 1.166,
+      "step": 10478
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549653289083886e-05,
+      "loss": 1.4805,
+      "step": 10479
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5495667044438216e-05,
+      "loss": 1.4492,
+      "step": 10480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54948011390051e-05,
+      "loss": 0.8926,
+      "step": 10481
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5493935174548815e-05,
+      "loss": 1.082,
+      "step": 10482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5493069151078658e-05,
+      "loss": 1.0947,
+      "step": 10483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5492203068603935e-05,
+      "loss": 1.1963,
+      "step": 10484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5491336927133946e-05,
+      "loss": 1.209,
+      "step": 10485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5490470726678003e-05,
+      "loss": 1.0947,
+      "step": 10486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54896044672454e-05,
+      "loss": 0.7964,
+      "step": 10487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5488738148845444e-05,
+      "loss": 1.0063,
+      "step": 10488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548787177148745e-05,
+      "loss": 1.041,
+      "step": 10489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548700533518071e-05,
+      "loss": 0.9395,
+      "step": 10490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5486138839934545e-05,
+      "loss": 1.2188,
+      "step": 10491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548527228575825e-05,
+      "loss": 0.8867,
+      "step": 10492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5484405672661146e-05,
+      "loss": 1.5498,
+      "step": 10493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5483539000652532e-05,
+      "loss": 1.1934,
+      "step": 10494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5482672269741718e-05,
+      "loss": 1.0938,
+      "step": 10495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5481805479938018e-05,
+      "loss": 1.2529,
+      "step": 10496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5480938631250745e-05,
+      "loss": 1.2568,
+      "step": 10497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5480071723689204e-05,
+      "loss": 1.0166,
+      "step": 10498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5479204757262708e-05,
+      "loss": 1.1265,
+      "step": 10499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5478337731980576e-05,
+      "loss": 1.0684,
+      "step": 10500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5477470647852113e-05,
+      "loss": 1.0137,
+      "step": 10501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5476603504886638e-05,
+      "loss": 1.2344,
+      "step": 10502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5475736303093464e-05,
+      "loss": 1.0195,
+      "step": 10503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5474869042481906e-05,
+      "loss": 0.9897,
+      "step": 10504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5474001723061277e-05,
+      "loss": 0.9155,
+      "step": 10505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5473134344840905e-05,
+      "loss": 0.8843,
+      "step": 10506
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5472266907830094e-05,
+      "loss": 0.937,
+      "step": 10507
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5471399412038164e-05,
+      "loss": 0.9521,
+      "step": 10508
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547053185747444e-05,
+      "loss": 1.124,
+      "step": 10509
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546966424414823e-05,
+      "loss": 1.1904,
+      "step": 10510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5468796572068864e-05,
+      "loss": 1.3867,
+      "step": 10511
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546792884124566e-05,
+      "loss": 1.0762,
+      "step": 10512
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5467061051687936e-05,
+      "loss": 1.1406,
+      "step": 10513
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5466193203405017e-05,
+      "loss": 1.1797,
+      "step": 10514
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5465325296406218e-05,
+      "loss": 1.21,
+      "step": 10515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546445733070087e-05,
+      "loss": 0.8994,
+      "step": 10516
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546358930629829e-05,
+      "loss": 0.7974,
+      "step": 10517
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546272122320781e-05,
+      "loss": 1.333,
+      "step": 10518
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5461853081438745e-05,
+      "loss": 0.9316,
+      "step": 10519
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5460984881000428e-05,
+      "loss": 0.7256,
+      "step": 10520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546011662190218e-05,
+      "loss": 1.334,
+      "step": 10521
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545924830415333e-05,
+      "loss": 1.2119,
+      "step": 10522
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54583799277632e-05,
+      "loss": 1.1855,
+      "step": 10523
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5457511492741128e-05,
+      "loss": 1.2812,
+      "step": 10524
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5456642999096437e-05,
+      "loss": 1.2383,
+      "step": 10525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5455774446838453e-05,
+      "loss": 1.1001,
+      "step": 10526
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5454905835976508e-05,
+      "loss": 0.9204,
+      "step": 10527
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5454037166519934e-05,
+      "loss": 1.2344,
+      "step": 10528
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5453168438478063e-05,
+      "loss": 1.2832,
+      "step": 10529
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5452299651860217e-05,
+      "loss": 0.7988,
+      "step": 10530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5451430806675742e-05,
+      "loss": 1.2114,
+      "step": 10531
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545056190293396e-05,
+      "loss": 1.2852,
+      "step": 10532
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5449692940644213e-05,
+      "loss": 1.0752,
+      "step": 10533
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544882391981583e-05,
+      "loss": 1.3223,
+      "step": 10534
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5447954840458145e-05,
+      "loss": 0.6416,
+      "step": 10535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5447085702580493e-05,
+      "loss": 0.9961,
+      "step": 10536
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5446216506192214e-05,
+      "loss": 1.0254,
+      "step": 10537
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5445347251302642e-05,
+      "loss": 1.1416,
+      "step": 10538
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5444477937921115e-05,
+      "loss": 1.1455,
+      "step": 10539
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5443608566056968e-05,
+      "loss": 0.96,
+      "step": 10540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5442739135719547e-05,
+      "loss": 1.3242,
+      "step": 10541
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544186964691818e-05,
+      "loss": 1.3262,
+      "step": 10542
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5441000099662215e-05,
+      "loss": 1.1943,
+      "step": 10543
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544013049396099e-05,
+      "loss": 1.4502,
+      "step": 10544
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5439260829823847e-05,
+      "loss": 1.2207,
+      "step": 10545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5438391107260128e-05,
+      "loss": 1.0244,
+      "step": 10546
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543752132627917e-05,
+      "loss": 1.04,
+      "step": 10547
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5436651486890323e-05,
+      "loss": 0.9219,
+      "step": 10548
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5435781589102928e-05,
+      "loss": 0.9424,
+      "step": 10549
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5434911632926322e-05,
+      "loss": 0.8799,
+      "step": 10550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5434041618369864e-05,
+      "loss": 0.9268,
+      "step": 10551
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543317154544289e-05,
+      "loss": 1.0254,
+      "step": 10552
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5432301414154744e-05,
+      "loss": 1.4277,
+      "step": 10553
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5431431224514777e-05,
+      "loss": 0.9551,
+      "step": 10554
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543056097653234e-05,
+      "loss": 1.2432,
+      "step": 10555
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5429690670216773e-05,
+      "loss": 0.7397,
+      "step": 10556
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5428820305577427e-05,
+      "loss": 1.25,
+      "step": 10557
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5427949882623655e-05,
+      "loss": 0.9766,
+      "step": 10558
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5427079401364803e-05,
+      "loss": 1.2656,
+      "step": 10559
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542620886181022e-05,
+      "loss": 1.4941,
+      "step": 10560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542533826396926e-05,
+      "loss": 1.3965,
+      "step": 10561
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5424467607851274e-05,
+      "loss": 1.0977,
+      "step": 10562
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5423596893465615e-05,
+      "loss": 1.5918,
+      "step": 10563
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5422726120821635e-05,
+      "loss": 0.7695,
+      "step": 10564
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5421855289928687e-05,
+      "loss": 0.8291,
+      "step": 10565
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5420984400796126e-05,
+      "loss": 0.873,
+      "step": 10566
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5420113453433305e-05,
+      "loss": 0.5542,
+      "step": 10567
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541924244784958e-05,
+      "loss": 1.291,
+      "step": 10568
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5418371384054315e-05,
+      "loss": 1.2256,
+      "step": 10569
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541750026205685e-05,
+      "loss": 1.4512,
+      "step": 10570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5416629081866558e-05,
+      "loss": 1.3398,
+      "step": 10571
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5415757843492788e-05,
+      "loss": 0.8867,
+      "step": 10572
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5414886546944904e-05,
+      "loss": 1.5742,
+      "step": 10573
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5414015192232263e-05,
+      "loss": 1.3105,
+      "step": 10574
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541314377936422e-05,
+      "loss": 1.2012,
+      "step": 10575
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5412272308350142e-05,
+      "loss": 1.3633,
+      "step": 10576
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5411400779199387e-05,
+      "loss": 1.4121,
+      "step": 10577
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541052919192132e-05,
+      "loss": 1.4004,
+      "step": 10578
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5409657546525297e-05,
+      "loss": 1.3848,
+      "step": 10579
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5408785843020687e-05,
+      "loss": 1.0615,
+      "step": 10580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540791408141685e-05,
+      "loss": 0.7847,
+      "step": 10581
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540704226172315e-05,
+      "loss": 0.8818,
+      "step": 10582
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5406170383948955e-05,
+      "loss": 1.0933,
+      "step": 10583
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540529844810363e-05,
+      "loss": 1.168,
+      "step": 10584
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5404426454196534e-05,
+      "loss": 1.4277,
+      "step": 10585
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540355440223704e-05,
+      "loss": 1.0176,
+      "step": 10586
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540268229223452e-05,
+      "loss": 1.5996,
+      "step": 10587
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540181012419833e-05,
+      "loss": 1.1689,
+      "step": 10588
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5400937898137848e-05,
+      "loss": 1.0977,
+      "step": 10589
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5400065614062437e-05,
+      "loss": 1.062,
+      "step": 10590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539919327198147e-05,
+      "loss": 1.5469,
+      "step": 10591
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539832087190432e-05,
+      "loss": 0.54,
+      "step": 10592
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5397448413840353e-05,
+      "loss": 1.2598,
+      "step": 10593
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539657589779894e-05,
+      "loss": 1.2256,
+      "step": 10594
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539570332378946e-05,
+      "loss": 0.8501,
+      "step": 10595
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539483069182128e-05,
+      "loss": 1.3809,
+      "step": 10596
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5393958001903775e-05,
+      "loss": 1.0762,
+      "step": 10597
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5393085254046322e-05,
+      "loss": 0.7998,
+      "step": 10598
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539221244825829e-05,
+      "loss": 1.4141,
+      "step": 10599
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5391339584549058e-05,
+      "loss": 1.3066,
+      "step": 10600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5390466662928e-05,
+      "loss": 1.1143,
+      "step": 10601
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53895936834045e-05,
+      "loss": 1.3164,
+      "step": 10602
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538872064598792e-05,
+      "loss": 0.8867,
+      "step": 10603
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5387847550687656e-05,
+      "loss": 1.3169,
+      "step": 10604
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5386974397513076e-05,
+      "loss": 1.0811,
+      "step": 10605
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5386101186473556e-05,
+      "loss": 1.0889,
+      "step": 10606
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5385227917578483e-05,
+      "loss": 1.0967,
+      "step": 10607
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5384354590837233e-05,
+      "loss": 1.2017,
+      "step": 10608
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538348120625919e-05,
+      "loss": 0.9609,
+      "step": 10609
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5382607763853736e-05,
+      "loss": 0.7998,
+      "step": 10610
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5381734263630252e-05,
+      "loss": 0.8408,
+      "step": 10611
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5380860705598114e-05,
+      "loss": 1.375,
+      "step": 10612
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537998708976672e-05,
+      "loss": 1.124,
+      "step": 10613
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537911341614544e-05,
+      "loss": 1.3418,
+      "step": 10614
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5378239684743665e-05,
+      "loss": 1.3584,
+      "step": 10615
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537736589557078e-05,
+      "loss": 0.9731,
+      "step": 10616
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5376492048636173e-05,
+      "loss": 0.7949,
+      "step": 10617
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537561814394923e-05,
+      "loss": 0.9258,
+      "step": 10618
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537474418151933e-05,
+      "loss": 1.0195,
+      "step": 10619
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5373870161355877e-05,
+      "loss": 1.0508,
+      "step": 10620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5372996083468242e-05,
+      "loss": 1.002,
+      "step": 10621
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537212194786582e-05,
+      "loss": 1.3887,
+      "step": 10622
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5371247754558008e-05,
+      "loss": 1.0791,
+      "step": 10623
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537037350355419e-05,
+      "loss": 0.7715,
+      "step": 10624
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5369499194863754e-05,
+      "loss": 1.001,
+      "step": 10625
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5368624828496098e-05,
+      "loss": 0.9141,
+      "step": 10626
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536775040446061e-05,
+      "loss": 1.0332,
+      "step": 10627
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536687592276668e-05,
+      "loss": 1.1533,
+      "step": 10628
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5366001383423708e-05,
+      "loss": 0.9321,
+      "step": 10629
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5365126786441087e-05,
+      "loss": 1.6348,
+      "step": 10630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5364252131828213e-05,
+      "loss": 1.1025,
+      "step": 10631
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536337741959447e-05,
+      "loss": 1.2393,
+      "step": 10632
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536250264974927e-05,
+      "loss": 1.1104,
+      "step": 10633
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5361627822301997e-05,
+      "loss": 1.0449,
+      "step": 10634
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536075293726205e-05,
+      "loss": 1.0405,
+      "step": 10635
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5359877994638835e-05,
+      "loss": 0.9399,
+      "step": 10636
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535900299444174e-05,
+      "loss": 1.2891,
+      "step": 10637
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5358127936680172e-05,
+      "loss": 1.1104,
+      "step": 10638
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5357252821363524e-05,
+      "loss": 0.9307,
+      "step": 10639
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53563776485012e-05,
+      "loss": 0.9224,
+      "step": 10640
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5355502418102603e-05,
+      "loss": 0.8325,
+      "step": 10641
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5354627130177123e-05,
+      "loss": 1.3477,
+      "step": 10642
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5353751784734177e-05,
+      "loss": 0.9893,
+      "step": 10643
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535287638178316e-05,
+      "loss": 1.2441,
+      "step": 10644
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5352000921333476e-05,
+      "loss": 0.7983,
+      "step": 10645
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535112540339453e-05,
+      "loss": 1.2012,
+      "step": 10646
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5350249827975724e-05,
+      "loss": 0.7588,
+      "step": 10647
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5349374195086468e-05,
+      "loss": 1.1289,
+      "step": 10648
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534849850473616e-05,
+      "loss": 0.7881,
+      "step": 10649
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5347622756934214e-05,
+      "loss": 1.4023,
+      "step": 10650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5346746951690033e-05,
+      "loss": 0.6885,
+      "step": 10651
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5345871089013024e-05,
+      "loss": 1.1465,
+      "step": 10652
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53449951689126e-05,
+      "loss": 1.04,
+      "step": 10653
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5344119191398163e-05,
+      "loss": 1.3867,
+      "step": 10654
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5343243156479125e-05,
+      "loss": 1.3457,
+      "step": 10655
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53423670641649e-05,
+      "loss": 1.0352,
+      "step": 10656
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5341490914464896e-05,
+      "loss": 0.8379,
+      "step": 10657
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534061470738852e-05,
+      "loss": 0.9941,
+      "step": 10658
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5339738442945192e-05,
+      "loss": 1.3398,
+      "step": 10659
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5338862121144318e-05,
+      "loss": 1.1353,
+      "step": 10660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5337985741995312e-05,
+      "loss": 1.1982,
+      "step": 10661
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5337109305507593e-05,
+      "loss": 1.1592,
+      "step": 10662
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533623281169057e-05,
+      "loss": 0.9961,
+      "step": 10663
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533535626055366e-05,
+      "loss": 0.9844,
+      "step": 10664
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5334479652106274e-05,
+      "loss": 1.0317,
+      "step": 10665
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5333602986357833e-05,
+      "loss": 1.1123,
+      "step": 10666
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5332726263317757e-05,
+      "loss": 1.2861,
+      "step": 10667
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533184948299546e-05,
+      "loss": 0.709,
+      "step": 10668
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5330972645400355e-05,
+      "loss": 1.1973,
+      "step": 10669
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533009575054187e-05,
+      "loss": 1.3418,
+      "step": 10670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5329218798429416e-05,
+      "loss": 1.4434,
+      "step": 10671
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532834178907242e-05,
+      "loss": 0.9717,
+      "step": 10672
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5327464722480296e-05,
+      "loss": 1.0244,
+      "step": 10673
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5326587598662468e-05,
+      "loss": 0.915,
+      "step": 10674
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532571041762836e-05,
+      "loss": 1.1836,
+      "step": 10675
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5324833179387392e-05,
+      "loss": 1.5137,
+      "step": 10676
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5323955883948986e-05,
+      "loss": 1.3467,
+      "step": 10677
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5323078531322567e-05,
+      "loss": 1.21,
+      "step": 10678
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5322201121517562e-05,
+      "loss": 1.0361,
+      "step": 10679
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5321323654543387e-05,
+      "loss": 1.1274,
+      "step": 10680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532044613040948e-05,
+      "loss": 0.938,
+      "step": 10681
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5319568549125253e-05,
+      "loss": 1.3096,
+      "step": 10682
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5318690910700144e-05,
+      "loss": 1.4033,
+      "step": 10683
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5317813215143577e-05,
+      "loss": 1.333,
+      "step": 10684
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5316935462464978e-05,
+      "loss": 0.6475,
+      "step": 10685
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5316057652673776e-05,
+      "loss": 1.0146,
+      "step": 10686
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5315179785779402e-05,
+      "loss": 1.2715,
+      "step": 10687
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5314301861791286e-05,
+      "loss": 1.1265,
+      "step": 10688
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5313423880718852e-05,
+      "loss": 1.4355,
+      "step": 10689
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531254584257154e-05,
+      "loss": 0.9902,
+      "step": 10690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5311667747358776e-05,
+      "loss": 1.2256,
+      "step": 10691
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5310789595089993e-05,
+      "loss": 1.0049,
+      "step": 10692
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5309911385774628e-05,
+      "loss": 1.3145,
+      "step": 10693
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5309033119422107e-05,
+      "loss": 1.2051,
+      "step": 10694
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5308154796041866e-05,
+      "loss": 1.123,
+      "step": 10695
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5307276415643346e-05,
+      "loss": 1.2393,
+      "step": 10696
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5306397978235977e-05,
+      "loss": 0.8408,
+      "step": 10697
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5305519483829193e-05,
+      "loss": 1.3789,
+      "step": 10698
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5304640932432434e-05,
+      "loss": 1.0659,
+      "step": 10699
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5303762324055134e-05,
+      "loss": 1.0229,
+      "step": 10700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530288365870674e-05,
+      "loss": 0.8984,
+      "step": 10701
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5302004936396676e-05,
+      "loss": 0.876,
+      "step": 10702
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530112615713439e-05,
+      "loss": 1.3467,
+      "step": 10703
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5300247320929316e-05,
+      "loss": 1.5137,
+      "step": 10704
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5299368427790904e-05,
+      "loss": 1.1455,
+      "step": 10705
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298489477728587e-05,
+      "loss": 0.8345,
+      "step": 10706
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5297610470751806e-05,
+      "loss": 0.8818,
+      "step": 10707
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5296731406870003e-05,
+      "loss": 1.584,
+      "step": 10708
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5295852286092628e-05,
+      "loss": 1.2529,
+      "step": 10709
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5294973108429112e-05,
+      "loss": 0.9395,
+      "step": 10710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529409387388891e-05,
+      "loss": 1.0493,
+      "step": 10711
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529321458248146e-05,
+      "loss": 1.0571,
+      "step": 10712
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529233523421621e-05,
+      "loss": 0.8047,
+      "step": 10713
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5291455829102603e-05,
+      "loss": 1.5645,
+      "step": 10714
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5290576367150087e-05,
+      "loss": 1.2217,
+      "step": 10715
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528969684836811e-05,
+      "loss": 1.4805,
+      "step": 10716
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528881727276611e-05,
+      "loss": 1.21,
+      "step": 10717
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5287937640353555e-05,
+      "loss": 1.3428,
+      "step": 10718
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5287057951139877e-05,
+      "loss": 0.9346,
+      "step": 10719
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5286178205134527e-05,
+      "loss": 0.8467,
+      "step": 10720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528529840234696e-05,
+      "loss": 1.1401,
+      "step": 10721
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5284418542786625e-05,
+      "loss": 0.7979,
+      "step": 10722
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5283538626462973e-05,
+      "loss": 1.1748,
+      "step": 10723
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5282658653385454e-05,
+      "loss": 1.0889,
+      "step": 10724
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5281778623563524e-05,
+      "loss": 1.5039,
+      "step": 10725
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5280898537006633e-05,
+      "loss": 1.0225,
+      "step": 10726
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5280018393724236e-05,
+      "loss": 1.127,
+      "step": 10727
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5279138193725788e-05,
+      "loss": 0.854,
+      "step": 10728
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5278257937020736e-05,
+      "loss": 0.8672,
+      "step": 10729
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527737762361855e-05,
+      "loss": 0.9639,
+      "step": 10730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527649725352867e-05,
+      "loss": 0.793,
+      "step": 10731
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5275616826760566e-05,
+      "loss": 1.3008,
+      "step": 10732
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5274736343323683e-05,
+      "loss": 1.1729,
+      "step": 10733
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5273855803227493e-05,
+      "loss": 1.2705,
+      "step": 10734
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527297520648144e-05,
+      "loss": 1.2012,
+      "step": 10735
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5272094553094997e-05,
+      "loss": 1.2324,
+      "step": 10736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527121384307761e-05,
+      "loss": 1.1831,
+      "step": 10737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527033307643875e-05,
+      "loss": 1.0469,
+      "step": 10738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526945225318787e-05,
+      "loss": 1.3311,
+      "step": 10739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526857137333444e-05,
+      "loss": 0.5205,
+      "step": 10740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526769043688791e-05,
+      "loss": 1.4805,
+      "step": 10741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526680944385776e-05,
+      "loss": 0.9951,
+      "step": 10742
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5265928394253436e-05,
+      "loss": 0.8291,
+      "step": 10743
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526504728808441e-05,
+      "loss": 1.0723,
+      "step": 10744
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5264166125360147e-05,
+      "loss": 1.3145,
+      "step": 10745
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5263284906090107e-05,
+      "loss": 0.8848,
+      "step": 10746
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5262403630283765e-05,
+      "loss": 1.1113,
+      "step": 10747
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5261522297950578e-05,
+      "loss": 0.8066,
+      "step": 10748
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5260640909100022e-05,
+      "loss": 0.9097,
+      "step": 10749
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5259759463741554e-05,
+      "loss": 1.1787,
+      "step": 10750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525887796188465e-05,
+      "loss": 1.3027,
+      "step": 10751
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5257996403538777e-05,
+      "loss": 1.1328,
+      "step": 10752
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5257114788713402e-05,
+      "loss": 0.7046,
+      "step": 10753
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5256233117418e-05,
+      "loss": 1.3291,
+      "step": 10754
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5255351389662034e-05,
+      "loss": 1.0703,
+      "step": 10755
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5254469605454983e-05,
+      "loss": 0.8198,
+      "step": 10756
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5253587764806313e-05,
+      "loss": 1.3027,
+      "step": 10757
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52527058677255e-05,
+      "loss": 1.2412,
+      "step": 10758
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5251823914222014e-05,
+      "loss": 0.9468,
+      "step": 10759
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5250941904305335e-05,
+      "loss": 0.8447,
+      "step": 10760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5250059837984927e-05,
+      "loss": 1.0273,
+      "step": 10761
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5249177715270276e-05,
+      "loss": 1.4219,
+      "step": 10762
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5248295536170847e-05,
+      "loss": 1.4453,
+      "step": 10763
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5247413300696123e-05,
+      "loss": 1.0166,
+      "step": 10764
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.524653100885558e-05,
+      "loss": 1.2363,
+      "step": 10765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.524564866065869e-05,
+      "loss": 1.3809,
+      "step": 10766
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5244766256114939e-05,
+      "loss": 1.0498,
+      "step": 10767
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5243883795233801e-05,
+      "loss": 0.9609,
+      "step": 10768
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5243001278024752e-05,
+      "loss": 0.9702,
+      "step": 10769
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5242118704497277e-05,
+      "loss": 1.3643,
+      "step": 10770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5241236074660854e-05,
+      "loss": 1.248,
+      "step": 10771
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5240353388524964e-05,
+      "loss": 1.415,
+      "step": 10772
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5239470646099088e-05,
+      "loss": 1.1309,
+      "step": 10773
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5238587847392712e-05,
+      "loss": 0.9072,
+      "step": 10774
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5237704992415311e-05,
+      "loss": 1.4531,
+      "step": 10775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5236822081176376e-05,
+      "loss": 1.2959,
+      "step": 10776
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5235939113685385e-05,
+      "loss": 0.7695,
+      "step": 10777
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523505608995183e-05,
+      "loss": 1.1318,
+      "step": 10778
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5234173009985185e-05,
+      "loss": 1.0879,
+      "step": 10779
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5233289873794947e-05,
+      "loss": 1.4883,
+      "step": 10780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5232406681390597e-05,
+      "loss": 1.2832,
+      "step": 10781
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5231523432781621e-05,
+      "loss": 1.1807,
+      "step": 10782
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230640127977507e-05,
+      "loss": 1.5469,
+      "step": 10783
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5229756766987745e-05,
+      "loss": 1.1445,
+      "step": 10784
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522887334982182e-05,
+      "loss": 1.209,
+      "step": 10785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5227989876489227e-05,
+      "loss": 1.167,
+      "step": 10786
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5227106346999454e-05,
+      "loss": 1.1846,
+      "step": 10787
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5226222761361989e-05,
+      "loss": 0.7783,
+      "step": 10788
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5225339119586327e-05,
+      "loss": 1.2051,
+      "step": 10789
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5224455421681951e-05,
+      "loss": 0.7725,
+      "step": 10790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5223571667658367e-05,
+      "loss": 1.1709,
+      "step": 10791
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5222687857525058e-05,
+      "loss": 0.9802,
+      "step": 10792
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5221803991291522e-05,
+      "loss": 0.915,
+      "step": 10793
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522092006896725e-05,
+      "loss": 1.1875,
+      "step": 10794
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5220036090561739e-05,
+      "loss": 1.3945,
+      "step": 10795
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5219152056084483e-05,
+      "loss": 1.0908,
+      "step": 10796
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5218267965544979e-05,
+      "loss": 1.4658,
+      "step": 10797
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5217383818952724e-05,
+      "loss": 1.4883,
+      "step": 10798
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5216499616317215e-05,
+      "loss": 0.8599,
+      "step": 10799
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5215615357647946e-05,
+      "loss": 0.707,
+      "step": 10800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.521473104295442e-05,
+      "loss": 0.834,
+      "step": 10801
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5213846672246139e-05,
+      "loss": 1.0039,
+      "step": 10802
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5212962245532592e-05,
+      "loss": 1.2188,
+      "step": 10803
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5212077762823291e-05,
+      "loss": 1.457,
+      "step": 10804
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5211193224127727e-05,
+      "loss": 1.0254,
+      "step": 10805
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5210308629455409e-05,
+      "loss": 1.0781,
+      "step": 10806
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5209423978815834e-05,
+      "loss": 1.2246,
+      "step": 10807
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5208539272218505e-05,
+      "loss": 0.9229,
+      "step": 10808
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5207654509672929e-05,
+      "loss": 1.1797,
+      "step": 10809
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5206769691188604e-05,
+      "loss": 1.3809,
+      "step": 10810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5205884816775039e-05,
+      "loss": 1.2051,
+      "step": 10811
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5204999886441738e-05,
+      "loss": 1.0508,
+      "step": 10812
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5204114900198206e-05,
+      "loss": 1.0195,
+      "step": 10813
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520322985805395e-05,
+      "loss": 1.2734,
+      "step": 10814
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5202344760018474e-05,
+      "loss": 1.0649,
+      "step": 10815
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520145960610129e-05,
+      "loss": 1.2012,
+      "step": 10816
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5200574396311905e-05,
+      "loss": 1.0107,
+      "step": 10817
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5199689130659822e-05,
+      "loss": 1.0547,
+      "step": 10818
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198803809154558e-05,
+      "loss": 0.96,
+      "step": 10819
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5197918431805618e-05,
+      "loss": 1.002,
+      "step": 10820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5197032998622513e-05,
+      "loss": 0.8955,
+      "step": 10821
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5196147509614756e-05,
+      "loss": 1.3096,
+      "step": 10822
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5195261964791854e-05,
+      "loss": 0.8013,
+      "step": 10823
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5194376364163327e-05,
+      "loss": 0.9131,
+      "step": 10824
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5193490707738683e-05,
+      "loss": 0.8076,
+      "step": 10825
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5192604995527433e-05,
+      "loss": 1.2773,
+      "step": 10826
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5191719227539095e-05,
+      "loss": 1.0967,
+      "step": 10827
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5190833403783183e-05,
+      "loss": 0.6221,
+      "step": 10828
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5189947524269208e-05,
+      "loss": 0.8013,
+      "step": 10829
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5189061589006693e-05,
+      "loss": 0.9248,
+      "step": 10830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5188175598005149e-05,
+      "loss": 0.7812,
+      "step": 10831
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5187289551274097e-05,
+      "loss": 0.9736,
+      "step": 10832
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518640344882305e-05,
+      "loss": 1.1011,
+      "step": 10833
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5185517290661526e-05,
+      "loss": 1.1299,
+      "step": 10834
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5184631076799053e-05,
+      "loss": 0.8442,
+      "step": 10835
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5183744807245135e-05,
+      "loss": 1.085,
+      "step": 10836
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182858482009306e-05,
+      "loss": 0.8906,
+      "step": 10837
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5181972101101083e-05,
+      "loss": 0.8887,
+      "step": 10838
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5181085664529982e-05,
+      "loss": 1.3955,
+      "step": 10839
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5180199172305527e-05,
+      "loss": 1.1201,
+      "step": 10840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5179312624437246e-05,
+      "loss": 1.2461,
+      "step": 10841
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5178426020934653e-05,
+      "loss": 0.9121,
+      "step": 10842
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5177539361807279e-05,
+      "loss": 0.958,
+      "step": 10843
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5176652647064645e-05,
+      "loss": 0.9805,
+      "step": 10844
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5175765876716276e-05,
+      "loss": 1.165,
+      "step": 10845
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5174879050771699e-05,
+      "loss": 1.2979,
+      "step": 10846
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173992169240433e-05,
+      "loss": 1.5703,
+      "step": 10847
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173105232132014e-05,
+      "loss": 0.9106,
+      "step": 10848
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5172218239455963e-05,
+      "loss": 1.1318,
+      "step": 10849
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5171331191221814e-05,
+      "loss": 1.0898,
+      "step": 10850
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5170444087439087e-05,
+      "loss": 0.9106,
+      "step": 10851
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5169556928117315e-05,
+      "loss": 1.0771,
+      "step": 10852
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5168669713266031e-05,
+      "loss": 1.4492,
+      "step": 10853
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5167782442894758e-05,
+      "loss": 1.333,
+      "step": 10854
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5166895117013035e-05,
+      "loss": 1.4473,
+      "step": 10855
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5166007735630386e-05,
+      "loss": 1.1172,
+      "step": 10856
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5165120298756347e-05,
+      "loss": 0.8613,
+      "step": 10857
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5164232806400447e-05,
+      "loss": 1.4219,
+      "step": 10858
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5163345258572224e-05,
+      "loss": 0.9839,
+      "step": 10859
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5162457655281207e-05,
+      "loss": 0.8291,
+      "step": 10860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5161569996536938e-05,
+      "loss": 0.999,
+      "step": 10861
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5160682282348939e-05,
+      "loss": 0.9031,
+      "step": 10862
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5159794512726755e-05,
+      "loss": 0.9199,
+      "step": 10863
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5158906687679924e-05,
+      "loss": 1.4219,
+      "step": 10864
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5158018807217976e-05,
+      "loss": 0.8408,
+      "step": 10865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5157130871350451e-05,
+      "loss": 0.9443,
+      "step": 10866
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5156242880086884e-05,
+      "loss": 1.1953,
+      "step": 10867
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5155354833436823e-05,
+      "loss": 1.2021,
+      "step": 10868
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5154466731409798e-05,
+      "loss": 0.665,
+      "step": 10869
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5153578574015348e-05,
+      "loss": 1.1709,
+      "step": 10870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515269036126302e-05,
+      "loss": 1.0093,
+      "step": 10871
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5151802093162348e-05,
+      "loss": 1.0771,
+      "step": 10872
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515091376972288e-05,
+      "loss": 1.5312,
+      "step": 10873
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5150025390954153e-05,
+      "loss": 1.2861,
+      "step": 10874
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5149136956865711e-05,
+      "loss": 1.2842,
+      "step": 10875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5148248467467097e-05,
+      "loss": 0.7061,
+      "step": 10876
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5147359922767862e-05,
+      "loss": 0.8154,
+      "step": 10877
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5146471322777534e-05,
+      "loss": 1.1367,
+      "step": 10878
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5145582667505676e-05,
+      "loss": 0.9951,
+      "step": 10879
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514469395696182e-05,
+      "loss": 1.207,
+      "step": 10880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514380519115552e-05,
+      "loss": 0.958,
+      "step": 10881
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5142916370096324e-05,
+      "loss": 1.252,
+      "step": 10882
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514202749379377e-05,
+      "loss": 1.2891,
+      "step": 10883
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5141138562257416e-05,
+      "loss": 1.1826,
+      "step": 10884
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5140249575496804e-05,
+      "loss": 0.9185,
+      "step": 10885
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5139360533521488e-05,
+      "loss": 1.1895,
+      "step": 10886
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5138471436341013e-05,
+      "loss": 0.7979,
+      "step": 10887
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513758228396493e-05,
+      "loss": 1.1299,
+      "step": 10888
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5136693076402798e-05,
+      "loss": 0.6152,
+      "step": 10889
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5135803813664159e-05,
+      "loss": 1.5352,
+      "step": 10890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5134914495758568e-05,
+      "loss": 1.5957,
+      "step": 10891
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5134025122695581e-05,
+      "loss": 1.041,
+      "step": 10892
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5133135694484744e-05,
+      "loss": 1.2295,
+      "step": 10893
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513224621113562e-05,
+      "loss": 1.1055,
+      "step": 10894
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5131356672657758e-05,
+      "loss": 1.2695,
+      "step": 10895
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5130467079060712e-05,
+      "loss": 1.2739,
+      "step": 10896
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512957743035404e-05,
+      "loss": 1.0005,
+      "step": 10897
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5128687726547297e-05,
+      "loss": 1.1055,
+      "step": 10898
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5127797967650044e-05,
+      "loss": 0.8271,
+      "step": 10899
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5126908153671834e-05,
+      "loss": 1.0762,
+      "step": 10900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5126018284622227e-05,
+      "loss": 0.834,
+      "step": 10901
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5125128360510778e-05,
+      "loss": 1.0088,
+      "step": 10902
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5124238381347054e-05,
+      "loss": 0.8594,
+      "step": 10903
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5123348347140606e-05,
+      "loss": 0.6614,
+      "step": 10904
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5122458257901002e-05,
+      "loss": 1.4473,
+      "step": 10905
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5121568113637796e-05,
+      "loss": 1.2783,
+      "step": 10906
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512067791436056e-05,
+      "loss": 1.3174,
+      "step": 10907
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5119787660078844e-05,
+      "loss": 1.0195,
+      "step": 10908
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5118897350802212e-05,
+      "loss": 0.8027,
+      "step": 10909
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511800698654024e-05,
+      "loss": 0.9282,
+      "step": 10910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5117116567302478e-05,
+      "loss": 1.291,
+      "step": 10911
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5116226093098503e-05,
+      "loss": 1.209,
+      "step": 10912
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5115335563937868e-05,
+      "loss": 0.9814,
+      "step": 10913
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5114444979830145e-05,
+      "loss": 1.2227,
+      "step": 10914
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5113554340784897e-05,
+      "loss": 1.5215,
+      "step": 10915
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5112663646811696e-05,
+      "loss": 1.0557,
+      "step": 10916
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5111772897920105e-05,
+      "loss": 1.417,
+      "step": 10917
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5110882094119696e-05,
+      "loss": 0.729,
+      "step": 10918
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5109991235420034e-05,
+      "loss": 0.9961,
+      "step": 10919
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5109100321830693e-05,
+      "loss": 1.1016,
+      "step": 10920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5108209353361236e-05,
+      "loss": 1.3125,
+      "step": 10921
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5107318330021237e-05,
+      "loss": 0.8662,
+      "step": 10922
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.510642725182027e-05,
+      "loss": 1.1523,
+      "step": 10923
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51055361187679e-05,
+      "loss": 1.3223,
+      "step": 10924
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5104644930873708e-05,
+      "loss": 1.1543,
+      "step": 10925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.510375368814726e-05,
+      "loss": 1.1982,
+      "step": 10926
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5102862390598132e-05,
+      "loss": 0.9395,
+      "step": 10927
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5101971038235897e-05,
+      "loss": 1.4863,
+      "step": 10928
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5101079631070128e-05,
+      "loss": 0.9834,
+      "step": 10929
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5100188169110403e-05,
+      "loss": 1.0352,
+      "step": 10930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5099296652366297e-05,
+      "loss": 0.9844,
+      "step": 10931
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5098405080847387e-05,
+      "loss": 1.1338,
+      "step": 10932
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5097513454563251e-05,
+      "loss": 0.8916,
+      "step": 10933
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5096621773523462e-05,
+      "loss": 1.4795,
+      "step": 10934
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5095730037737599e-05,
+      "loss": 1.2334,
+      "step": 10935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5094838247215248e-05,
+      "loss": 1.0469,
+      "step": 10936
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5093946401965976e-05,
+      "loss": 1.2354,
+      "step": 10937
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5093054501999376e-05,
+      "loss": 1.0791,
+      "step": 10938
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509216254732502e-05,
+      "loss": 0.8755,
+      "step": 10939
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5091270537952492e-05,
+      "loss": 1.3301,
+      "step": 10940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509037847389137e-05,
+      "loss": 1.3184,
+      "step": 10941
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508948635515124e-05,
+      "loss": 1.1338,
+      "step": 10942
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5088594181741685e-05,
+      "loss": 0.9932,
+      "step": 10943
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5087701953672285e-05,
+      "loss": 1.1855,
+      "step": 10944
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508680967095263e-05,
+      "loss": 1.2617,
+      "step": 10945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5085917333592299e-05,
+      "loss": 1.4033,
+      "step": 10946
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5085024941600877e-05,
+      "loss": 0.9229,
+      "step": 10947
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5084132494987956e-05,
+      "loss": 1.04,
+      "step": 10948
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5083239993763116e-05,
+      "loss": 0.9863,
+      "step": 10949
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5082347437935944e-05,
+      "loss": 1.5254,
+      "step": 10950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5081454827516036e-05,
+      "loss": 0.9492,
+      "step": 10951
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5080562162512965e-05,
+      "loss": 1.1934,
+      "step": 10952
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5079669442936333e-05,
+      "loss": 0.916,
+      "step": 10953
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5078776668795727e-05,
+      "loss": 0.915,
+      "step": 10954
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5077883840100728e-05,
+      "loss": 1.2852,
+      "step": 10955
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5076990956860939e-05,
+      "loss": 1.1123,
+      "step": 10956
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5076098019085943e-05,
+      "loss": 1.2871,
+      "step": 10957
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5075205026785331e-05,
+      "loss": 1.0195,
+      "step": 10958
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5074311979968701e-05,
+      "loss": 1.5791,
+      "step": 10959
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5073418878645643e-05,
+      "loss": 1.125,
+      "step": 10960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5072525722825745e-05,
+      "loss": 1.3018,
+      "step": 10961
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5071632512518612e-05,
+      "loss": 0.9336,
+      "step": 10962
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507073924773383e-05,
+      "loss": 0.9229,
+      "step": 10963
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5069845928480997e-05,
+      "loss": 0.8125,
+      "step": 10964
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5068952554769707e-05,
+      "loss": 0.7979,
+      "step": 10965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5068059126609557e-05,
+      "loss": 1.1934,
+      "step": 10966
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5067165644010149e-05,
+      "loss": 0.9473,
+      "step": 10967
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506627210698107e-05,
+      "loss": 0.8906,
+      "step": 10968
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5065378515531927e-05,
+      "loss": 1.0352,
+      "step": 10969
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5064484869672318e-05,
+      "loss": 0.9697,
+      "step": 10970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5063591169411836e-05,
+      "loss": 1.2002,
+      "step": 10971
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5062697414760085e-05,
+      "loss": 0.772,
+      "step": 10972
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5061803605726666e-05,
+      "loss": 0.834,
+      "step": 10973
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5060909742321177e-05,
+      "loss": 1.2441,
+      "step": 10974
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5060015824553224e-05,
+      "loss": 0.8535,
+      "step": 10975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5059121852432408e-05,
+      "loss": 1.168,
+      "step": 10976
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5058227825968325e-05,
+      "loss": 0.7197,
+      "step": 10977
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.505733374517059e-05,
+      "loss": 1.1499,
+      "step": 10978
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5056439610048799e-05,
+      "loss": 0.834,
+      "step": 10979
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5055545420612557e-05,
+      "loss": 1.5508,
+      "step": 10980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5054651176871468e-05,
+      "loss": 0.792,
+      "step": 10981
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5053756878835144e-05,
+      "loss": 1.3955,
+      "step": 10982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5052862526513184e-05,
+      "loss": 0.9302,
+      "step": 10983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5051968119915201e-05,
+      "loss": 1.0918,
+      "step": 10984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5051073659050795e-05,
+      "loss": 0.8794,
+      "step": 10985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.505017914392958e-05,
+      "loss": 1.0859,
+      "step": 10986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5049284574561162e-05,
+      "loss": 1.1299,
+      "step": 10987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5048389950955153e-05,
+      "loss": 0.4597,
+      "step": 10988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5047495273121162e-05,
+      "loss": 1.2188,
+      "step": 10989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5046600541068795e-05,
+      "loss": 1.2344,
+      "step": 10990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5045705754807664e-05,
+      "loss": 0.9316,
+      "step": 10991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5044810914347387e-05,
+      "loss": 1.166,
+      "step": 10992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5043916019697569e-05,
+      "loss": 0.8203,
+      "step": 10993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5043021070867825e-05,
+      "loss": 1.0625,
+      "step": 10994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5042126067867768e-05,
+      "loss": 1.1104,
+      "step": 10995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5041231010707012e-05,
+      "loss": 1.1123,
+      "step": 10996
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5040335899395175e-05,
+      "loss": 0.894,
+      "step": 10997
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5039440733941863e-05,
+      "loss": 0.8701,
+      "step": 10998
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5038545514356696e-05,
+      "loss": 0.7334,
+      "step": 10999
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5037650240649297e-05,
+      "loss": 1.0215,
+      "step": 11000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5036754912829275e-05,
+      "loss": 1.373,
+      "step": 11001
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5035859530906247e-05,
+      "loss": 1.1533,
+      "step": 11002
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5034964094889834e-05,
+      "loss": 1.1748,
+      "step": 11003
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5034068604789654e-05,
+      "loss": 0.9082,
+      "step": 11004
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5033173060615322e-05,
+      "loss": 1.2734,
+      "step": 11005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5032277462376467e-05,
+      "loss": 1.1318,
+      "step": 11006
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5031381810082696e-05,
+      "loss": 1.1475,
+      "step": 11007
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5030486103743643e-05,
+      "loss": 1.1387,
+      "step": 11008
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502959034336892e-05,
+      "loss": 1.2393,
+      "step": 11009
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5028694528968153e-05,
+      "loss": 0.8047,
+      "step": 11010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5027798660550965e-05,
+      "loss": 0.9023,
+      "step": 11011
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5026902738126976e-05,
+      "loss": 1.1455,
+      "step": 11012
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5026006761705816e-05,
+      "loss": 1.0884,
+      "step": 11013
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5025110731297101e-05,
+      "loss": 1.1221,
+      "step": 11014
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5024214646910457e-05,
+      "loss": 1.207,
+      "step": 11015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5023318508555519e-05,
+      "loss": 1.0664,
+      "step": 11016
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.50224223162419e-05,
+      "loss": 1.1406,
+      "step": 11017
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5021526069979232e-05,
+      "loss": 1.0645,
+      "step": 11018
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5020629769777146e-05,
+      "loss": 0.9229,
+      "step": 11019
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5019733415645266e-05,
+      "loss": 1.2275,
+      "step": 11020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501883700759322e-05,
+      "loss": 1.0742,
+      "step": 11021
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5017940545630639e-05,
+      "loss": 1.3242,
+      "step": 11022
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501704402976715e-05,
+      "loss": 1.085,
+      "step": 11023
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5016147460012386e-05,
+      "loss": 1.1191,
+      "step": 11024
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5015250836375973e-05,
+      "loss": 1.3057,
+      "step": 11025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5014354158867549e-05,
+      "loss": 1.0723,
+      "step": 11026
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5013457427496736e-05,
+      "loss": 0.9678,
+      "step": 11027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501256064227318e-05,
+      "loss": 1.3516,
+      "step": 11028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5011663803206501e-05,
+      "loss": 0.8379,
+      "step": 11029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5010766910306338e-05,
+      "loss": 1.5039,
+      "step": 11030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5009869963582326e-05,
+      "loss": 0.9731,
+      "step": 11031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008972963044097e-05,
+      "loss": 1.3164,
+      "step": 11032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008075908701289e-05,
+      "loss": 0.6206,
+      "step": 11033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5007178800563535e-05,
+      "loss": 0.9209,
+      "step": 11034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5006281638640474e-05,
+      "loss": 0.7485,
+      "step": 11035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.500538442294174e-05,
+      "loss": 1.3809,
+      "step": 11036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5004487153476976e-05,
+      "loss": 0.7305,
+      "step": 11037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5003589830255813e-05,
+      "loss": 0.873,
+      "step": 11038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5002692453287899e-05,
+      "loss": 1.0156,
+      "step": 11039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5001795022582863e-05,
+      "loss": 1.3779,
+      "step": 11040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5000897538150353e-05,
+      "loss": 1.3594,
+      "step": 11041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.9023,
+      "step": 11042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4999102408141457e-05,
+      "loss": 0.9785,
+      "step": 11043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4998204762584356e-05,
+      "loss": 0.9697,
+      "step": 11044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4997307063338346e-05,
+      "loss": 1.3018,
+      "step": 11045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4996409310413063e-05,
+      "loss": 1.3389,
+      "step": 11046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4995511503818156e-05,
+      "loss": 0.8682,
+      "step": 11047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4994613643563268e-05,
+      "loss": 1.4502,
+      "step": 11048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4993715729658041e-05,
+      "loss": 0.7634,
+      "step": 11049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4992817762112122e-05,
+      "loss": 1.2139,
+      "step": 11050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4991919740935155e-05,
+      "loss": 0.8584,
+      "step": 11051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4991021666136786e-05,
+      "loss": 1.1084,
+      "step": 11052
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4990123537726664e-05,
+      "loss": 0.9502,
+      "step": 11053
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4989225355714436e-05,
+      "loss": 1.082,
+      "step": 11054
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4988327120109752e-05,
+      "loss": 1.1367,
+      "step": 11055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4987428830922253e-05,
+      "loss": 0.8633,
+      "step": 11056
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4986530488161596e-05,
+      "loss": 1.0859,
+      "step": 11057
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4985632091837428e-05,
+      "loss": 1.0332,
+      "step": 11058
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4984733641959399e-05,
+      "loss": 1.2354,
+      "step": 11059
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4983835138537158e-05,
+      "loss": 0.9307,
+      "step": 11060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.498293658158036e-05,
+      "loss": 1.3604,
+      "step": 11061
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4982037971098654e-05,
+      "loss": 0.9072,
+      "step": 11062
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4981139307101693e-05,
+      "loss": 1.0405,
+      "step": 11063
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4980240589599132e-05,
+      "loss": 0.7229,
+      "step": 11064
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497934181860062e-05,
+      "loss": 1.1416,
+      "step": 11065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4978442994115821e-05,
+      "loss": 1.2217,
+      "step": 11066
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4977544116154379e-05,
+      "loss": 1.0977,
+      "step": 11067
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4976645184725958e-05,
+      "loss": 0.9727,
+      "step": 11068
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4975746199840204e-05,
+      "loss": 0.8428,
+      "step": 11069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4974847161506787e-05,
+      "loss": 1.2383,
+      "step": 11070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497394806973535e-05,
+      "loss": 0.8955,
+      "step": 11071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4973048924535562e-05,
+      "loss": 0.7319,
+      "step": 11072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4972149725917076e-05,
+      "loss": 0.9736,
+      "step": 11073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4971250473889552e-05,
+      "loss": 1.3438,
+      "step": 11074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4970351168462647e-05,
+      "loss": 0.9443,
+      "step": 11075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4969451809646023e-05,
+      "loss": 0.8389,
+      "step": 11076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4968552397449341e-05,
+      "loss": 1.1719,
+      "step": 11077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4967652931882261e-05,
+      "loss": 1.0645,
+      "step": 11078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4966753412954449e-05,
+      "loss": 1.166,
+      "step": 11079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4965853840675562e-05,
+      "loss": 1.2656,
+      "step": 11080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4964954215055266e-05,
+      "loss": 1.0352,
+      "step": 11081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4964054536103221e-05,
+      "loss": 1.4414,
+      "step": 11082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4963154803829093e-05,
+      "loss": 1.0752,
+      "step": 11083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4962255018242548e-05,
+      "loss": 1.1426,
+      "step": 11084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4961355179353252e-05,
+      "loss": 1.3086,
+      "step": 11085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4960455287170868e-05,
+      "loss": 1.2021,
+      "step": 11086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4959555341705061e-05,
+      "loss": 1.1553,
+      "step": 11087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.49586553429655e-05,
+      "loss": 1.1387,
+      "step": 11088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4957755290961853e-05,
+      "loss": 0.9541,
+      "step": 11089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4956855185703787e-05,
+      "loss": 1.4854,
+      "step": 11090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4955955027200973e-05,
+      "loss": 1.3057,
+      "step": 11091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4955054815463077e-05,
+      "loss": 0.8926,
+      "step": 11092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.495415455049977e-05,
+      "loss": 1.2842,
+      "step": 11093
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4953254232320724e-05,
+      "loss": 1.0762,
+      "step": 11094
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4952353860935605e-05,
+      "loss": 0.9033,
+      "step": 11095
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.495145343635409e-05,
+      "loss": 1.1875,
+      "step": 11096
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4950552958585847e-05,
+      "loss": 1.4111,
+      "step": 11097
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494965242764055e-05,
+      "loss": 1.083,
+      "step": 11098
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4948751843527875e-05,
+      "loss": 1.0596,
+      "step": 11099
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494785120625749e-05,
+      "loss": 1.1396,
+      "step": 11100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946950515839078e-05,
+      "loss": 1.0342,
+      "step": 11101
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946049772282303e-05,
+      "loss": 0.9727,
+      "step": 11102
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494514897559685e-05,
+      "loss": 1.1094,
+      "step": 11103
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4944248125792388e-05,
+      "loss": 1.2041,
+      "step": 11104
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4943347222878597e-05,
+      "loss": 1.3008,
+      "step": 11105
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4942446266865154e-05,
+      "loss": 0.9727,
+      "step": 11106
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4941545257761739e-05,
+      "loss": 1.4922,
+      "step": 11107
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4940644195578025e-05,
+      "loss": 0.8496,
+      "step": 11108
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4939743080323692e-05,
+      "loss": 1.0396,
+      "step": 11109
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4938841912008423e-05,
+      "loss": 1.2334,
+      "step": 11110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4937940690641899e-05,
+      "loss": 0.8232,
+      "step": 11111
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4937039416233796e-05,
+      "loss": 1.208,
+      "step": 11112
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4936138088793795e-05,
+      "loss": 1.0879,
+      "step": 11113
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4935236708331582e-05,
+      "loss": 1.208,
+      "step": 11114
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4934335274856832e-05,
+      "loss": 1.0361,
+      "step": 11115
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493343378837924e-05,
+      "loss": 0.8896,
+      "step": 11116
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4932532248908477e-05,
+      "loss": 0.9404,
+      "step": 11117
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4931630656454234e-05,
+      "loss": 1.0518,
+      "step": 11118
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4930729011026196e-05,
+      "loss": 0.7812,
+      "step": 11119
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4929827312634043e-05,
+      "loss": 1.1895,
+      "step": 11120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4928925561287464e-05,
+      "loss": 0.9453,
+      "step": 11121
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4928023756996147e-05,
+      "loss": 0.9307,
+      "step": 11122
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4927121899769775e-05,
+      "loss": 0.9688,
+      "step": 11123
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4926219989618038e-05,
+      "loss": 0.8467,
+      "step": 11124
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4925318026550626e-05,
+      "loss": 1.1309,
+      "step": 11125
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4924416010577218e-05,
+      "loss": 0.8555,
+      "step": 11126
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4923513941707516e-05,
+      "loss": 1.2598,
+      "step": 11127
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.49226118199512e-05,
+      "loss": 1.2192,
+      "step": 11128
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4921709645317968e-05,
+      "loss": 1.2993,
+      "step": 11129
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4920807417817504e-05,
+      "loss": 1.2471,
+      "step": 11130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4919905137459503e-05,
+      "loss": 1.2139,
+      "step": 11131
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4919002804253654e-05,
+      "loss": 1.1426,
+      "step": 11132
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4918100418209655e-05,
+      "loss": 1.4707,
+      "step": 11133
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4917197979337195e-05,
+      "loss": 1.0637,
+      "step": 11134
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4916295487645966e-05,
+      "loss": 0.8066,
+      "step": 11135
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4915392943145668e-05,
+      "loss": 1.165,
+      "step": 11136
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491449034584599e-05,
+      "loss": 1.1309,
+      "step": 11137
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4913587695756632e-05,
+      "loss": 1.0518,
+      "step": 11138
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4912684992887286e-05,
+      "loss": 1.1777,
+      "step": 11139
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4911782237247651e-05,
+      "loss": 1.0664,
+      "step": 11140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4910879428847422e-05,
+      "loss": 0.9863,
+      "step": 11141
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4909976567696303e-05,
+      "loss": 0.8901,
+      "step": 11142
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4909073653803982e-05,
+      "loss": 1.0918,
+      "step": 11143
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4908170687180163e-05,
+      "loss": 1.3135,
+      "step": 11144
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490726766783455e-05,
+      "loss": 1.3877,
+      "step": 11145
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4906364595776834e-05,
+      "loss": 1.3047,
+      "step": 11146
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4905461471016722e-05,
+      "loss": 1.5996,
+      "step": 11147
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4904558293563913e-05,
+      "loss": 0.7192,
+      "step": 11148
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4903655063428107e-05,
+      "loss": 1.0537,
+      "step": 11149
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490275178061901e-05,
+      "loss": 1.0967,
+      "step": 11150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4901848445146319e-05,
+      "loss": 0.8491,
+      "step": 11151
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4900945057019742e-05,
+      "loss": 1.0918,
+      "step": 11152
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4900041616248982e-05,
+      "loss": 1.2363,
+      "step": 11153
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4899138122843742e-05,
+      "loss": 1.0439,
+      "step": 11154
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4898234576813728e-05,
+      "loss": 0.9365,
+      "step": 11155
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4897330978168648e-05,
+      "loss": 1.084,
+      "step": 11156
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4896427326918201e-05,
+      "loss": 1.1953,
+      "step": 11157
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4895523623072103e-05,
+      "loss": 1.1133,
+      "step": 11158
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4894619866640052e-05,
+      "loss": 0.7043,
+      "step": 11159
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4893716057631766e-05,
+      "loss": 1.2207,
+      "step": 11160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4892812196056945e-05,
+      "loss": 0.7119,
+      "step": 11161
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48919082819253e-05,
+      "loss": 0.8491,
+      "step": 11162
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489100431524654e-05,
+      "loss": 0.9941,
+      "step": 11163
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4890100296030378e-05,
+      "loss": 1.0859,
+      "step": 11164
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4889196224286524e-05,
+      "loss": 1.5146,
+      "step": 11165
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4888292100024688e-05,
+      "loss": 1.5879,
+      "step": 11166
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4887387923254581e-05,
+      "loss": 0.9102,
+      "step": 11167
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4886483693985916e-05,
+      "loss": 1.0547,
+      "step": 11168
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4885579412228409e-05,
+      "loss": 1.0186,
+      "step": 11169
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4884675077991766e-05,
+      "loss": 0.7783,
+      "step": 11170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488377069128571e-05,
+      "loss": 0.9697,
+      "step": 11171
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488286625211995e-05,
+      "loss": 1.2363,
+      "step": 11172
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4881961760504202e-05,
+      "loss": 1.4863,
+      "step": 11173
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4881057216448183e-05,
+      "loss": 0.9785,
+      "step": 11174
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4880152619961607e-05,
+      "loss": 1.1387,
+      "step": 11175
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4879247971054193e-05,
+      "loss": 1.1582,
+      "step": 11176
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487834326973566e-05,
+      "loss": 1.2324,
+      "step": 11177
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487743851601572e-05,
+      "loss": 1.4121,
+      "step": 11178
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4876533709904098e-05,
+      "loss": 1.0635,
+      "step": 11179
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4875628851410509e-05,
+      "loss": 1.1729,
+      "step": 11180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4874723940544674e-05,
+      "loss": 1.3389,
+      "step": 11181
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4873818977316314e-05,
+      "loss": 1.5293,
+      "step": 11182
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4872913961735149e-05,
+      "loss": 1.0938,
+      "step": 11183
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4872008893810904e-05,
+      "loss": 1.228,
+      "step": 11184
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487110377355329e-05,
+      "loss": 0.9326,
+      "step": 11185
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4870198600972043e-05,
+      "loss": 1.1528,
+      "step": 11186
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4869293376076876e-05,
+      "loss": 1.04,
+      "step": 11187
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4868388098877517e-05,
+      "loss": 1.0498,
+      "step": 11188
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4867482769383692e-05,
+      "loss": 0.9346,
+      "step": 11189
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4866577387605119e-05,
+      "loss": 1.0054,
+      "step": 11190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4865671953551533e-05,
+      "loss": 0.7988,
+      "step": 11191
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.486476646723265e-05,
+      "loss": 1.3369,
+      "step": 11192
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4863860928658203e-05,
+      "loss": 1.0889,
+      "step": 11193
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4862955337837917e-05,
+      "loss": 1.0098,
+      "step": 11194
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4862049694781518e-05,
+      "loss": 1.1836,
+      "step": 11195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4861143999498735e-05,
+      "loss": 1.0752,
+      "step": 11196
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4860238251999301e-05,
+      "loss": 0.9961,
+      "step": 11197
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4859332452292937e-05,
+      "loss": 1.0508,
+      "step": 11198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4858426600389377e-05,
+      "loss": 0.689,
+      "step": 11199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4857520696298355e-05,
+      "loss": 1.0771,
+      "step": 11200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4856614740029595e-05,
+      "loss": 1.1016,
+      "step": 11201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4855708731592835e-05,
+      "loss": 1.1152,
+      "step": 11202
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48548026709978e-05,
+      "loss": 1.082,
+      "step": 11203
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485389655825423e-05,
+      "loss": 1.1367,
+      "step": 11204
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4852990393371853e-05,
+      "loss": 1.1748,
+      "step": 11205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4852084176360404e-05,
+      "loss": 0.7383,
+      "step": 11206
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4851177907229617e-05,
+      "loss": 1.0615,
+      "step": 11207
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4850271585989227e-05,
+      "loss": 1.2617,
+      "step": 11208
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484936521264897e-05,
+      "loss": 1.0947,
+      "step": 11209
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4848458787218583e-05,
+      "loss": 0.9521,
+      "step": 11210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48475523097078e-05,
+      "loss": 0.7407,
+      "step": 11211
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484664578012636e-05,
+      "loss": 0.9209,
+      "step": 11212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4845739198484e-05,
+      "loss": 1.3105,
+      "step": 11213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4844832564790454e-05,
+      "loss": 1.0181,
+      "step": 11214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484392587905547e-05,
+      "loss": 1.2383,
+      "step": 11215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4843019141288777e-05,
+      "loss": 1.0352,
+      "step": 11216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4842112351500124e-05,
+      "loss": 0.8975,
+      "step": 11217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4841205509699243e-05,
+      "loss": 1.2852,
+      "step": 11218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4840298615895883e-05,
+      "loss": 0.958,
+      "step": 11219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4839391670099782e-05,
+      "loss": 1.1299,
+      "step": 11220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4838484672320679e-05,
+      "loss": 1.0479,
+      "step": 11221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483757762256832e-05,
+      "loss": 0.9248,
+      "step": 11222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4836670520852448e-05,
+      "loss": 0.8003,
+      "step": 11223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4835763367182804e-05,
+      "loss": 0.9932,
+      "step": 11224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4834856161569138e-05,
+      "loss": 1.3955,
+      "step": 11225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4833948904021192e-05,
+      "loss": 1.0684,
+      "step": 11226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4833041594548708e-05,
+      "loss": 1.1963,
+      "step": 11227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483213423316144e-05,
+      "loss": 0.8257,
+      "step": 11228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4831226819869123e-05,
+      "loss": 0.8672,
+      "step": 11229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4830319354681516e-05,
+      "loss": 1.1797,
+      "step": 11230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4829411837608358e-05,
+      "loss": 0.812,
+      "step": 11231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4828504268659403e-05,
+      "loss": 0.7773,
+      "step": 11232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4827596647844396e-05,
+      "loss": 0.9814,
+      "step": 11233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4826688975173085e-05,
+      "loss": 0.7927,
+      "step": 11234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4825781250655227e-05,
+      "loss": 0.8525,
+      "step": 11235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4824873474300565e-05,
+      "loss": 0.8672,
+      "step": 11236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4823965646118855e-05,
+      "loss": 1.0459,
+      "step": 11237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4823057766119847e-05,
+      "loss": 1.085,
+      "step": 11238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4822149834313288e-05,
+      "loss": 1.0564,
+      "step": 11239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.482124185070894e-05,
+      "loss": 1.0669,
+      "step": 11240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4820333815316552e-05,
+      "loss": 1.3242,
+      "step": 11241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4819425728145875e-05,
+      "loss": 1.3242,
+      "step": 11242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4818517589206664e-05,
+      "loss": 1.1807,
+      "step": 11243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4817609398508681e-05,
+      "loss": 0.7827,
+      "step": 11244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4816701156061673e-05,
+      "loss": 0.8477,
+      "step": 11245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48157928618754e-05,
+      "loss": 1.083,
+      "step": 11246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4814884515959614e-05,
+      "loss": 1.5098,
+      "step": 11247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481397611832408e-05,
+      "loss": 1.3965,
+      "step": 11248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4813067668978551e-05,
+      "loss": 0.958,
+      "step": 11249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4812159167932784e-05,
+      "loss": 0.9688,
+      "step": 11250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481125061519654e-05,
+      "loss": 1.1836,
+      "step": 11251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4810342010779578e-05,
+      "loss": 1.3125,
+      "step": 11252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4809433354691658e-05,
+      "loss": 1.3076,
+      "step": 11253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.480852464694254e-05,
+      "loss": 1.1631,
+      "step": 11254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4807615887541986e-05,
+      "loss": 1.1387,
+      "step": 11255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4806707076499756e-05,
+      "loss": 0.7988,
+      "step": 11256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4805798213825613e-05,
+      "loss": 1.3477,
+      "step": 11257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.480488929952932e-05,
+      "loss": 1.0059,
+      "step": 11258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.480398033362064e-05,
+      "loss": 1.2422,
+      "step": 11259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4803071316109333e-05,
+      "loss": 1.5801,
+      "step": 11260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4802162247005171e-05,
+      "loss": 1.127,
+      "step": 11261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4801253126317913e-05,
+      "loss": 1.2715,
+      "step": 11262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4800343954057328e-05,
+      "loss": 0.877,
+      "step": 11263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4799434730233178e-05,
+      "loss": 1.0176,
+      "step": 11264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4798525454855233e-05,
+      "loss": 0.9688,
+      "step": 11265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797616127933259e-05,
+      "loss": 0.8442,
+      "step": 11266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4796706749477023e-05,
+      "loss": 1.2051,
+      "step": 11267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4795797319496293e-05,
+      "loss": 1.2939,
+      "step": 11268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4794887838000839e-05,
+      "loss": 1.1758,
+      "step": 11269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.479397830500043e-05,
+      "loss": 1.3853,
+      "step": 11270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4793068720504837e-05,
+      "loss": 1.2002,
+      "step": 11271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4792159084523832e-05,
+      "loss": 0.7559,
+      "step": 11272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4791249397067176e-05,
+      "loss": 1.0166,
+      "step": 11273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4790339658144652e-05,
+      "loss": 1.2441,
+      "step": 11274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4789429867766027e-05,
+      "loss": 1.1162,
+      "step": 11275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4788520025941076e-05,
+      "loss": 1.1494,
+      "step": 11276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4787610132679569e-05,
+      "loss": 0.6343,
+      "step": 11277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478670018799128e-05,
+      "loss": 0.9648,
+      "step": 11278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785790191885988e-05,
+      "loss": 1.0771,
+      "step": 11279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478488014437346e-05,
+      "loss": 0.7822,
+      "step": 11280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4783970045463481e-05,
+      "loss": 1.1191,
+      "step": 11281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4783059895165818e-05,
+      "loss": 1.249,
+      "step": 11282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4782149693490253e-05,
+      "loss": 0.9258,
+      "step": 11283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4781239440446562e-05,
+      "loss": 1.3496,
+      "step": 11284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4780329136044522e-05,
+      "loss": 1.2891,
+      "step": 11285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477941878029391e-05,
+      "loss": 0.8274,
+      "step": 11286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4778508373204508e-05,
+      "loss": 1.0596,
+      "step": 11287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4777597914786089e-05,
+      "loss": 1.2119,
+      "step": 11288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477668740504844e-05,
+      "loss": 1.3516,
+      "step": 11289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477577684400134e-05,
+      "loss": 1.3926,
+      "step": 11290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4774866231654563e-05,
+      "loss": 0.937,
+      "step": 11291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4773955568017902e-05,
+      "loss": 1.0322,
+      "step": 11292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4773044853101128e-05,
+      "loss": 0.9775,
+      "step": 11293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4772134086914032e-05,
+      "loss": 1.2021,
+      "step": 11294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477122326946639e-05,
+      "loss": 1.2832,
+      "step": 11295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4770312400767992e-05,
+      "loss": 1.3027,
+      "step": 11296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4769401480828617e-05,
+      "loss": 1.2305,
+      "step": 11297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4768490509658053e-05,
+      "loss": 0.8838,
+      "step": 11298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4767579487266086e-05,
+      "loss": 1.1079,
+      "step": 11299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4766668413662497e-05,
+      "loss": 0.9238,
+      "step": 11300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4765757288857079e-05,
+      "loss": 0.8867,
+      "step": 11301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4764846112859614e-05,
+      "loss": 1.6348,
+      "step": 11302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4763934885679893e-05,
+      "loss": 1.3945,
+      "step": 11303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.47630236073277e-05,
+      "loss": 0.9844,
+      "step": 11304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4762112277812827e-05,
+      "loss": 1.1016,
+      "step": 11305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4761200897145063e-05,
+      "loss": 1.4219,
+      "step": 11306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4760289465334197e-05,
+      "loss": 1.2236,
+      "step": 11307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4759377982390017e-05,
+      "loss": 1.248,
+      "step": 11308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4758466448322318e-05,
+      "loss": 1.2773,
+      "step": 11309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4757554863140886e-05,
+      "loss": 1.0107,
+      "step": 11310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4756643226855521e-05,
+      "loss": 0.5869,
+      "step": 11311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4755731539476007e-05,
+      "loss": 1.1064,
+      "step": 11312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4754819801012141e-05,
+      "loss": 1.1104,
+      "step": 11313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4753908011473717e-05,
+      "loss": 1.0347,
+      "step": 11314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4752996170870528e-05,
+      "loss": 1.1211,
+      "step": 11315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4752084279212373e-05,
+      "loss": 1.3984,
+      "step": 11316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4751172336509036e-05,
+      "loss": 1.0908,
+      "step": 11317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4750260342770328e-05,
+      "loss": 1.25,
+      "step": 11318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4749348298006033e-05,
+      "loss": 1.2754,
+      "step": 11319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4748436202225951e-05,
+      "loss": 0.8936,
+      "step": 11320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4747524055439883e-05,
+      "loss": 1.3398,
+      "step": 11321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4746611857657624e-05,
+      "loss": 1.0635,
+      "step": 11322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4745699608888972e-05,
+      "loss": 1.6006,
+      "step": 11323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4744787309143727e-05,
+      "loss": 0.8965,
+      "step": 11324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474387495843169e-05,
+      "loss": 0.8467,
+      "step": 11325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474296255676266e-05,
+      "loss": 1.3125,
+      "step": 11326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4742050104146435e-05,
+      "loss": 1.0127,
+      "step": 11327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474113760059282e-05,
+      "loss": 1.0518,
+      "step": 11328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474022504611162e-05,
+      "loss": 1.3604,
+      "step": 11329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4739312440712627e-05,
+      "loss": 0.9053,
+      "step": 11330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4738399784405652e-05,
+      "loss": 1.002,
+      "step": 11331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4737487077200494e-05,
+      "loss": 0.9463,
+      "step": 11332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4736574319106962e-05,
+      "loss": 1.1387,
+      "step": 11333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4735661510134858e-05,
+      "loss": 1.0674,
+      "step": 11334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4734748650293985e-05,
+      "loss": 1.3271,
+      "step": 11335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4733835739594152e-05,
+      "loss": 1.0557,
+      "step": 11336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4732922778045162e-05,
+      "loss": 1.3242,
+      "step": 11337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4732009765656824e-05,
+      "loss": 0.959,
+      "step": 11338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4731096702438942e-05,
+      "loss": 0.9155,
+      "step": 11339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4730183588401327e-05,
+      "loss": 1.2178,
+      "step": 11340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4729270423553788e-05,
+      "loss": 0.9814,
+      "step": 11341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4728357207906133e-05,
+      "loss": 0.9551,
+      "step": 11342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472744394146817e-05,
+      "loss": 1.1074,
+      "step": 11343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4726530624249708e-05,
+      "loss": 0.9888,
+      "step": 11344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4725617256260563e-05,
+      "loss": 1.2266,
+      "step": 11345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4724703837510542e-05,
+      "loss": 1.0586,
+      "step": 11346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4723790368009456e-05,
+      "loss": 1.084,
+      "step": 11347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722876847767115e-05,
+      "loss": 1.5312,
+      "step": 11348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4721963276793341e-05,
+      "loss": 1.0098,
+      "step": 11349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4721049655097937e-05,
+      "loss": 1.0918,
+      "step": 11350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4720135982690724e-05,
+      "loss": 1.1699,
+      "step": 11351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4719222259581512e-05,
+      "loss": 1.4912,
+      "step": 11352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4718308485780117e-05,
+      "loss": 0.8906,
+      "step": 11353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4717394661296353e-05,
+      "loss": 0.7773,
+      "step": 11354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4716480786140039e-05,
+      "loss": 1.2803,
+      "step": 11355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471556686032099e-05,
+      "loss": 1.1406,
+      "step": 11356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4714652883849023e-05,
+      "loss": 0.6035,
+      "step": 11357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4713738856733956e-05,
+      "loss": 0.9414,
+      "step": 11358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4712824778985608e-05,
+      "loss": 1.2666,
+      "step": 11359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4711910650613799e-05,
+      "loss": 0.9844,
+      "step": 11360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471099647162834e-05,
+      "loss": 0.9761,
+      "step": 11361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471008224203906e-05,
+      "loss": 1.0391,
+      "step": 11362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4709167961855774e-05,
+      "loss": 1.3867,
+      "step": 11363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470825363108831e-05,
+      "loss": 1.1074,
+      "step": 11364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470733924974648e-05,
+      "loss": 0.7803,
+      "step": 11365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470642481784011e-05,
+      "loss": 1.2031,
+      "step": 11366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4705510335379023e-05,
+      "loss": 1.5332,
+      "step": 11367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4704595802373043e-05,
+      "loss": 0.9932,
+      "step": 11368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4703681218831991e-05,
+      "loss": 1.5547,
+      "step": 11369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4702766584765693e-05,
+      "loss": 1.2012,
+      "step": 11370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4701851900183975e-05,
+      "loss": 0.9546,
+      "step": 11371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4700937165096658e-05,
+      "loss": 1.0664,
+      "step": 11372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4700022379513573e-05,
+      "loss": 0.9658,
+      "step": 11373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4699107543444537e-05,
+      "loss": 1.0371,
+      "step": 11374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4698192656899391e-05,
+      "loss": 0.9512,
+      "step": 11375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4697277719887948e-05,
+      "loss": 1.0869,
+      "step": 11376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4696362732420047e-05,
+      "loss": 0.9858,
+      "step": 11377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4695447694505512e-05,
+      "loss": 0.9319,
+      "step": 11378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4694532606154165e-05,
+      "loss": 0.8657,
+      "step": 11379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469361746737585e-05,
+      "loss": 0.8882,
+      "step": 11380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469270227818039e-05,
+      "loss": 1.2676,
+      "step": 11381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4691787038577614e-05,
+      "loss": 0.8896,
+      "step": 11382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4690871748577352e-05,
+      "loss": 0.7979,
+      "step": 11383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468995640818944e-05,
+      "loss": 1.25,
+      "step": 11384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4689041017423708e-05,
+      "loss": 0.9688,
+      "step": 11385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4688125576289992e-05,
+      "loss": 0.8408,
+      "step": 11386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4687210084798118e-05,
+      "loss": 1.1416,
+      "step": 11387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4686294542957927e-05,
+      "loss": 1.0654,
+      "step": 11388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468537895077925e-05,
+      "loss": 1.5352,
+      "step": 11389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468446330827192e-05,
+      "loss": 1.0146,
+      "step": 11390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4683547615445784e-05,
+      "loss": 1.1152,
+      "step": 11391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468263187231066e-05,
+      "loss": 0.8154,
+      "step": 11392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.46817160788764e-05,
+      "loss": 1.2295,
+      "step": 11393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4680800235152833e-05,
+      "loss": 0.8975,
+      "step": 11394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.46798843411498e-05,
+      "loss": 1.0605,
+      "step": 11395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4678968396877135e-05,
+      "loss": 0.8774,
+      "step": 11396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4678052402344683e-05,
+      "loss": 0.7305,
+      "step": 11397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4677136357562276e-05,
+      "loss": 0.8164,
+      "step": 11398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467622026253976e-05,
+      "loss": 1.0371,
+      "step": 11399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4675304117286974e-05,
+      "loss": 1.2383,
+      "step": 11400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4674387921813758e-05,
+      "loss": 0.7412,
+      "step": 11401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467347167612995e-05,
+      "loss": 1.166,
+      "step": 11402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4672555380245398e-05,
+      "loss": 1.0557,
+      "step": 11403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4671639034169946e-05,
+      "loss": 0.9648,
+      "step": 11404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4670722637913428e-05,
+      "loss": 1.1963,
+      "step": 11405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4669806191485695e-05,
+      "loss": 1.1479,
+      "step": 11406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4668889694896586e-05,
+      "loss": 1.3159,
+      "step": 11407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4667973148155951e-05,
+      "loss": 1.1221,
+      "step": 11408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4667056551273633e-05,
+      "loss": 0.9917,
+      "step": 11409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4666139904259478e-05,
+      "loss": 1.4902,
+      "step": 11410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466522320712333e-05,
+      "loss": 1.2871,
+      "step": 11411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4664306459875035e-05,
+      "loss": 0.9761,
+      "step": 11412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4663389662524447e-05,
+      "loss": 1.1562,
+      "step": 11413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4662472815081408e-05,
+      "loss": 1.2021,
+      "step": 11414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4661555917555768e-05,
+      "loss": 1.1904,
+      "step": 11415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4660638969957378e-05,
+      "loss": 0.8188,
+      "step": 11416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4659721972296086e-05,
+      "loss": 1.168,
+      "step": 11417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4658804924581738e-05,
+      "loss": 1.2686,
+      "step": 11418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4657887826824193e-05,
+      "loss": 1.1191,
+      "step": 11419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4656970679033293e-05,
+      "loss": 0.9614,
+      "step": 11420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.46560534812189e-05,
+      "loss": 1.1426,
+      "step": 11421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4655136233390858e-05,
+      "loss": 0.8857,
+      "step": 11422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4654218935559022e-05,
+      "loss": 1.1465,
+      "step": 11423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4653301587733243e-05,
+      "loss": 0.9204,
+      "step": 11424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465238418992338e-05,
+      "loss": 1.1045,
+      "step": 11425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4651466742139284e-05,
+      "loss": 0.9697,
+      "step": 11426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465054924439081e-05,
+      "loss": 1.1201,
+      "step": 11427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4649631696687814e-05,
+      "loss": 1.4414,
+      "step": 11428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4648714099040153e-05,
+      "loss": 1.4189,
+      "step": 11429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4647796451457683e-05,
+      "loss": 1.0498,
+      "step": 11430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4646878753950257e-05,
+      "loss": 0.9658,
+      "step": 11431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464596100652774e-05,
+      "loss": 1.1377,
+      "step": 11432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4645043209199984e-05,
+      "loss": 1.0244,
+      "step": 11433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464412536197685e-05,
+      "loss": 0.9668,
+      "step": 11434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4643207464868198e-05,
+      "loss": 0.9473,
+      "step": 11435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4642289517883886e-05,
+      "loss": 0.8257,
+      "step": 11436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4641371521033775e-05,
+      "loss": 0.9854,
+      "step": 11437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4640453474327725e-05,
+      "loss": 1.1357,
+      "step": 11438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.46395353777756e-05,
+      "loss": 1.5234,
+      "step": 11439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463861723138726e-05,
+      "loss": 1.0361,
+      "step": 11440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4637699035172565e-05,
+      "loss": 1.376,
+      "step": 11441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4636780789141383e-05,
+      "loss": 1.2197,
+      "step": 11442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4635862493303572e-05,
+      "loss": 1.127,
+      "step": 11443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4634944147669e-05,
+      "loss": 1.1006,
+      "step": 11444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4634025752247532e-05,
+      "loss": 1.125,
+      "step": 11445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4633107307049028e-05,
+      "loss": 1.0312,
+      "step": 11446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4632188812083359e-05,
+      "loss": 1.0068,
+      "step": 11447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463127026736039e-05,
+      "loss": 0.957,
+      "step": 11448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630351672889984e-05,
+      "loss": 0.835,
+      "step": 11449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4629433028682014e-05,
+      "loss": 0.8848,
+      "step": 11450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4628514334746342e-05,
+      "loss": 1.2178,
+      "step": 11451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4627595591092842e-05,
+      "loss": 1.0625,
+      "step": 11452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4626676797731379e-05,
+      "loss": 0.8301,
+      "step": 11453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4625757954671823e-05,
+      "loss": 0.8604,
+      "step": 11454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462483906192404e-05,
+      "loss": 0.8975,
+      "step": 11455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462392011949791e-05,
+      "loss": 1.1562,
+      "step": 11456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4623001127403296e-05,
+      "loss": 1.0913,
+      "step": 11457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4622082085650072e-05,
+      "loss": 1.2764,
+      "step": 11458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462116299424811e-05,
+      "loss": 1.4551,
+      "step": 11459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4620243853207282e-05,
+      "loss": 1.3506,
+      "step": 11460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4619324662537463e-05,
+      "loss": 1.377,
+      "step": 11461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4618405422248522e-05,
+      "loss": 1.2051,
+      "step": 11462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4617486132350343e-05,
+      "loss": 1.2285,
+      "step": 11463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4616566792852787e-05,
+      "loss": 1.2637,
+      "step": 11464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461564740376574e-05,
+      "loss": 0.8398,
+      "step": 11465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4614727965099074e-05,
+      "loss": 1.2061,
+      "step": 11466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4613808476862664e-05,
+      "loss": 1.1689,
+      "step": 11467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4612888939066388e-05,
+      "loss": 1.5664,
+      "step": 11468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611969351720121e-05,
+      "loss": 1.251,
+      "step": 11469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611049714833747e-05,
+      "loss": 1.3203,
+      "step": 11470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461013002841714e-05,
+      "loss": 1.1216,
+      "step": 11471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4609210292480178e-05,
+      "loss": 1.2012,
+      "step": 11472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4608290507032743e-05,
+      "loss": 0.8945,
+      "step": 11473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4607370672084717e-05,
+      "loss": 1.2095,
+      "step": 11474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4606450787645971e-05,
+      "loss": 1.1973,
+      "step": 11475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.46055308537264e-05,
+      "loss": 1.3525,
+      "step": 11476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4604610870335872e-05,
+      "loss": 0.7217,
+      "step": 11477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4603690837484279e-05,
+      "loss": 1.2734,
+      "step": 11478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.46027707551815e-05,
+      "loss": 1.1025,
+      "step": 11479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4601850623437415e-05,
+      "loss": 1.4375,
+      "step": 11480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4600930442261916e-05,
+      "loss": 1.0049,
+      "step": 11481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4600010211664877e-05,
+      "loss": 0.9473,
+      "step": 11482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4599089931656191e-05,
+      "loss": 1.0801,
+      "step": 11483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4598169602245742e-05,
+      "loss": 0.7227,
+      "step": 11484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4597249223443413e-05,
+      "loss": 0.9502,
+      "step": 11485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459632879525909e-05,
+      "loss": 0.9604,
+      "step": 11486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4595408317702663e-05,
+      "loss": 1.0674,
+      "step": 11487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4594487790784015e-05,
+      "loss": 1.0859,
+      "step": 11488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459356721451304e-05,
+      "loss": 1.1943,
+      "step": 11489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4592646588899623e-05,
+      "loss": 0.8843,
+      "step": 11490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4591725913953654e-05,
+      "loss": 1.0508,
+      "step": 11491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4590805189685022e-05,
+      "loss": 0.9453,
+      "step": 11492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4589884416103614e-05,
+      "loss": 0.917,
+      "step": 11493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4588963593219327e-05,
+      "loss": 1.085,
+      "step": 11494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4588042721042046e-05,
+      "loss": 1.3193,
+      "step": 11495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4587121799581669e-05,
+      "loss": 0.9746,
+      "step": 11496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4586200828848083e-05,
+      "loss": 0.9697,
+      "step": 11497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.458527980885118e-05,
+      "loss": 1.2705,
+      "step": 11498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4584358739600859e-05,
+      "loss": 1.0151,
+      "step": 11499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4583437621107008e-05,
+      "loss": 1.5469,
+      "step": 11500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4582516453379525e-05,
+      "loss": 0.96,
+      "step": 11501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4581595236428303e-05,
+      "loss": 0.9819,
+      "step": 11502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.458067397026324e-05,
+      "loss": 0.9209,
+      "step": 11503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4579752654894229e-05,
+      "loss": 0.9019,
+      "step": 11504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4578831290331168e-05,
+      "loss": 1.2085,
+      "step": 11505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4577909876583951e-05,
+      "loss": 1.1943,
+      "step": 11506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.457698841366248e-05,
+      "loss": 1.1914,
+      "step": 11507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.457606690157665e-05,
+      "loss": 1.8457,
+      "step": 11508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4575145340336363e-05,
+      "loss": 1.209,
+      "step": 11509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4574223729951511e-05,
+      "loss": 1.0537,
+      "step": 11510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4573302070432001e-05,
+      "loss": 1.0537,
+      "step": 11511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.457238036178773e-05,
+      "loss": 1.1484,
+      "step": 11512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45714586040286e-05,
+      "loss": 1.4629,
+      "step": 11513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4570536797164507e-05,
+      "loss": 0.8311,
+      "step": 11514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4569614941205359e-05,
+      "loss": 0.9209,
+      "step": 11515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4568693036161057e-05,
+      "loss": 1.0713,
+      "step": 11516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4567771082041503e-05,
+      "loss": 0.8716,
+      "step": 11517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4566849078856597e-05,
+      "loss": 1.418,
+      "step": 11518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456592702661625e-05,
+      "loss": 1.124,
+      "step": 11519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4565004925330362e-05,
+      "loss": 1.1533,
+      "step": 11520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4564082775008834e-05,
+      "loss": 1.0537,
+      "step": 11521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456316057566158e-05,
+      "loss": 1.1475,
+      "step": 11522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4562238327298501e-05,
+      "loss": 0.8867,
+      "step": 11523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45613160299295e-05,
+      "loss": 0.8672,
+      "step": 11524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4560393683564494e-05,
+      "loss": 1.5586,
+      "step": 11525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4559471288213378e-05,
+      "loss": 0.9189,
+      "step": 11526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4558548843886073e-05,
+      "loss": 1.4365,
+      "step": 11527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4557626350592477e-05,
+      "loss": 1.0303,
+      "step": 11528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4556703808342506e-05,
+      "loss": 1.0439,
+      "step": 11529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4555781217146063e-05,
+      "loss": 1.0977,
+      "step": 11530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4554858577013064e-05,
+      "loss": 0.9624,
+      "step": 11531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4553935887953416e-05,
+      "loss": 0.957,
+      "step": 11532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4553013149977033e-05,
+      "loss": 1.1426,
+      "step": 11533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4552090363093827e-05,
+      "loss": 1.1597,
+      "step": 11534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4551167527313707e-05,
+      "loss": 1.1387,
+      "step": 11535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455024464264659e-05,
+      "loss": 1.3613,
+      "step": 11536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4549321709102382e-05,
+      "loss": 0.6875,
+      "step": 11537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4548398726691006e-05,
+      "loss": 1.1323,
+      "step": 11538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4547475695422368e-05,
+      "loss": 0.9429,
+      "step": 11539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4546552615306393e-05,
+      "loss": 0.9678,
+      "step": 11540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4545629486352985e-05,
+      "loss": 1.21,
+      "step": 11541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4544706308572068e-05,
+      "loss": 0.7817,
+      "step": 11542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4543783081973553e-05,
+      "loss": 1.1992,
+      "step": 11543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454285980656736e-05,
+      "loss": 0.8174,
+      "step": 11544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4541936482363407e-05,
+      "loss": 0.9609,
+      "step": 11545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454101310937161e-05,
+      "loss": 1.2754,
+      "step": 11546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454008968760189e-05,
+      "loss": 1.0361,
+      "step": 11547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4539166217064166e-05,
+      "loss": 1.0586,
+      "step": 11548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4538242697768356e-05,
+      "loss": 1.124,
+      "step": 11549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4537319129724376e-05,
+      "loss": 1.292,
+      "step": 11550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4536395512942157e-05,
+      "loss": 0.8628,
+      "step": 11551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4535471847431607e-05,
+      "loss": 0.9043,
+      "step": 11552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4534548133202662e-05,
+      "loss": 0.8408,
+      "step": 11553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4533624370265233e-05,
+      "loss": 1.4775,
+      "step": 11554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453270055862925e-05,
+      "loss": 1.1992,
+      "step": 11555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4531776698304629e-05,
+      "loss": 1.1201,
+      "step": 11556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45308527893013e-05,
+      "loss": 1.1201,
+      "step": 11557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4529928831629185e-05,
+      "loss": 0.8936,
+      "step": 11558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4529004825298207e-05,
+      "loss": 0.8359,
+      "step": 11559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4528080770318296e-05,
+      "loss": 0.9482,
+      "step": 11560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4527156666699374e-05,
+      "loss": 1.25,
+      "step": 11561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4526232514451369e-05,
+      "loss": 0.9814,
+      "step": 11562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4525308313584208e-05,
+      "loss": 1.1465,
+      "step": 11563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4524384064107816e-05,
+      "loss": 0.8467,
+      "step": 11564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4523459766032124e-05,
+      "loss": 1.1689,
+      "step": 11565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452253541936706e-05,
+      "loss": 1.0518,
+      "step": 11566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452161102412255e-05,
+      "loss": 1.501,
+      "step": 11567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4520686580308528e-05,
+      "loss": 0.9307,
+      "step": 11568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451976208793492e-05,
+      "loss": 1.3867,
+      "step": 11569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4518837547011657e-05,
+      "loss": 1.1338,
+      "step": 11570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4517912957548676e-05,
+      "loss": 1.1387,
+      "step": 11571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4516988319555902e-05,
+      "loss": 1.2266,
+      "step": 11572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4516063633043268e-05,
+      "loss": 1.3438,
+      "step": 11573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451513889802071e-05,
+      "loss": 0.8271,
+      "step": 11574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451421411449816e-05,
+      "loss": 0.9521,
+      "step": 11575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4513289282485548e-05,
+      "loss": 0.8398,
+      "step": 11576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451236440199281e-05,
+      "loss": 0.9033,
+      "step": 11577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4511439473029884e-05,
+      "loss": 1.4668,
+      "step": 11578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4510514495606702e-05,
+      "loss": 1.1416,
+      "step": 11579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4509589469733206e-05,
+      "loss": 0.8657,
+      "step": 11580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4508664395419319e-05,
+      "loss": 1.0703,
+      "step": 11581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450773927267499e-05,
+      "loss": 1.1582,
+      "step": 11582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450681410151015e-05,
+      "loss": 1.2012,
+      "step": 11583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4505888881934742e-05,
+      "loss": 1.4502,
+      "step": 11584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45049636139587e-05,
+      "loss": 1.3643,
+      "step": 11585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4504038297591962e-05,
+      "loss": 1.1191,
+      "step": 11586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4503112932844472e-05,
+      "loss": 0.9556,
+      "step": 11587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4502187519726168e-05,
+      "loss": 0.9805,
+      "step": 11588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450126205824699e-05,
+      "loss": 0.7979,
+      "step": 11589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4500336548416878e-05,
+      "loss": 0.5835,
+      "step": 11590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4499410990245774e-05,
+      "loss": 0.8721,
+      "step": 11591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449848538374362e-05,
+      "loss": 0.9453,
+      "step": 11592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4497559728920363e-05,
+      "loss": 0.9248,
+      "step": 11593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4496634025785938e-05,
+      "loss": 0.8843,
+      "step": 11594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4495708274350298e-05,
+      "loss": 1.2109,
+      "step": 11595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4494782474623377e-05,
+      "loss": 1.0234,
+      "step": 11596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4493856626615127e-05,
+      "loss": 0.9365,
+      "step": 11597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4492930730335489e-05,
+      "loss": 0.7102,
+      "step": 11598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4492004785794409e-05,
+      "loss": 0.9194,
+      "step": 11599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4491078793001836e-05,
+      "loss": 1.4424,
+      "step": 11600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4490152751967715e-05,
+      "loss": 1.543,
+      "step": 11601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4489226662701994e-05,
+      "loss": 1.23,
+      "step": 11602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4488300525214619e-05,
+      "loss": 0.9429,
+      "step": 11603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4487374339515539e-05,
+      "loss": 1.0161,
+      "step": 11604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4486448105614704e-05,
+      "loss": 0.4595,
+      "step": 11605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448552182352206e-05,
+      "loss": 1.1758,
+      "step": 11606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448459549324756e-05,
+      "loss": 1.0781,
+      "step": 11607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4483669114801155e-05,
+      "loss": 1.1045,
+      "step": 11608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448274268819279e-05,
+      "loss": 1.1357,
+      "step": 11609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4481816213432428e-05,
+      "loss": 1.0469,
+      "step": 11610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4480889690530007e-05,
+      "loss": 1.0361,
+      "step": 11611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4479963119495485e-05,
+      "loss": 1.1875,
+      "step": 11612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447903650033882e-05,
+      "loss": 1.2803,
+      "step": 11613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4478109833069956e-05,
+      "loss": 0.9482,
+      "step": 11614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4477183117698856e-05,
+      "loss": 0.9922,
+      "step": 11615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4476256354235469e-05,
+      "loss": 1.0107,
+      "step": 11616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4475329542689752e-05,
+      "loss": 1.1382,
+      "step": 11617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4474402683071659e-05,
+      "loss": 0.7773,
+      "step": 11618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4473475775391146e-05,
+      "loss": 1.1201,
+      "step": 11619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447254881965817e-05,
+      "loss": 1.123,
+      "step": 11620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4471621815882692e-05,
+      "loss": 1.0283,
+      "step": 11621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4470694764074663e-05,
+      "loss": 1.2969,
+      "step": 11622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4469767664244043e-05,
+      "loss": 1.0713,
+      "step": 11623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4468840516400793e-05,
+      "loss": 1.0361,
+      "step": 11624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.446791332055487e-05,
+      "loss": 1.1387,
+      "step": 11625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4466986076716235e-05,
+      "loss": 0.7334,
+      "step": 11626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4466058784894846e-05,
+      "loss": 0.9365,
+      "step": 11627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4465131445100667e-05,
+      "loss": 0.9155,
+      "step": 11628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4464204057343657e-05,
+      "loss": 1.1499,
+      "step": 11629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4463276621633775e-05,
+      "loss": 0.9883,
+      "step": 11630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4462349137980987e-05,
+      "loss": 1.333,
+      "step": 11631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4461421606395253e-05,
+      "loss": 1.21,
+      "step": 11632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4460494026886542e-05,
+      "loss": 1.3086,
+      "step": 11633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445956639946481e-05,
+      "loss": 0.9312,
+      "step": 11634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4458638724140027e-05,
+      "loss": 0.9062,
+      "step": 11635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4457711000922155e-05,
+      "loss": 1.4688,
+      "step": 11636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4456783229821159e-05,
+      "loss": 1.1328,
+      "step": 11637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4455855410847006e-05,
+      "loss": 1.2656,
+      "step": 11638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4454927544009665e-05,
+      "loss": 1.0859,
+      "step": 11639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4453999629319093e-05,
+      "loss": 1.0547,
+      "step": 11640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445307166678527e-05,
+      "loss": 0.9072,
+      "step": 11641
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4452143656418157e-05,
+      "loss": 1.0146,
+      "step": 11642
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4451215598227721e-05,
+      "loss": 1.1729,
+      "step": 11643
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4450287492223933e-05,
+      "loss": 1.167,
+      "step": 11644
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4449359338416761e-05,
+      "loss": 1.1748,
+      "step": 11645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444843113681618e-05,
+      "loss": 0.9492,
+      "step": 11646
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4447502887432157e-05,
+      "loss": 0.9976,
+      "step": 11647
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444657459027466e-05,
+      "loss": 1.0576,
+      "step": 11648
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4445646245353664e-05,
+      "loss": 1.502,
+      "step": 11649
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4444717852679139e-05,
+      "loss": 1.4102,
+      "step": 11650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444378941226106e-05,
+      "loss": 1.1738,
+      "step": 11651
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44428609241094e-05,
+      "loss": 1.1025,
+      "step": 11652
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4441932388234128e-05,
+      "loss": 1.0098,
+      "step": 11653
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4441003804645224e-05,
+      "loss": 1.4092,
+      "step": 11654
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4440075173352658e-05,
+      "loss": 1.5811,
+      "step": 11655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4439146494366405e-05,
+      "loss": 0.9658,
+      "step": 11656
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4438217767696445e-05,
+      "loss": 1.2754,
+      "step": 11657
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443728899335275e-05,
+      "loss": 0.9082,
+      "step": 11658
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4436360171345296e-05,
+      "loss": 0.96,
+      "step": 11659
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4435431301684061e-05,
+      "loss": 1.1279,
+      "step": 11660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4434502384379026e-05,
+      "loss": 1.5039,
+      "step": 11661
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4433573419440166e-05,
+      "loss": 1.2139,
+      "step": 11662
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4432644406877458e-05,
+      "loss": 0.9014,
+      "step": 11663
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4431715346700886e-05,
+      "loss": 1.124,
+      "step": 11664
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4430786238920424e-05,
+      "loss": 0.8877,
+      "step": 11665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4429857083546054e-05,
+      "loss": 0.9619,
+      "step": 11666
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4428927880587765e-05,
+      "loss": 1.3037,
+      "step": 11667
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4427998630055521e-05,
+      "loss": 0.9907,
+      "step": 11668
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4427069331959318e-05,
+      "loss": 0.8916,
+      "step": 11669
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4426139986309136e-05,
+      "loss": 1.3906,
+      "step": 11670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4425210593114951e-05,
+      "loss": 1.084,
+      "step": 11671
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4424281152386751e-05,
+      "loss": 0.9268,
+      "step": 11672
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4423351664134521e-05,
+      "loss": 0.7998,
+      "step": 11673
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4422422128368242e-05,
+      "loss": 1.0557,
+      "step": 11674
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4421492545097901e-05,
+      "loss": 0.873,
+      "step": 11675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4420562914333482e-05,
+      "loss": 0.915,
+      "step": 11676
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4419633236084974e-05,
+      "loss": 1.502,
+      "step": 11677
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4418703510362355e-05,
+      "loss": 1.168,
+      "step": 11678
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441777373717562e-05,
+      "loss": 1.1309,
+      "step": 11679
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4416843916534755e-05,
+      "loss": 0.9854,
+      "step": 11680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4415914048449744e-05,
+      "loss": 1.123,
+      "step": 11681
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4414984132930579e-05,
+      "loss": 0.9062,
+      "step": 11682
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4414054169987248e-05,
+      "loss": 0.7739,
+      "step": 11683
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4413124159629734e-05,
+      "loss": 0.7954,
+      "step": 11684
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441219410186804e-05,
+      "loss": 1.1094,
+      "step": 11685
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4411263996712144e-05,
+      "loss": 1.0811,
+      "step": 11686
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4410333844172046e-05,
+      "loss": 1.3594,
+      "step": 11687
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4409403644257732e-05,
+      "loss": 0.7588,
+      "step": 11688
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4408473396979193e-05,
+      "loss": 1.2129,
+      "step": 11689
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4407543102346423e-05,
+      "loss": 1.1973,
+      "step": 11690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4406612760369416e-05,
+      "loss": 0.9146,
+      "step": 11691
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4405682371058165e-05,
+      "loss": 1.0332,
+      "step": 11692
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4404751934422665e-05,
+      "loss": 1.1455,
+      "step": 11693
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4403821450472908e-05,
+      "loss": 1.1846,
+      "step": 11694
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4402890919218889e-05,
+      "loss": 1.1445,
+      "step": 11695
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4401960340670606e-05,
+      "loss": 1.3584,
+      "step": 11696
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4401029714838051e-05,
+      "loss": 1.2773,
+      "step": 11697
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4400099041731226e-05,
+      "loss": 0.9629,
+      "step": 11698
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439916832136012e-05,
+      "loss": 1.0537,
+      "step": 11699
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439823755373474e-05,
+      "loss": 1.1201,
+      "step": 11700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4397306738865076e-05,
+      "loss": 0.6851,
+      "step": 11701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439637587676113e-05,
+      "loss": 1.0039,
+      "step": 11702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.43954449674329e-05,
+      "loss": 0.9385,
+      "step": 11703
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439451401089039e-05,
+      "loss": 1.1807,
+      "step": 11704
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4393583007143592e-05,
+      "loss": 1.2861,
+      "step": 11705
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4392651956202512e-05,
+      "loss": 1.0156,
+      "step": 11706
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439172085807715e-05,
+      "loss": 0.8262,
+      "step": 11707
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4390789712777506e-05,
+      "loss": 0.9502,
+      "step": 11708
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4389858520313584e-05,
+      "loss": 1.1426,
+      "step": 11709
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4388927280695385e-05,
+      "loss": 1.3301,
+      "step": 11710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4387995993932916e-05,
+      "loss": 1.1592,
+      "step": 11711
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438706466003617e-05,
+      "loss": 1.1016,
+      "step": 11712
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4386133279015165e-05,
+      "loss": 1.3398,
+      "step": 11713
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4385201850879895e-05,
+      "loss": 1.4277,
+      "step": 11714
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438427037564037e-05,
+      "loss": 1.1016,
+      "step": 11715
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4383338853306597e-05,
+      "loss": 0.8777,
+      "step": 11716
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4382407283888578e-05,
+      "loss": 1.0498,
+      "step": 11717
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438147566739632e-05,
+      "loss": 1.2793,
+      "step": 11718
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4380544003839832e-05,
+      "loss": 0.9531,
+      "step": 11719
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.437961229322912e-05,
+      "loss": 1.2793,
+      "step": 11720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4378680535574195e-05,
+      "loss": 1.1387,
+      "step": 11721
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4377748730885061e-05,
+      "loss": 1.1309,
+      "step": 11722
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4376816879171732e-05,
+      "loss": 1.043,
+      "step": 11723
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4375884980444213e-05,
+      "loss": 0.8418,
+      "step": 11724
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4374953034712518e-05,
+      "loss": 1.0264,
+      "step": 11725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4374021041986655e-05,
+      "loss": 1.2041,
+      "step": 11726
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.437308900227664e-05,
+      "loss": 1.1162,
+      "step": 11727
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4372156915592474e-05,
+      "loss": 1.5273,
+      "step": 11728
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371224781944182e-05,
+      "loss": 1.2373,
+      "step": 11729
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4370292601341764e-05,
+      "loss": 1.1104,
+      "step": 11730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4369360373795247e-05,
+      "loss": 0.9395,
+      "step": 11731
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4368428099314633e-05,
+      "loss": 1.0889,
+      "step": 11732
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4367495777909943e-05,
+      "loss": 1.1797,
+      "step": 11733
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4366563409591187e-05,
+      "loss": 1.0254,
+      "step": 11734
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4365630994368385e-05,
+      "loss": 0.9258,
+      "step": 11735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4364698532251546e-05,
+      "loss": 0.9033,
+      "step": 11736
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4363766023250693e-05,
+      "loss": 0.7559,
+      "step": 11737
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4362833467375839e-05,
+      "loss": 0.9873,
+      "step": 11738
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4361900864637001e-05,
+      "loss": 1.7148,
+      "step": 11739
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4360968215044202e-05,
+      "loss": 1.0654,
+      "step": 11740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.436003551860745e-05,
+      "loss": 1.3135,
+      "step": 11741
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4359102775336775e-05,
+      "loss": 1.4072,
+      "step": 11742
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4358169985242188e-05,
+      "loss": 1.2969,
+      "step": 11743
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435723714833371e-05,
+      "loss": 1.4922,
+      "step": 11744
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4356304264621366e-05,
+      "loss": 1.3213,
+      "step": 11745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435537133411517e-05,
+      "loss": 1.0576,
+      "step": 11746
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4354438356825152e-05,
+      "loss": 1.3018,
+      "step": 11747
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4353505332761324e-05,
+      "loss": 1.0527,
+      "step": 11748
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4352572261933713e-05,
+      "loss": 0.9941,
+      "step": 11749
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435163914435234e-05,
+      "loss": 1.0957,
+      "step": 11750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435070598002723e-05,
+      "loss": 1.2539,
+      "step": 11751
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4349772768968408e-05,
+      "loss": 1.1348,
+      "step": 11752
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4348839511185899e-05,
+      "loss": 1.1318,
+      "step": 11753
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4347906206689719e-05,
+      "loss": 1.2041,
+      "step": 11754
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4346972855489902e-05,
+      "loss": 1.0098,
+      "step": 11755
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.434603945759647e-05,
+      "loss": 1.125,
+      "step": 11756
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4345106013019453e-05,
+      "loss": 0.958,
+      "step": 11757
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4344172521768873e-05,
+      "loss": 0.9795,
+      "step": 11758
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4343238983854754e-05,
+      "loss": 1.1348,
+      "step": 11759
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4342305399287137e-05,
+      "loss": 1.1104,
+      "step": 11760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4341371768076038e-05,
+      "loss": 1.0869,
+      "step": 11761
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4340438090231492e-05,
+      "loss": 0.9951,
+      "step": 11762
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4339504365763526e-05,
+      "loss": 1.1084,
+      "step": 11763
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4338570594682169e-05,
+      "loss": 1.3066,
+      "step": 11764
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4337636776997453e-05,
+      "loss": 1.0127,
+      "step": 11765
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4336702912719407e-05,
+      "loss": 0.8149,
+      "step": 11766
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433576900185806e-05,
+      "loss": 0.9922,
+      "step": 11767
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433483504442345e-05,
+      "loss": 0.5474,
+      "step": 11768
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4333901040425606e-05,
+      "loss": 0.9932,
+      "step": 11769
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433296698987456e-05,
+      "loss": 1.1145,
+      "step": 11770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4332032892780348e-05,
+      "loss": 1.0527,
+      "step": 11771
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4331098749152998e-05,
+      "loss": 1.3652,
+      "step": 11772
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4330164559002553e-05,
+      "loss": 1.1074,
+      "step": 11773
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4329230322339036e-05,
+      "loss": 1.3794,
+      "step": 11774
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4328296039172496e-05,
+      "loss": 1.2734,
+      "step": 11775
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4327361709512957e-05,
+      "loss": 1.543,
+      "step": 11776
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4326427333370461e-05,
+      "loss": 1.4805,
+      "step": 11777
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4325492910755043e-05,
+      "loss": 1.2471,
+      "step": 11778
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4324558441676742e-05,
+      "loss": 1.1357,
+      "step": 11779
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4323623926145593e-05,
+      "loss": 1.3105,
+      "step": 11780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4322689364171635e-05,
+      "loss": 0.6738,
+      "step": 11781
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4321754755764912e-05,
+      "loss": 1.0869,
+      "step": 11782
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4320820100935455e-05,
+      "loss": 0.9795,
+      "step": 11783
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4319885399693312e-05,
+      "loss": 1.5605,
+      "step": 11784
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4318950652048512e-05,
+      "loss": 1.1387,
+      "step": 11785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4318015858011108e-05,
+      "loss": 1.418,
+      "step": 11786
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431708101759113e-05,
+      "loss": 1.0693,
+      "step": 11787
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4316146130798632e-05,
+      "loss": 1.2046,
+      "step": 11788
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4315211197643644e-05,
+      "loss": 0.9902,
+      "step": 11789
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4314276218136218e-05,
+      "loss": 1.0566,
+      "step": 11790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4313341192286392e-05,
+      "loss": 0.9619,
+      "step": 11791
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431240612010421e-05,
+      "loss": 1.1113,
+      "step": 11792
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4311471001599717e-05,
+      "loss": 1.2051,
+      "step": 11793
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431053583678296e-05,
+      "loss": 1.2227,
+      "step": 11794
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4309600625663983e-05,
+      "loss": 0.9893,
+      "step": 11795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430866536825283e-05,
+      "loss": 1.1621,
+      "step": 11796
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430773006455955e-05,
+      "loss": 1.0156,
+      "step": 11797
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4306794714594183e-05,
+      "loss": 1.333,
+      "step": 11798
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4305859318366785e-05,
+      "loss": 1.0186,
+      "step": 11799
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4304923875887398e-05,
+      "loss": 1.0801,
+      "step": 11800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4303988387166074e-05,
+      "loss": 1.1279,
+      "step": 11801
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430305285221286e-05,
+      "loss": 1.0596,
+      "step": 11802
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4302117271037803e-05,
+      "loss": 1.3691,
+      "step": 11803
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4301181643650954e-05,
+      "loss": 1.1553,
+      "step": 11804
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4300245970062362e-05,
+      "loss": 1.3457,
+      "step": 11805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4299310250282086e-05,
+      "loss": 0.9614,
+      "step": 11806
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4298374484320165e-05,
+      "loss": 1.1133,
+      "step": 11807
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4297438672186658e-05,
+      "loss": 1.082,
+      "step": 11808
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4296502813891616e-05,
+      "loss": 0.9294,
+      "step": 11809
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429556690944509e-05,
+      "loss": 1.166,
+      "step": 11810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4294630958857135e-05,
+      "loss": 0.8174,
+      "step": 11811
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4293694962137801e-05,
+      "loss": 1.1265,
+      "step": 11812
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4292758919297148e-05,
+      "loss": 1.4395,
+      "step": 11813
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4291822830345225e-05,
+      "loss": 1.0693,
+      "step": 11814
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4290886695292093e-05,
+      "loss": 1.0078,
+      "step": 11815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4289950514147802e-05,
+      "loss": 1.2793,
+      "step": 11816
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428901428692241e-05,
+      "loss": 1.1543,
+      "step": 11817
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4288078013625975e-05,
+      "loss": 0.8794,
+      "step": 11818
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4287141694268554e-05,
+      "loss": 1.1133,
+      "step": 11819
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4286205328860204e-05,
+      "loss": 1.4395,
+      "step": 11820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428526891741098e-05,
+      "loss": 1.1377,
+      "step": 11821
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4284332459930946e-05,
+      "loss": 1.0776,
+      "step": 11822
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4283395956430157e-05,
+      "loss": 1.123,
+      "step": 11823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4282459406918674e-05,
+      "loss": 0.9106,
+      "step": 11824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4281522811406559e-05,
+      "loss": 1.5156,
+      "step": 11825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4280586169903869e-05,
+      "loss": 1.043,
+      "step": 11826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4279649482420665e-05,
+      "loss": 1.5293,
+      "step": 11827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4278712748967017e-05,
+      "loss": 1.3809,
+      "step": 11828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4277775969552974e-05,
+      "loss": 1.0273,
+      "step": 11829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4276839144188608e-05,
+      "loss": 1.0273,
+      "step": 11830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4275902272883977e-05,
+      "loss": 1.1318,
+      "step": 11831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427496535564915e-05,
+      "loss": 0.8867,
+      "step": 11832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4274028392494186e-05,
+      "loss": 1.4502,
+      "step": 11833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427309138342915e-05,
+      "loss": 0.9395,
+      "step": 11834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427215432846411e-05,
+      "loss": 1.3574,
+      "step": 11835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4271217227609127e-05,
+      "loss": 0.9648,
+      "step": 11836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4270280080874272e-05,
+      "loss": 1.2754,
+      "step": 11837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4269342888269609e-05,
+      "loss": 1.0088,
+      "step": 11838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4268405649805202e-05,
+      "loss": 0.9121,
+      "step": 11839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4267468365491125e-05,
+      "loss": 1.249,
+      "step": 11840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4266531035337443e-05,
+      "loss": 1.0059,
+      "step": 11841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426559365935422e-05,
+      "loss": 0.8198,
+      "step": 11842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4264656237551532e-05,
+      "loss": 0.6895,
+      "step": 11843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4263718769939444e-05,
+      "loss": 1.3906,
+      "step": 11844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426278125652803e-05,
+      "loss": 1.1631,
+      "step": 11845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4261843697327355e-05,
+      "loss": 1.0796,
+      "step": 11846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4260906092347495e-05,
+      "loss": 1.001,
+      "step": 11847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4259968441598518e-05,
+      "loss": 0.8457,
+      "step": 11848
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4259030745090495e-05,
+      "loss": 1.1699,
+      "step": 11849
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4258093002833504e-05,
+      "loss": 1.2119,
+      "step": 11850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4257155214837613e-05,
+      "loss": 1.0024,
+      "step": 11851
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4256217381112895e-05,
+      "loss": 1.1846,
+      "step": 11852
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4255279501669429e-05,
+      "loss": 1.2471,
+      "step": 11853
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425434157651728e-05,
+      "loss": 0.9395,
+      "step": 11854
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4253403605666534e-05,
+      "loss": 1.1533,
+      "step": 11855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4252465589127262e-05,
+      "loss": 1.1689,
+      "step": 11856
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4251527526909535e-05,
+      "loss": 0.918,
+      "step": 11857
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4250589419023435e-05,
+      "loss": 1.0088,
+      "step": 11858
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424965126547904e-05,
+      "loss": 1.248,
+      "step": 11859
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4248713066286417e-05,
+      "loss": 1.126,
+      "step": 11860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4247774821455657e-05,
+      "loss": 1.1875,
+      "step": 11861
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424683653099683e-05,
+      "loss": 1.0518,
+      "step": 11862
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424589819492002e-05,
+      "loss": 1.2217,
+      "step": 11863
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4244959813235304e-05,
+      "loss": 1.0439,
+      "step": 11864
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4244021385952758e-05,
+      "loss": 1.4268,
+      "step": 11865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4243082913082467e-05,
+      "loss": 1.2217,
+      "step": 11866
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424214439463451e-05,
+      "loss": 1.3682,
+      "step": 11867
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4241205830618969e-05,
+      "loss": 0.9902,
+      "step": 11868
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4240267221045926e-05,
+      "loss": 1.0078,
+      "step": 11869
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4239328565925461e-05,
+      "loss": 0.8101,
+      "step": 11870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4238389865267656e-05,
+      "loss": 0.981,
+      "step": 11871
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4237451119082602e-05,
+      "loss": 1.124,
+      "step": 11872
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4236512327380371e-05,
+      "loss": 0.8594,
+      "step": 11873
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4235573490171057e-05,
+      "loss": 1.0596,
+      "step": 11874
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4234634607464739e-05,
+      "loss": 0.78,
+      "step": 11875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4233695679271506e-05,
+      "loss": 1.1133,
+      "step": 11876
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.423275670560144e-05,
+      "loss": 1.127,
+      "step": 11877
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4231817686464631e-05,
+      "loss": 0.8472,
+      "step": 11878
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.423087862187116e-05,
+      "loss": 0.8438,
+      "step": 11879
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.422993951183112e-05,
+      "loss": 1.5215,
+      "step": 11880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4229000356354595e-05,
+      "loss": 0.8057,
+      "step": 11881
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4228061155451671e-05,
+      "loss": 0.9111,
+      "step": 11882
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4227121909132442e-05,
+      "loss": 1.4473,
+      "step": 11883
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4226182617406996e-05,
+      "loss": 1.2588,
+      "step": 11884
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4225243280285421e-05,
+      "loss": 1.1836,
+      "step": 11885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4224303897777805e-05,
+      "loss": 0.9385,
+      "step": 11886
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4223364469894243e-05,
+      "loss": 1.0283,
+      "step": 11887
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.422242499664482e-05,
+      "loss": 0.916,
+      "step": 11888
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4221485478039635e-05,
+      "loss": 1.1875,
+      "step": 11889
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4220545914088777e-05,
+      "loss": 1.4043,
+      "step": 11890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4219606304802334e-05,
+      "loss": 1.0703,
+      "step": 11891
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4218666650190405e-05,
+      "loss": 1.1152,
+      "step": 11892
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4217726950263079e-05,
+      "loss": 1.1953,
+      "step": 11893
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4216787205030453e-05,
+      "loss": 1.5244,
+      "step": 11894
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4215847414502619e-05,
+      "loss": 0.9062,
+      "step": 11895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4214907578689676e-05,
+      "loss": 1.3516,
+      "step": 11896
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4213967697601713e-05,
+      "loss": 0.9609,
+      "step": 11897
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4213027771248833e-05,
+      "loss": 1.1396,
+      "step": 11898
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4212087799641125e-05,
+      "loss": 1.0918,
+      "step": 11899
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4211147782788695e-05,
+      "loss": 1.0566,
+      "step": 11900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421020772070163e-05,
+      "loss": 1.1382,
+      "step": 11901
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4209267613390036e-05,
+      "loss": 0.6372,
+      "step": 11902
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4208327460864006e-05,
+      "loss": 1.0225,
+      "step": 11903
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4207387263133638e-05,
+      "loss": 1.125,
+      "step": 11904
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.420644702020904e-05,
+      "loss": 1.1152,
+      "step": 11905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4205506732100303e-05,
+      "loss": 0.8818,
+      "step": 11906
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4204566398817531e-05,
+      "loss": 0.9531,
+      "step": 11907
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4203626020370824e-05,
+      "loss": 0.8926,
+      "step": 11908
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4202685596770281e-05,
+      "loss": 1.252,
+      "step": 11909
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4201745128026007e-05,
+      "loss": 0.9824,
+      "step": 11910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4200804614148103e-05,
+      "loss": 0.7651,
+      "step": 11911
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419986405514667e-05,
+      "loss": 1.2734,
+      "step": 11912
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4198923451031814e-05,
+      "loss": 1.0869,
+      "step": 11913
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4197982801813637e-05,
+      "loss": 1.0381,
+      "step": 11914
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4197042107502243e-05,
+      "loss": 1.1611,
+      "step": 11915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419610136810774e-05,
+      "loss": 0.9541,
+      "step": 11916
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4195160583640224e-05,
+      "loss": 1.4775,
+      "step": 11917
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4194219754109812e-05,
+      "loss": 1.3594,
+      "step": 11918
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4193278879526602e-05,
+      "loss": 0.9111,
+      "step": 11919
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4192337959900707e-05,
+      "loss": 0.9932,
+      "step": 11920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4191396995242226e-05,
+      "loss": 1.3848,
+      "step": 11921
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4190455985561273e-05,
+      "loss": 0.9082,
+      "step": 11922
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4189514930867954e-05,
+      "loss": 0.9023,
+      "step": 11923
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4188573831172373e-05,
+      "loss": 1.0459,
+      "step": 11924
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4187632686484647e-05,
+      "loss": 0.9873,
+      "step": 11925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4186691496814881e-05,
+      "loss": 0.9277,
+      "step": 11926
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4185750262173186e-05,
+      "loss": 0.9385,
+      "step": 11927
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4184808982569671e-05,
+      "loss": 1.21,
+      "step": 11928
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418386765801445e-05,
+      "loss": 1.0801,
+      "step": 11929
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4182926288517628e-05,
+      "loss": 1.2109,
+      "step": 11930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4181984874089328e-05,
+      "loss": 1.2061,
+      "step": 11931
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4181043414739648e-05,
+      "loss": 1.4209,
+      "step": 11932
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4180101910478711e-05,
+      "loss": 1.1543,
+      "step": 11933
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4179160361316629e-05,
+      "loss": 1.2168,
+      "step": 11934
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4178218767263511e-05,
+      "loss": 1.2812,
+      "step": 11935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4177277128329479e-05,
+      "loss": 0.8369,
+      "step": 11936
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.417633544452464e-05,
+      "loss": 1.3447,
+      "step": 11937
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4175393715859112e-05,
+      "loss": 0.793,
+      "step": 11938
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4174451942343013e-05,
+      "loss": 0.8101,
+      "step": 11939
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4173510123986457e-05,
+      "loss": 1.1221,
+      "step": 11940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4172568260799562e-05,
+      "loss": 1.2988,
+      "step": 11941
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4171626352792442e-05,
+      "loss": 0.9121,
+      "step": 11942
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4170684399975217e-05,
+      "loss": 0.8916,
+      "step": 11943
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4169742402358008e-05,
+      "loss": 0.8555,
+      "step": 11944
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4168800359950927e-05,
+      "loss": 1.209,
+      "step": 11945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4167858272764101e-05,
+      "loss": 1.2344,
+      "step": 11946
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4166916140807642e-05,
+      "loss": 1.0361,
+      "step": 11947
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.416597396409167e-05,
+      "loss": 1.2246,
+      "step": 11948
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4165031742626317e-05,
+      "loss": 1.2598,
+      "step": 11949
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4164089476421688e-05,
+      "loss": 1.3311,
+      "step": 11950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4163147165487919e-05,
+      "loss": 1.5137,
+      "step": 11951
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.416220480983512e-05,
+      "loss": 1.1436,
+      "step": 11952
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.416126240947342e-05,
+      "loss": 1.4258,
+      "step": 11953
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4160319964412943e-05,
+      "loss": 1.3984,
+      "step": 11954
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4159377474663807e-05,
+      "loss": 1.0801,
+      "step": 11955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4158434940236141e-05,
+      "loss": 1.4062,
+      "step": 11956
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4157492361140065e-05,
+      "loss": 1.3145,
+      "step": 11957
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4156549737385708e-05,
+      "loss": 0.9189,
+      "step": 11958
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4155607068983193e-05,
+      "loss": 0.896,
+      "step": 11959
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4154664355942647e-05,
+      "loss": 1.2051,
+      "step": 11960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4153721598274194e-05,
+      "loss": 1.3975,
+      "step": 11961
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4152778795987965e-05,
+      "loss": 1.4229,
+      "step": 11962
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.415183594909408e-05,
+      "loss": 1.1348,
+      "step": 11963
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4150893057602675e-05,
+      "loss": 1.1133,
+      "step": 11964
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4149950121523874e-05,
+      "loss": 0.8652,
+      "step": 11965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4149007140867805e-05,
+      "loss": 1.4307,
+      "step": 11966
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41480641156446e-05,
+      "loss": 1.0527,
+      "step": 11967
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4147121045864384e-05,
+      "loss": 0.8799,
+      "step": 11968
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4146177931537293e-05,
+      "loss": 1.0664,
+      "step": 11969
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4145234772673452e-05,
+      "loss": 1.2178,
+      "step": 11970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4144291569282995e-05,
+      "loss": 0.9038,
+      "step": 11971
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4143348321376056e-05,
+      "loss": 0.877,
+      "step": 11972
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4142405028962763e-05,
+      "loss": 1.4512,
+      "step": 11973
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4141461692053247e-05,
+      "loss": 1.1436,
+      "step": 11974
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414051831065765e-05,
+      "loss": 1.1572,
+      "step": 11975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4139574884786092e-05,
+      "loss": 0.9834,
+      "step": 11976
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413863141444872e-05,
+      "loss": 1.0308,
+      "step": 11977
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4137687899655662e-05,
+      "loss": 1.0654,
+      "step": 11978
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4136744340417048e-05,
+      "loss": 1.168,
+      "step": 11979
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4135800736743024e-05,
+      "loss": 0.9473,
+      "step": 11980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4134857088643719e-05,
+      "loss": 0.8613,
+      "step": 11981
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413391339612927e-05,
+      "loss": 0.9668,
+      "step": 11982
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4132969659209817e-05,
+      "loss": 0.9746,
+      "step": 11983
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4132025877895495e-05,
+      "loss": 1.0029,
+      "step": 11984
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413108205219644e-05,
+      "loss": 1.2676,
+      "step": 11985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4130138182122793e-05,
+      "loss": 1.0879,
+      "step": 11986
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4129194267684692e-05,
+      "loss": 1.1504,
+      "step": 11987
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412825030889228e-05,
+      "loss": 0.7944,
+      "step": 11988
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4127306305755686e-05,
+      "loss": 1.1279,
+      "step": 11989
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412636225828506e-05,
+      "loss": 0.9058,
+      "step": 11990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412541816649054e-05,
+      "loss": 1.0195,
+      "step": 11991
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4124474030382264e-05,
+      "loss": 0.7734,
+      "step": 11992
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4123529849970378e-05,
+      "loss": 0.8848,
+      "step": 11993
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4122585625265019e-05,
+      "loss": 1.4375,
+      "step": 11994
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412164135627634e-05,
+      "loss": 0.9619,
+      "step": 11995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4120697043014471e-05,
+      "loss": 0.9258,
+      "step": 11996
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4119752685489563e-05,
+      "loss": 1.082,
+      "step": 11997
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4118808283711757e-05,
+      "loss": 1.1758,
+      "step": 11998
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4117863837691198e-05,
+      "loss": 0.9854,
+      "step": 11999
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4116919347438032e-05,
+      "loss": 1.2979,
+      "step": 12000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4115974812962408e-05,
+      "loss": 1.4326,
+      "step": 12001
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4115030234274461e-05,
+      "loss": 0.8926,
+      "step": 12002
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4114085611384348e-05,
+      "loss": 1.3594,
+      "step": 12003
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4113140944302208e-05,
+      "loss": 1.2451,
+      "step": 12004
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4112196233038196e-05,
+      "loss": 0.6965,
+      "step": 12005
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4111251477602457e-05,
+      "loss": 1.1865,
+      "step": 12006
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4110306678005133e-05,
+      "loss": 0.9468,
+      "step": 12007
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4109361834256381e-05,
+      "loss": 1.4473,
+      "step": 12008
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4108416946366345e-05,
+      "loss": 1.1875,
+      "step": 12009
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4107472014345179e-05,
+      "loss": 1.0537,
+      "step": 12010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4106527038203029e-05,
+      "loss": 1.1465,
+      "step": 12011
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4105582017950046e-05,
+      "loss": 1.0679,
+      "step": 12012
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4104636953596384e-05,
+      "loss": 0.8857,
+      "step": 12013
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4103691845152192e-05,
+      "loss": 1.2695,
+      "step": 12014
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4102746692627625e-05,
+      "loss": 1.0713,
+      "step": 12015
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410180149603283e-05,
+      "loss": 0.9751,
+      "step": 12016
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4100856255377964e-05,
+      "loss": 0.9551,
+      "step": 12017
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409991097067318e-05,
+      "loss": 0.9082,
+      "step": 12018
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4098965641928635e-05,
+      "loss": 1.1768,
+      "step": 12019
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4098020269154476e-05,
+      "loss": 1.1562,
+      "step": 12020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4097074852360866e-05,
+      "loss": 0.7339,
+      "step": 12021
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409612939155795e-05,
+      "loss": 0.9238,
+      "step": 12022
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4095183886755894e-05,
+      "loss": 0.9243,
+      "step": 12023
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4094238337964852e-05,
+      "loss": 1.3291,
+      "step": 12024
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4093292745194975e-05,
+      "loss": 1.1592,
+      "step": 12025
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4092347108456425e-05,
+      "loss": 0.9814,
+      "step": 12026
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4091401427759361e-05,
+      "loss": 0.8906,
+      "step": 12027
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4090455703113937e-05,
+      "loss": 0.8564,
+      "step": 12028
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4089509934530315e-05,
+      "loss": 1.1807,
+      "step": 12029
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4088564122018653e-05,
+      "loss": 0.71,
+      "step": 12030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.408761826558911e-05,
+      "loss": 1.1426,
+      "step": 12031
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4086672365251847e-05,
+      "loss": 1.3018,
+      "step": 12032
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408572642101702e-05,
+      "loss": 1.5449,
+      "step": 12033
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.40847804328948e-05,
+      "loss": 1.1641,
+      "step": 12034
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4083834400895338e-05,
+      "loss": 0.9395,
+      "step": 12035
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4082888325028803e-05,
+      "loss": 1.4365,
+      "step": 12036
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4081942205305351e-05,
+      "loss": 0.9814,
+      "step": 12037
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4080996041735153e-05,
+      "loss": 1.1348,
+      "step": 12038
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4080049834328366e-05,
+      "loss": 1.0049,
+      "step": 12039
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4079103583095157e-05,
+      "loss": 1.126,
+      "step": 12040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4078157288045686e-05,
+      "loss": 1.0498,
+      "step": 12041
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4077210949190124e-05,
+      "loss": 0.8237,
+      "step": 12042
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4076264566538633e-05,
+      "loss": 1.0684,
+      "step": 12043
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4075318140101377e-05,
+      "loss": 1.0596,
+      "step": 12044
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4074371669888526e-05,
+      "loss": 0.9775,
+      "step": 12045
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407342515591024e-05,
+      "loss": 0.9365,
+      "step": 12046
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072478598176698e-05,
+      "loss": 1.2939,
+      "step": 12047
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4071531996698054e-05,
+      "loss": 0.9067,
+      "step": 12048
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4070585351484485e-05,
+      "loss": 1.1895,
+      "step": 12049
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4069638662546158e-05,
+      "loss": 1.1758,
+      "step": 12050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4068691929893237e-05,
+      "loss": 1.042,
+      "step": 12051
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4067745153535901e-05,
+      "loss": 0.7302,
+      "step": 12052
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4066798333484308e-05,
+      "loss": 1.1318,
+      "step": 12053
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406585146974864e-05,
+      "loss": 0.8496,
+      "step": 12054
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4064904562339059e-05,
+      "loss": 0.8179,
+      "step": 12055
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4063957611265741e-05,
+      "loss": 1.293,
+      "step": 12056
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4063010616538856e-05,
+      "loss": 1.1367,
+      "step": 12057
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4062063578168577e-05,
+      "loss": 1.1797,
+      "step": 12058
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4061116496165077e-05,
+      "loss": 1.0713,
+      "step": 12059
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406016937053853e-05,
+      "loss": 1.1094,
+      "step": 12060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4059222201299107e-05,
+      "loss": 0.9844,
+      "step": 12061
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4058274988456983e-05,
+      "loss": 1.0713,
+      "step": 12062
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4057327732022337e-05,
+      "loss": 0.9502,
+      "step": 12063
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4056380432005336e-05,
+      "loss": 0.9814,
+      "step": 12064
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4055433088416163e-05,
+      "loss": 1.0947,
+      "step": 12065
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4054485701264989e-05,
+      "loss": 1.3896,
+      "step": 12066
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4053538270561996e-05,
+      "loss": 0.9033,
+      "step": 12067
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4052590796317355e-05,
+      "loss": 1.3027,
+      "step": 12068
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4051643278541246e-05,
+      "loss": 1.4463,
+      "step": 12069
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4050695717243846e-05,
+      "loss": 1.0352,
+      "step": 12070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4049748112435335e-05,
+      "loss": 1.4395,
+      "step": 12071
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404880046412589e-05,
+      "loss": 0.8699,
+      "step": 12072
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4047852772325692e-05,
+      "loss": 1.2754,
+      "step": 12073
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404690503704492e-05,
+      "loss": 1.3232,
+      "step": 12074
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4045957258293754e-05,
+      "loss": 1.2031,
+      "step": 12075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4045009436082377e-05,
+      "loss": 0.6626,
+      "step": 12076
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4044061570420964e-05,
+      "loss": 1.0811,
+      "step": 12077
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4043113661319705e-05,
+      "loss": 1.27,
+      "step": 12078
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4042165708788773e-05,
+      "loss": 1.0654,
+      "step": 12079
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404121771283836e-05,
+      "loss": 0.8115,
+      "step": 12080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4040269673478644e-05,
+      "loss": 1.2051,
+      "step": 12081
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4039321590719807e-05,
+      "loss": 0.9072,
+      "step": 12082
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4038373464572036e-05,
+      "loss": 1.0801,
+      "step": 12083
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4037425295045513e-05,
+      "loss": 1.6543,
+      "step": 12084
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4036477082150426e-05,
+      "loss": 0.9248,
+      "step": 12085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4035528825896956e-05,
+      "loss": 1.3857,
+      "step": 12086
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4034580526295294e-05,
+      "loss": 0.8345,
+      "step": 12087
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4033632183355621e-05,
+      "loss": 1.1982,
+      "step": 12088
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4032683797088131e-05,
+      "loss": 0.8115,
+      "step": 12089
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4031735367503e-05,
+      "loss": 1.127,
+      "step": 12090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4030786894610428e-05,
+      "loss": 1.6191,
+      "step": 12091
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4029838378420595e-05,
+      "loss": 1.249,
+      "step": 12092
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4028889818943694e-05,
+      "loss": 0.811,
+      "step": 12093
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402794121618991e-05,
+      "loss": 0.6873,
+      "step": 12094
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4026992570169435e-05,
+      "loss": 1.3838,
+      "step": 12095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4026043880892464e-05,
+      "loss": 1.0664,
+      "step": 12096
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4025095148369175e-05,
+      "loss": 0.6455,
+      "step": 12097
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402414637260977e-05,
+      "loss": 0.9438,
+      "step": 12098
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4023197553624438e-05,
+      "loss": 1.2891,
+      "step": 12099
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4022248691423366e-05,
+      "loss": 1.9316,
+      "step": 12100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4021299786016749e-05,
+      "loss": 0.9434,
+      "step": 12101
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4020350837414785e-05,
+      "loss": 1.292,
+      "step": 12102
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4019401845627658e-05,
+      "loss": 0.7983,
+      "step": 12103
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401845281066557e-05,
+      "loss": 1.4185,
+      "step": 12104
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4017503732538712e-05,
+      "loss": 0.9336,
+      "step": 12105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4016554611257276e-05,
+      "loss": 1.5273,
+      "step": 12106
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4015605446831464e-05,
+      "loss": 1.0996,
+      "step": 12107
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401465623927146e-05,
+      "loss": 0.9727,
+      "step": 12108
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4013706988587475e-05,
+      "loss": 0.9756,
+      "step": 12109
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4012757694789691e-05,
+      "loss": 1.0469,
+      "step": 12110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4011808357888317e-05,
+      "loss": 1.0303,
+      "step": 12111
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4010858977893542e-05,
+      "loss": 0.9072,
+      "step": 12112
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4009909554815568e-05,
+      "loss": 0.9629,
+      "step": 12113
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4008960088664591e-05,
+      "loss": 1.1826,
+      "step": 12114
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4008010579450812e-05,
+      "loss": 0.8916,
+      "step": 12115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4007061027184432e-05,
+      "loss": 1.3604,
+      "step": 12116
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4006111431875647e-05,
+      "loss": 0.7363,
+      "step": 12117
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4005161793534656e-05,
+      "loss": 1.1006,
+      "step": 12118
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004212112171662e-05,
+      "loss": 1.1436,
+      "step": 12119
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4003262387796872e-05,
+      "loss": 0.9131,
+      "step": 12120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4002312620420476e-05,
+      "loss": 1.0391,
+      "step": 12121
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4001362810052684e-05,
+      "loss": 1.0371,
+      "step": 12122
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4000412956703695e-05,
+      "loss": 1.3604,
+      "step": 12123
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3999463060383716e-05,
+      "loss": 0.7993,
+      "step": 12124
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3998513121102945e-05,
+      "loss": 1.0781,
+      "step": 12125
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399756313887159e-05,
+      "loss": 1.1309,
+      "step": 12126
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3996613113699855e-05,
+      "loss": 1.0088,
+      "step": 12127
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3995663045597943e-05,
+      "loss": 0.9639,
+      "step": 12128
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399471293457606e-05,
+      "loss": 1.3555,
+      "step": 12129
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3993762780644413e-05,
+      "loss": 1.1719,
+      "step": 12130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3992812583813207e-05,
+      "loss": 0.9863,
+      "step": 12131
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3991862344092649e-05,
+      "loss": 0.7754,
+      "step": 12132
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3990912061492945e-05,
+      "loss": 0.8232,
+      "step": 12133
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3989961736024302e-05,
+      "loss": 0.4783,
+      "step": 12134
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3989011367696934e-05,
+      "loss": 1.1689,
+      "step": 12135
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398806095652104e-05,
+      "loss": 1.4102,
+      "step": 12136
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398711050250684e-05,
+      "loss": 0.79,
+      "step": 12137
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3986160005664533e-05,
+      "loss": 1.1826,
+      "step": 12138
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3985209466004333e-05,
+      "loss": 0.8584,
+      "step": 12139
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3984258883536455e-05,
+      "loss": 0.8906,
+      "step": 12140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3983308258271102e-05,
+      "loss": 1.1934,
+      "step": 12141
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3982357590218491e-05,
+      "loss": 0.998,
+      "step": 12142
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3981406879388831e-05,
+      "loss": 1.2051,
+      "step": 12143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3980456125792335e-05,
+      "loss": 1.1689,
+      "step": 12144
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3979505329439215e-05,
+      "loss": 0.7656,
+      "step": 12145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3978554490339687e-05,
+      "loss": 1.1543,
+      "step": 12146
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3977603608503958e-05,
+      "loss": 0.8477,
+      "step": 12147
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3976652683942247e-05,
+      "loss": 0.8799,
+      "step": 12148
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.397570171666477e-05,
+      "loss": 1.0762,
+      "step": 12149
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3974750706681736e-05,
+      "loss": 1.2393,
+      "step": 12150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.397379965400337e-05,
+      "loss": 0.8926,
+      "step": 12151
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3972848558639878e-05,
+      "loss": 1.1816,
+      "step": 12152
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3971897420601482e-05,
+      "loss": 0.9619,
+      "step": 12153
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3970946239898394e-05,
+      "loss": 1.1484,
+      "step": 12154
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3969995016540835e-05,
+      "loss": 1.2979,
+      "step": 12155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3969043750539024e-05,
+      "loss": 0.8467,
+      "step": 12156
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3968092441903176e-05,
+      "loss": 1.2822,
+      "step": 12157
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396714109064351e-05,
+      "loss": 1.1641,
+      "step": 12158
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3966189696770248e-05,
+      "loss": 1.1455,
+      "step": 12159
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3965238260293603e-05,
+      "loss": 0.6763,
+      "step": 12160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3964286781223804e-05,
+      "loss": 1.1387,
+      "step": 12161
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3963335259571063e-05,
+      "loss": 1.0508,
+      "step": 12162
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3962383695345607e-05,
+      "loss": 1.0933,
+      "step": 12163
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3961432088557656e-05,
+      "loss": 0.9434,
+      "step": 12164
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3960480439217427e-05,
+      "loss": 1.0693,
+      "step": 12165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3959528747335149e-05,
+      "loss": 1.2925,
+      "step": 12166
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.395857701292104e-05,
+      "loss": 0.9453,
+      "step": 12167
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3957625235985327e-05,
+      "loss": 1.1514,
+      "step": 12168
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.395667341653823e-05,
+      "loss": 1.2412,
+      "step": 12169
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3955721554589979e-05,
+      "loss": 0.8213,
+      "step": 12170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.395476965015079e-05,
+      "loss": 1.1113,
+      "step": 12171
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3953817703230892e-05,
+      "loss": 1.543,
+      "step": 12172
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3952865713840514e-05,
+      "loss": 0.9727,
+      "step": 12173
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3951913681989875e-05,
+      "loss": 0.8857,
+      "step": 12174
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3950961607689206e-05,
+      "loss": 1.0771,
+      "step": 12175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3950009490948734e-05,
+      "loss": 1.3086,
+      "step": 12176
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3949057331778688e-05,
+      "loss": 1.1187,
+      "step": 12177
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3948105130189287e-05,
+      "loss": 1.4941,
+      "step": 12178
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3947152886190772e-05,
+      "loss": 0.7705,
+      "step": 12179
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3946200599793358e-05,
+      "loss": 1.1084,
+      "step": 12180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3945248271007286e-05,
+      "loss": 0.8955,
+      "step": 12181
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3944295899842778e-05,
+      "loss": 0.9102,
+      "step": 12182
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3943343486310069e-05,
+      "loss": 1.3262,
+      "step": 12183
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3942391030419385e-05,
+      "loss": 0.8555,
+      "step": 12184
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3941438532180956e-05,
+      "loss": 1.6738,
+      "step": 12185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3940485991605022e-05,
+      "loss": 0.7383,
+      "step": 12186
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3939533408701806e-05,
+      "loss": 1.3057,
+      "step": 12187
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3938580783481545e-05,
+      "loss": 1.0898,
+      "step": 12188
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3937628115954469e-05,
+      "loss": 0.9502,
+      "step": 12189
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3936675406130812e-05,
+      "loss": 1.2217,
+      "step": 12190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3935722654020811e-05,
+      "loss": 0.7295,
+      "step": 12191
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3934769859634693e-05,
+      "loss": 1.1177,
+      "step": 12192
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3933817022982701e-05,
+      "loss": 1.249,
+      "step": 12193
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3932864144075064e-05,
+      "loss": 1.0596,
+      "step": 12194
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3931911222922024e-05,
+      "loss": 0.9526,
+      "step": 12195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3930958259533805e-05,
+      "loss": 1.1875,
+      "step": 12196
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3930005253920656e-05,
+      "loss": 1.3311,
+      "step": 12197
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3929052206092803e-05,
+      "loss": 1.2783,
+      "step": 12198
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3928099116060495e-05,
+      "loss": 1.1523,
+      "step": 12199
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3927145983833961e-05,
+      "loss": 1.0879,
+      "step": 12200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392619280942344e-05,
+      "loss": 1.0552,
+      "step": 12201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3925239592839175e-05,
+      "loss": 1.2168,
+      "step": 12202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3924286334091402e-05,
+      "loss": 1.2939,
+      "step": 12203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392333303319036e-05,
+      "loss": 1.2793,
+      "step": 12204
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392237969014629e-05,
+      "loss": 1.2803,
+      "step": 12205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921426304969435e-05,
+      "loss": 1.1934,
+      "step": 12206
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392047287767003e-05,
+      "loss": 1.2031,
+      "step": 12207
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3919519408258327e-05,
+      "loss": 1.1006,
+      "step": 12208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3918565896744552e-05,
+      "loss": 0.9409,
+      "step": 12209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3917612343138962e-05,
+      "loss": 1.2539,
+      "step": 12210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3916658747451788e-05,
+      "loss": 1.3828,
+      "step": 12211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3915705109693286e-05,
+      "loss": 1.0391,
+      "step": 12212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3914751429873688e-05,
+      "loss": 0.8804,
+      "step": 12213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3913797708003244e-05,
+      "loss": 1.168,
+      "step": 12214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3912843944092194e-05,
+      "loss": 1.1816,
+      "step": 12215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3911890138150788e-05,
+      "loss": 1.1748,
+      "step": 12216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3910936290189269e-05,
+      "loss": 1.124,
+      "step": 12217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3909982400217885e-05,
+      "loss": 0.9033,
+      "step": 12218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3909028468246881e-05,
+      "loss": 1.0303,
+      "step": 12219
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3908074494286504e-05,
+      "loss": 1.0444,
+      "step": 12220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3907120478347e-05,
+      "loss": 1.0771,
+      "step": 12221
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3906166420438618e-05,
+      "loss": 1.3965,
+      "step": 12222
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3905212320571606e-05,
+      "loss": 0.959,
+      "step": 12223
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3904258178756209e-05,
+      "loss": 1.1182,
+      "step": 12224
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3903303995002684e-05,
+      "loss": 1.1787,
+      "step": 12225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3902349769321272e-05,
+      "loss": 0.8188,
+      "step": 12226
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390139550172223e-05,
+      "loss": 0.916,
+      "step": 12227
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3900441192215803e-05,
+      "loss": 1.0273,
+      "step": 12228
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3899486840812242e-05,
+      "loss": 1.4619,
+      "step": 12229
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3898532447521804e-05,
+      "loss": 1.0439,
+      "step": 12230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3897578012354738e-05,
+      "loss": 1.0156,
+      "step": 12231
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3896623535321292e-05,
+      "loss": 1.082,
+      "step": 12232
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3895669016431723e-05,
+      "loss": 1.6641,
+      "step": 12233
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3894714455696283e-05,
+      "loss": 1.1572,
+      "step": 12234
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3893759853125227e-05,
+      "loss": 1.1484,
+      "step": 12235
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3892805208728807e-05,
+      "loss": 0.7031,
+      "step": 12236
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3891850522517276e-05,
+      "loss": 1.2031,
+      "step": 12237
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3890895794500891e-05,
+      "loss": 0.5933,
+      "step": 12238
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388994102468991e-05,
+      "loss": 1.2988,
+      "step": 12239
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3888986213094584e-05,
+      "loss": 1.0237,
+      "step": 12240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3888031359725173e-05,
+      "loss": 0.7407,
+      "step": 12241
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388707646459193e-05,
+      "loss": 1.0239,
+      "step": 12242
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3886121527705116e-05,
+      "loss": 1.2432,
+      "step": 12243
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3885166549074984e-05,
+      "loss": 1.6152,
+      "step": 12244
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3884211528711798e-05,
+      "loss": 0.8945,
+      "step": 12245
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388325646662581e-05,
+      "loss": 1.2178,
+      "step": 12246
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3882301362827283e-05,
+      "loss": 0.9893,
+      "step": 12247
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3881346217326477e-05,
+      "loss": 1.1191,
+      "step": 12248
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3880391030133652e-05,
+      "loss": 1.2197,
+      "step": 12249
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387943580125906e-05,
+      "loss": 1.0098,
+      "step": 12250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3878480530712975e-05,
+      "loss": 1.4102,
+      "step": 12251
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387752521850565e-05,
+      "loss": 0.9639,
+      "step": 12252
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3876569864647345e-05,
+      "loss": 1.2822,
+      "step": 12253
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3875614469148329e-05,
+      "loss": 1.3281,
+      "step": 12254
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3874659032018857e-05,
+      "loss": 1.3506,
+      "step": 12255
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38737035532692e-05,
+      "loss": 0.9841,
+      "step": 12256
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3872748032909615e-05,
+      "loss": 0.7549,
+      "step": 12257
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3871792470950367e-05,
+      "loss": 1.3398,
+      "step": 12258
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3870836867401723e-05,
+      "loss": 1.0654,
+      "step": 12259
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3869881222273945e-05,
+      "loss": 1.3896,
+      "step": 12260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3868925535577298e-05,
+      "loss": 1.2529,
+      "step": 12261
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386796980732205e-05,
+      "loss": 1.0732,
+      "step": 12262
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3867014037518467e-05,
+      "loss": 0.915,
+      "step": 12263
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3866058226176812e-05,
+      "loss": 1.0459,
+      "step": 12264
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3865102373307358e-05,
+      "loss": 1.248,
+      "step": 12265
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3864146478920366e-05,
+      "loss": 0.8564,
+      "step": 12266
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3863190543026108e-05,
+      "loss": 0.6816,
+      "step": 12267
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3862234565634849e-05,
+      "loss": 0.9058,
+      "step": 12268
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3861278546756864e-05,
+      "loss": 1.4707,
+      "step": 12269
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3860322486402416e-05,
+      "loss": 0.856,
+      "step": 12270
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3859366384581776e-05,
+      "loss": 0.9248,
+      "step": 12271
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3858410241305214e-05,
+      "loss": 1.2344,
+      "step": 12272
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3857454056583002e-05,
+      "loss": 0.9072,
+      "step": 12273
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385649783042541e-05,
+      "loss": 1.0283,
+      "step": 12274
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3855541562842711e-05,
+      "loss": 1.0566,
+      "step": 12275
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3854585253845176e-05,
+      "loss": 1.1387,
+      "step": 12276
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3853628903443075e-05,
+      "loss": 1.0625,
+      "step": 12277
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3852672511646683e-05,
+      "loss": 1.2754,
+      "step": 12278
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3851716078466275e-05,
+      "loss": 0.877,
+      "step": 12279
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3850759603912122e-05,
+      "loss": 1.0645,
+      "step": 12280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3849803087994496e-05,
+      "loss": 1.103,
+      "step": 12281
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3848846530723678e-05,
+      "loss": 1.1475,
+      "step": 12282
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384788993210994e-05,
+      "loss": 1.2012,
+      "step": 12283
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384693329216355e-05,
+      "loss": 1.0566,
+      "step": 12284
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3845976610894795e-05,
+      "loss": 1.1455,
+      "step": 12285
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3845019888313946e-05,
+      "loss": 1.1172,
+      "step": 12286
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3844063124431284e-05,
+      "loss": 0.9941,
+      "step": 12287
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384310631925708e-05,
+      "loss": 1.3291,
+      "step": 12288
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3842149472801616e-05,
+      "loss": 1.043,
+      "step": 12289
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3841192585075164e-05,
+      "loss": 0.9072,
+      "step": 12290
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3840235656088012e-05,
+      "loss": 0.8906,
+      "step": 12291
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3839278685850434e-05,
+      "loss": 1.1006,
+      "step": 12292
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3838321674372708e-05,
+      "loss": 0.9077,
+      "step": 12293
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3837364621665114e-05,
+      "loss": 1.2285,
+      "step": 12294
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3836407527737938e-05,
+      "loss": 0.9639,
+      "step": 12295
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3835450392601453e-05,
+      "loss": 1.5898,
+      "step": 12296
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3834493216265945e-05,
+      "loss": 1.2666,
+      "step": 12297
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3833535998741696e-05,
+      "loss": 0.8848,
+      "step": 12298
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3832578740038983e-05,
+      "loss": 1.0068,
+      "step": 12299
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3831621440168097e-05,
+      "loss": 1.1289,
+      "step": 12300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3830664099139312e-05,
+      "loss": 1.209,
+      "step": 12301
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3829706716962916e-05,
+      "loss": 1.022,
+      "step": 12302
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3828749293649193e-05,
+      "loss": 1.0889,
+      "step": 12303
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3827791829208425e-05,
+      "loss": 1.1885,
+      "step": 12304
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3826834323650899e-05,
+      "loss": 1.2441,
+      "step": 12305
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3825876776986899e-05,
+      "loss": 1.1133,
+      "step": 12306
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3824919189226714e-05,
+      "loss": 0.833,
+      "step": 12307
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3823961560380623e-05,
+      "loss": 0.9756,
+      "step": 12308
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382300389045892e-05,
+      "loss": 0.9326,
+      "step": 12309
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3822046179471887e-05,
+      "loss": 1.3643,
+      "step": 12310
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3821088427429814e-05,
+      "loss": 0.821,
+      "step": 12311
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3820130634342985e-05,
+      "loss": 1.1211,
+      "step": 12312
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3819172800221693e-05,
+      "loss": 1.3613,
+      "step": 12313
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3818214925076226e-05,
+      "loss": 1.1162,
+      "step": 12314
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3817257008916871e-05,
+      "loss": 0.6494,
+      "step": 12315
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3816299051753919e-05,
+      "loss": 0.9961,
+      "step": 12316
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3815341053597657e-05,
+      "loss": 0.8213,
+      "step": 12317
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381438301445838e-05,
+      "loss": 1.2686,
+      "step": 12318
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3813424934346375e-05,
+      "loss": 0.541,
+      "step": 12319
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381246681327194e-05,
+      "loss": 1.2041,
+      "step": 12320
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3811508651245356e-05,
+      "loss": 1.3359,
+      "step": 12321
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3810550448276925e-05,
+      "loss": 0.8721,
+      "step": 12322
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3809592204376936e-05,
+      "loss": 1.2715,
+      "step": 12323
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3808633919555683e-05,
+      "loss": 0.9014,
+      "step": 12324
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3807675593823456e-05,
+      "loss": 0.8135,
+      "step": 12325
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3806717227190553e-05,
+      "loss": 0.9941,
+      "step": 12326
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3805758819667264e-05,
+      "loss": 0.8999,
+      "step": 12327
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3804800371263889e-05,
+      "loss": 0.6328,
+      "step": 12328
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3803841881990722e-05,
+      "loss": 1.4473,
+      "step": 12329
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3802883351858057e-05,
+      "loss": 1.042,
+      "step": 12330
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3801924780876194e-05,
+      "loss": 1.2754,
+      "step": 12331
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3800966169055426e-05,
+      "loss": 1.0498,
+      "step": 12332
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380000751640605e-05,
+      "loss": 1.0576,
+      "step": 12333
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.379904882293836e-05,
+      "loss": 1.4082,
+      "step": 12334
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3798090088662664e-05,
+      "loss": 1.5215,
+      "step": 12335
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3797131313589251e-05,
+      "loss": 0.8311,
+      "step": 12336
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3796172497728427e-05,
+      "loss": 1.4316,
+      "step": 12337
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3795213641090483e-05,
+      "loss": 1.0098,
+      "step": 12338
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3794254743685729e-05,
+      "loss": 1.0713,
+      "step": 12339
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3793295805524455e-05,
+      "loss": 1.1357,
+      "step": 12340
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3792336826616968e-05,
+      "loss": 1.1885,
+      "step": 12341
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.379137780697357e-05,
+      "loss": 1.1191,
+      "step": 12342
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3790418746604552e-05,
+      "loss": 1.0645,
+      "step": 12343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3789459645520231e-05,
+      "loss": 1.4502,
+      "step": 12344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3788500503730899e-05,
+      "loss": 0.9629,
+      "step": 12345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3787541321246863e-05,
+      "loss": 1.2119,
+      "step": 12346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3786582098078421e-05,
+      "loss": 1.0527,
+      "step": 12347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3785622834235883e-05,
+      "loss": 1.1807,
+      "step": 12348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3784663529729549e-05,
+      "loss": 1.001,
+      "step": 12349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3783704184569727e-05,
+      "loss": 0.9502,
+      "step": 12350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3782744798766716e-05,
+      "loss": 1.0312,
+      "step": 12351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3781785372330826e-05,
+      "loss": 1.123,
+      "step": 12352
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3780825905272367e-05,
+      "loss": 0.9219,
+      "step": 12353
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3779866397601635e-05,
+      "loss": 1.3652,
+      "step": 12354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3778906849328948e-05,
+      "loss": 1.0654,
+      "step": 12355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3777947260464599e-05,
+      "loss": 1.1621,
+      "step": 12356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.377698763101891e-05,
+      "loss": 0.7471,
+      "step": 12357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3776027961002178e-05,
+      "loss": 1.0088,
+      "step": 12358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3775068250424722e-05,
+      "loss": 1.6035,
+      "step": 12359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3774108499296838e-05,
+      "loss": 1.2051,
+      "step": 12360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3773148707628848e-05,
+      "loss": 0.9404,
+      "step": 12361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3772188875431054e-05,
+      "loss": 0.958,
+      "step": 12362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3771229002713767e-05,
+      "loss": 0.915,
+      "step": 12363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3770269089487297e-05,
+      "loss": 1.377,
+      "step": 12364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3769309135761961e-05,
+      "loss": 1.4121,
+      "step": 12365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3768349141548063e-05,
+      "loss": 1.126,
+      "step": 12366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376738910685592e-05,
+      "loss": 1.3105,
+      "step": 12367
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3766429031695844e-05,
+      "loss": 0.7646,
+      "step": 12368
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3765468916078141e-05,
+      "loss": 1.2012,
+      "step": 12369
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3764508760013136e-05,
+      "loss": 1.0449,
+      "step": 12370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376354856351113e-05,
+      "loss": 0.7358,
+      "step": 12371
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3762588326582448e-05,
+      "loss": 1.0303,
+      "step": 12372
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.37616280492374e-05,
+      "loss": 0.7729,
+      "step": 12373
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3760667731486296e-05,
+      "loss": 1.3359,
+      "step": 12374
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3759707373339462e-05,
+      "loss": 1.0254,
+      "step": 12375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3758746974807205e-05,
+      "loss": 1.1572,
+      "step": 12376
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3757786535899844e-05,
+      "loss": 0.7007,
+      "step": 12377
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3756826056627694e-05,
+      "loss": 1.0293,
+      "step": 12378
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3755865537001076e-05,
+      "loss": 1.0957,
+      "step": 12379
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3754904977030306e-05,
+      "loss": 0.9819,
+      "step": 12380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3753944376725704e-05,
+      "loss": 0.9309,
+      "step": 12381
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375298373609758e-05,
+      "loss": 0.8027,
+      "step": 12382
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375202305515626e-05,
+      "loss": 1.291,
+      "step": 12383
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3751062333912066e-05,
+      "loss": 1.1182,
+      "step": 12384
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3750101572375311e-05,
+      "loss": 0.9932,
+      "step": 12385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3749140770556322e-05,
+      "loss": 0.9541,
+      "step": 12386
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3748179928465408e-05,
+      "loss": 1.124,
+      "step": 12387
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3747219046112905e-05,
+      "loss": 1.04,
+      "step": 12388
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3746258123509123e-05,
+      "loss": 1.0195,
+      "step": 12389
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374529716066439e-05,
+      "loss": 1.0322,
+      "step": 12390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3744336157589025e-05,
+      "loss": 1.0444,
+      "step": 12391
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3743375114293352e-05,
+      "loss": 1.2793,
+      "step": 12392
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3742414030787692e-05,
+      "loss": 0.9277,
+      "step": 12393
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3741452907082373e-05,
+      "loss": 0.8721,
+      "step": 12394
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3740491743187718e-05,
+      "loss": 1.0518,
+      "step": 12395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3739530539114048e-05,
+      "loss": 0.9609,
+      "step": 12396
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373856929487169e-05,
+      "loss": 1.3145,
+      "step": 12397
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3737608010470973e-05,
+      "loss": 0.8789,
+      "step": 12398
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373664668592222e-05,
+      "loss": 1.4766,
+      "step": 12399
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373568532123575e-05,
+      "loss": 1.0322,
+      "step": 12400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3734723916421902e-05,
+      "loss": 0.8467,
+      "step": 12401
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3733762471490992e-05,
+      "loss": 1.0474,
+      "step": 12402
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373280098645336e-05,
+      "loss": 1.292,
+      "step": 12403
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373183946131932e-05,
+      "loss": 1.1094,
+      "step": 12404
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373087789609921e-05,
+      "loss": 0.8154,
+      "step": 12405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3729916290803354e-05,
+      "loss": 0.9521,
+      "step": 12406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3728954645442085e-05,
+      "loss": 0.9639,
+      "step": 12407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3727992960025729e-05,
+      "loss": 1.0234,
+      "step": 12408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3727031234564618e-05,
+      "loss": 1.4082,
+      "step": 12409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3726069469069083e-05,
+      "loss": 0.7612,
+      "step": 12410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3725107663549452e-05,
+      "loss": 0.9287,
+      "step": 12411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3724145818016064e-05,
+      "loss": 1.375,
+      "step": 12412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372318393247924e-05,
+      "loss": 1.0283,
+      "step": 12413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3722222006949322e-05,
+      "loss": 0.9463,
+      "step": 12414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3721260041436632e-05,
+      "loss": 1.2656,
+      "step": 12415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3720298035951513e-05,
+      "loss": 1.0903,
+      "step": 12416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3719335990504296e-05,
+      "loss": 1.1338,
+      "step": 12417
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3718373905105313e-05,
+      "loss": 0.8984,
+      "step": 12418
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3717411779764898e-05,
+      "loss": 0.5007,
+      "step": 12419
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3716449614493388e-05,
+      "loss": 1.2705,
+      "step": 12420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3715487409301117e-05,
+      "loss": 1.042,
+      "step": 12421
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3714525164198418e-05,
+      "loss": 1.2695,
+      "step": 12422
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3713562879195632e-05,
+      "loss": 0.6797,
+      "step": 12423
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3712600554303095e-05,
+      "loss": 0.9502,
+      "step": 12424
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3711638189531142e-05,
+      "loss": 0.6611,
+      "step": 12425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3710675784890108e-05,
+      "loss": 1.1797,
+      "step": 12426
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3709713340390336e-05,
+      "loss": 1.0996,
+      "step": 12427
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370875085604216e-05,
+      "loss": 1.085,
+      "step": 12428
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3707788331855923e-05,
+      "loss": 1.1016,
+      "step": 12429
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370682576784196e-05,
+      "loss": 0.8867,
+      "step": 12430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3705863164010611e-05,
+      "loss": 0.9922,
+      "step": 12431
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370490052037222e-05,
+      "loss": 0.6191,
+      "step": 12432
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370393783693712e-05,
+      "loss": 0.7441,
+      "step": 12433
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3702975113715661e-05,
+      "loss": 0.9136,
+      "step": 12434
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3702012350718178e-05,
+      "loss": 1.002,
+      "step": 12435
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3701049547955013e-05,
+      "loss": 0.9363,
+      "step": 12436
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3700086705436509e-05,
+      "loss": 1.2168,
+      "step": 12437
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3699123823173009e-05,
+      "loss": 0.8574,
+      "step": 12438
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3698160901174855e-05,
+      "loss": 0.9785,
+      "step": 12439
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3697197939452392e-05,
+      "loss": 1.0469,
+      "step": 12440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3696234938015962e-05,
+      "loss": 0.876,
+      "step": 12441
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369527189687591e-05,
+      "loss": 0.9414,
+      "step": 12442
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3694308816042582e-05,
+      "loss": 0.9697,
+      "step": 12443
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3693345695526318e-05,
+      "loss": 1.2588,
+      "step": 12444
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3692382535337473e-05,
+      "loss": 1.082,
+      "step": 12445
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369141933548638e-05,
+      "loss": 1.1113,
+      "step": 12446
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3690456095983397e-05,
+      "loss": 1.5508,
+      "step": 12447
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3689492816838867e-05,
+      "loss": 1.1211,
+      "step": 12448
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3688529498063132e-05,
+      "loss": 0.9131,
+      "step": 12449
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3687566139666549e-05,
+      "loss": 1.1016,
+      "step": 12450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3686602741659458e-05,
+      "loss": 0.9766,
+      "step": 12451
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3685639304052209e-05,
+      "loss": 1.1758,
+      "step": 12452
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3684675826855153e-05,
+      "loss": 0.9268,
+      "step": 12453
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3683712310078644e-05,
+      "loss": 1.1338,
+      "step": 12454
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368274875373302e-05,
+      "loss": 0.9526,
+      "step": 12455
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3681785157828642e-05,
+      "loss": 0.958,
+      "step": 12456
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3680821522375853e-05,
+      "loss": 1.291,
+      "step": 12457
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.367985784738501e-05,
+      "loss": 1.2607,
+      "step": 12458
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3678894132866458e-05,
+      "loss": 1.252,
+      "step": 12459
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3677930378830558e-05,
+      "loss": 1.3369,
+      "step": 12460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3676966585287654e-05,
+      "loss": 0.8936,
+      "step": 12461
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3676002752248102e-05,
+      "loss": 1.2012,
+      "step": 12462
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3675038879722254e-05,
+      "loss": 0.8574,
+      "step": 12463
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3674074967720464e-05,
+      "loss": 0.6582,
+      "step": 12464
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3673111016253086e-05,
+      "loss": 0.9932,
+      "step": 12465
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3672147025330477e-05,
+      "loss": 0.7007,
+      "step": 12466
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3671182994962989e-05,
+      "loss": 1.1826,
+      "step": 12467
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3670218925160977e-05,
+      "loss": 1.3945,
+      "step": 12468
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36692548159348e-05,
+      "loss": 1.1499,
+      "step": 12469
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3668290667294806e-05,
+      "loss": 1.1123,
+      "step": 12470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3667326479251363e-05,
+      "loss": 1.0518,
+      "step": 12471
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3666362251814817e-05,
+      "loss": 0.772,
+      "step": 12472
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3665397984995534e-05,
+      "loss": 1.4023,
+      "step": 12473
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.366443367880387e-05,
+      "loss": 1.0557,
+      "step": 12474
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3663469333250177e-05,
+      "loss": 1.1943,
+      "step": 12475
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3662504948344824e-05,
+      "loss": 0.9673,
+      "step": 12476
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3661540524098155e-05,
+      "loss": 0.9766,
+      "step": 12477
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.366057606052055e-05,
+      "loss": 1.6074,
+      "step": 12478
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3659611557622352e-05,
+      "loss": 1.4648,
+      "step": 12479
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3658647015413926e-05,
+      "loss": 1.0171,
+      "step": 12480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3657682433905635e-05,
+      "loss": 1.2207,
+      "step": 12481
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365671781310784e-05,
+      "loss": 1.2891,
+      "step": 12482
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36557531530309e-05,
+      "loss": 1.2578,
+      "step": 12483
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3654788453685179e-05,
+      "loss": 0.7983,
+      "step": 12484
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3653823715081041e-05,
+      "loss": 1.2676,
+      "step": 12485
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3652858937228845e-05,
+      "loss": 1.1094,
+      "step": 12486
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365189412013896e-05,
+      "loss": 0.8057,
+      "step": 12487
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365092926382174e-05,
+      "loss": 0.9404,
+      "step": 12488
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.364996436828756e-05,
+      "loss": 1.1016,
+      "step": 12489
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3648999433546775e-05,
+      "loss": 0.958,
+      "step": 12490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.364803445960976e-05,
+      "loss": 1.3584,
+      "step": 12491
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.364706944648687e-05,
+      "loss": 1.2686,
+      "step": 12492
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.364610439418848e-05,
+      "loss": 1.4414,
+      "step": 12493
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.364513930272495e-05,
+      "loss": 1.2754,
+      "step": 12494
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.364417417210665e-05,
+      "loss": 1.0928,
+      "step": 12495
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3643209002343945e-05,
+      "loss": 0.7661,
+      "step": 12496
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3642243793447202e-05,
+      "loss": 0.7334,
+      "step": 12497
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3641278545426793e-05,
+      "loss": 1.1328,
+      "step": 12498
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3640313258293083e-05,
+      "loss": 1.0435,
+      "step": 12499
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3639347932056444e-05,
+      "loss": 1.0352,
+      "step": 12500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3638382566727238e-05,
+      "loss": 1.0059,
+      "step": 12501
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363741716231584e-05,
+      "loss": 1.1191,
+      "step": 12502
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363645171883262e-05,
+      "loss": 1.1055,
+      "step": 12503
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3635486236287949e-05,
+      "loss": 0.8428,
+      "step": 12504
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3634520714692197e-05,
+      "loss": 0.7837,
+      "step": 12505
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3633555154055733e-05,
+      "loss": 0.9321,
+      "step": 12506
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3632589554388931e-05,
+      "loss": 1.0227,
+      "step": 12507
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3631623915702162e-05,
+      "loss": 1.1104,
+      "step": 12508
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3630658238005799e-05,
+      "loss": 1.2246,
+      "step": 12509
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3629692521310216e-05,
+      "loss": 1.4824,
+      "step": 12510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3628726765625785e-05,
+      "loss": 1.4609,
+      "step": 12511
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362776097096288e-05,
+      "loss": 1.0381,
+      "step": 12512
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3626795137331879e-05,
+      "loss": 0.96,
+      "step": 12513
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362582926474315e-05,
+      "loss": 1.1123,
+      "step": 12514
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3624863353207072e-05,
+      "loss": 1.1123,
+      "step": 12515
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3623897402734016e-05,
+      "loss": 0.873,
+      "step": 12516
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3622931413334368e-05,
+      "loss": 1.2715,
+      "step": 12517
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3621965385018496e-05,
+      "loss": 1.2324,
+      "step": 12518
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3620999317796776e-05,
+      "loss": 1.1445,
+      "step": 12519
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362003321167959e-05,
+      "loss": 1.0254,
+      "step": 12520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3619067066677312e-05,
+      "loss": 1.3037,
+      "step": 12521
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3618100882800324e-05,
+      "loss": 0.6938,
+      "step": 12522
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3617134660059001e-05,
+      "loss": 0.895,
+      "step": 12523
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361616839846372e-05,
+      "loss": 1.1279,
+      "step": 12524
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3615202098024867e-05,
+      "loss": 1.2803,
+      "step": 12525
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3614235758752814e-05,
+      "loss": 1.3184,
+      "step": 12526
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3613269380657946e-05,
+      "loss": 0.7202,
+      "step": 12527
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361230296375064e-05,
+      "loss": 1.5908,
+      "step": 12528
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3611336508041281e-05,
+      "loss": 0.8989,
+      "step": 12529
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361037001354025e-05,
+      "loss": 0.9048,
+      "step": 12530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3609403480257925e-05,
+      "loss": 1.2383,
+      "step": 12531
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3608436908204688e-05,
+      "loss": 1.3564,
+      "step": 12532
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3607470297390926e-05,
+      "loss": 0.9619,
+      "step": 12533
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3606503647827016e-05,
+      "loss": 1.0757,
+      "step": 12534
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3605536959523352e-05,
+      "loss": 1.1514,
+      "step": 12535
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3604570232490306e-05,
+      "loss": 1.0859,
+      "step": 12536
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3603603466738269e-05,
+      "loss": 1.1201,
+      "step": 12537
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3602636662277622e-05,
+      "loss": 1.0723,
+      "step": 12538
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3601669819118752e-05,
+      "loss": 0.9951,
+      "step": 12539
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3600702937272045e-05,
+      "loss": 1.2422,
+      "step": 12540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359973601674789e-05,
+      "loss": 1.2646,
+      "step": 12541
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3598769057556664e-05,
+      "loss": 1.2754,
+      "step": 12542
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359780205970876e-05,
+      "loss": 0.8496,
+      "step": 12543
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3596835023214567e-05,
+      "loss": 0.7827,
+      "step": 12544
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3595867948084467e-05,
+      "loss": 0.7471,
+      "step": 12545
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3594900834328855e-05,
+      "loss": 1.0215,
+      "step": 12546
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3593933681958109e-05,
+      "loss": 1.0117,
+      "step": 12547
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3592966490982631e-05,
+      "loss": 1.074,
+      "step": 12548
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3591999261412801e-05,
+      "loss": 1.4746,
+      "step": 12549
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359103199325901e-05,
+      "loss": 1.2988,
+      "step": 12550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3590064686531648e-05,
+      "loss": 1.333,
+      "step": 12551
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3589097341241106e-05,
+      "loss": 0.9756,
+      "step": 12552
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3588129957397778e-05,
+      "loss": 1.0664,
+      "step": 12553
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3587162535012051e-05,
+      "loss": 1.0684,
+      "step": 12554
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3586195074094319e-05,
+      "loss": 1.4688,
+      "step": 12555
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3585227574654974e-05,
+      "loss": 0.7959,
+      "step": 12556
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3584260036704406e-05,
+      "loss": 0.9988,
+      "step": 12557
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358329246025301e-05,
+      "loss": 1.1807,
+      "step": 12558
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3582324845311183e-05,
+      "loss": 1.1284,
+      "step": 12559
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3581357191889311e-05,
+      "loss": 0.8633,
+      "step": 12560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3580389499997795e-05,
+      "loss": 0.8633,
+      "step": 12561
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3579421769647026e-05,
+      "loss": 0.9141,
+      "step": 12562
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3578454000847399e-05,
+      "loss": 0.8535,
+      "step": 12563
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357748619360931e-05,
+      "loss": 1.2754,
+      "step": 12564
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3576518347943154e-05,
+      "loss": 0.7441,
+      "step": 12565
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3575550463859332e-05,
+      "loss": 1.001,
+      "step": 12566
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3574582541368235e-05,
+      "loss": 1.4395,
+      "step": 12567
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3573614580480262e-05,
+      "loss": 0.9824,
+      "step": 12568
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3572646581205811e-05,
+      "loss": 1.167,
+      "step": 12569
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357167854355528e-05,
+      "loss": 1.1064,
+      "step": 12570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3570710467539066e-05,
+      "loss": 1.0791,
+      "step": 12571
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3569742353167568e-05,
+      "loss": 1.4824,
+      "step": 12572
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3568774200451187e-05,
+      "loss": 1.0225,
+      "step": 12573
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356780600940032e-05,
+      "loss": 0.9014,
+      "step": 12574
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356683778002537e-05,
+      "loss": 1.5039,
+      "step": 12575
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3565869512336732e-05,
+      "loss": 1.1289,
+      "step": 12576
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3564901206344814e-05,
+      "loss": 1.1592,
+      "step": 12577
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356393286206001e-05,
+      "loss": 1.2715,
+      "step": 12578
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356296447949273e-05,
+      "loss": 1.2637,
+      "step": 12579
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3561996058653366e-05,
+      "loss": 1.3271,
+      "step": 12580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3561027599552327e-05,
+      "loss": 1.2188,
+      "step": 12581
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3560059102200014e-05,
+      "loss": 1.2324,
+      "step": 12582
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3559090566606832e-05,
+      "loss": 1.0454,
+      "step": 12583
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3558121992783183e-05,
+      "loss": 1.0186,
+      "step": 12584
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3557153380739471e-05,
+      "loss": 1.374,
+      "step": 12585
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35561847304861e-05,
+      "loss": 0.9624,
+      "step": 12586
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3555216042033477e-05,
+      "loss": 0.9648,
+      "step": 12587
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3554247315392008e-05,
+      "loss": 1.1992,
+      "step": 12588
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3553278550572091e-05,
+      "loss": 1.1191,
+      "step": 12589
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3552309747584144e-05,
+      "loss": 1.2607,
+      "step": 12590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3551340906438564e-05,
+      "loss": 0.9185,
+      "step": 12591
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3550372027145761e-05,
+      "loss": 1.124,
+      "step": 12592
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3549403109716146e-05,
+      "loss": 0.8369,
+      "step": 12593
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3548434154160122e-05,
+      "loss": 1.002,
+      "step": 12594
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3547465160488096e-05,
+      "loss": 0.8735,
+      "step": 12595
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354649612871048e-05,
+      "loss": 0.8447,
+      "step": 12596
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3545527058837685e-05,
+      "loss": 1.1123,
+      "step": 12597
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3544557950880115e-05,
+      "loss": 0.8604,
+      "step": 12598
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3543588804848183e-05,
+      "loss": 1.1816,
+      "step": 12599
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35426196207523e-05,
+      "loss": 0.9756,
+      "step": 12600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3541650398602872e-05,
+      "loss": 1.1719,
+      "step": 12601
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3540681138410317e-05,
+      "loss": 1.0322,
+      "step": 12602
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3539711840185042e-05,
+      "loss": 0.9849,
+      "step": 12603
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3538742503937457e-05,
+      "loss": 1.3252,
+      "step": 12604
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353777312967798e-05,
+      "loss": 1.4375,
+      "step": 12605
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3536803717417017e-05,
+      "loss": 1.0986,
+      "step": 12606
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3535834267164986e-05,
+      "loss": 1.2783,
+      "step": 12607
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3534864778932303e-05,
+      "loss": 1.1191,
+      "step": 12608
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353389525272937e-05,
+      "loss": 0.9292,
+      "step": 12609
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3532925688566616e-05,
+      "loss": 1.2363,
+      "step": 12610
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3531956086454447e-05,
+      "loss": 1.042,
+      "step": 12611
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353098644640328e-05,
+      "loss": 0.6316,
+      "step": 12612
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353001676842353e-05,
+      "loss": 0.7393,
+      "step": 12613
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3529047052525615e-05,
+      "loss": 0.8794,
+      "step": 12614
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3528077298719949e-05,
+      "loss": 1.3965,
+      "step": 12615
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352710750701695e-05,
+      "loss": 1.0908,
+      "step": 12616
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3526137677427031e-05,
+      "loss": 0.6143,
+      "step": 12617
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3525167809960618e-05,
+      "loss": 0.8135,
+      "step": 12618
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3524197904628125e-05,
+      "loss": 0.6865,
+      "step": 12619
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3523227961439965e-05,
+      "loss": 0.9014,
+      "step": 12620
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3522257980406565e-05,
+      "loss": 1.0176,
+      "step": 12621
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3521287961538339e-05,
+      "loss": 0.9111,
+      "step": 12622
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352031790484571e-05,
+      "loss": 1.0146,
+      "step": 12623
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3519347810339093e-05,
+      "loss": 1.1494,
+      "step": 12624
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3518377678028915e-05,
+      "loss": 1.5889,
+      "step": 12625
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3517407507925593e-05,
+      "loss": 0.8799,
+      "step": 12626
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3516437300039547e-05,
+      "loss": 1.1699,
+      "step": 12627
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35154670543812e-05,
+      "loss": 1.1621,
+      "step": 12628
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3514496770960976e-05,
+      "loss": 1.1914,
+      "step": 12629
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3513526449789294e-05,
+      "loss": 0.979,
+      "step": 12630
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.351255609087658e-05,
+      "loss": 1.2949,
+      "step": 12631
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3511585694233255e-05,
+      "loss": 1.166,
+      "step": 12632
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3510615259869746e-05,
+      "loss": 0.7729,
+      "step": 12633
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3509644787796475e-05,
+      "loss": 0.9438,
+      "step": 12634
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350867427802386e-05,
+      "loss": 1.0771,
+      "step": 12635
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3507703730562338e-05,
+      "loss": 1.6035,
+      "step": 12636
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3506733145422324e-05,
+      "loss": 1.2346,
+      "step": 12637
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350576252261425e-05,
+      "loss": 1.4551,
+      "step": 12638
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3504791862148538e-05,
+      "loss": 1.042,
+      "step": 12639
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3503821164035618e-05,
+      "loss": 1.002,
+      "step": 12640
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3502850428285914e-05,
+      "loss": 1.2881,
+      "step": 12641
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3501879654909855e-05,
+      "loss": 0.8853,
+      "step": 12642
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350090884391787e-05,
+      "loss": 1.1318,
+      "step": 12643
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3499937995320385e-05,
+      "loss": 0.5415,
+      "step": 12644
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3498967109127827e-05,
+      "loss": 0.6318,
+      "step": 12645
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3497996185350629e-05,
+      "loss": 1.1299,
+      "step": 12646
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3497025223999218e-05,
+      "loss": 1.166,
+      "step": 12647
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3496054225084019e-05,
+      "loss": 1.1914,
+      "step": 12648
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3495083188615471e-05,
+      "loss": 0.7349,
+      "step": 12649
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3494112114604001e-05,
+      "loss": 1.373,
+      "step": 12650
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3493141003060039e-05,
+      "loss": 1.0186,
+      "step": 12651
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3492169853994014e-05,
+      "loss": 1.1064,
+      "step": 12652
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3491198667416363e-05,
+      "loss": 0.9756,
+      "step": 12653
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3490227443337516e-05,
+      "loss": 1.084,
+      "step": 12654
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3489256181767905e-05,
+      "loss": 1.1963,
+      "step": 12655
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3488284882717961e-05,
+      "loss": 0.8115,
+      "step": 12656
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.348731354619812e-05,
+      "loss": 1.5703,
+      "step": 12657
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3486342172218817e-05,
+      "loss": 0.7031,
+      "step": 12658
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3485370760790481e-05,
+      "loss": 0.8213,
+      "step": 12659
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3484399311923555e-05,
+      "loss": 0.9502,
+      "step": 12660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3483427825628463e-05,
+      "loss": 0.8003,
+      "step": 12661
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3482456301915653e-05,
+      "loss": 1.0381,
+      "step": 12662
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3481484740795545e-05,
+      "loss": 1.1777,
+      "step": 12663
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3480513142278589e-05,
+      "loss": 0.751,
+      "step": 12664
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3479541506375218e-05,
+      "loss": 0.5173,
+      "step": 12665
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3478569833095864e-05,
+      "loss": 1.1895,
+      "step": 12666
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3477598122450971e-05,
+      "loss": 0.9263,
+      "step": 12667
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3476626374450974e-05,
+      "loss": 0.9023,
+      "step": 12668
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3475654589106308e-05,
+      "loss": 1.2578,
+      "step": 12669
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3474682766427416e-05,
+      "loss": 1.0205,
+      "step": 12670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3473710906424734e-05,
+      "loss": 1.3623,
+      "step": 12671
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3472739009108707e-05,
+      "loss": 1.0469,
+      "step": 12672
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3471767074489765e-05,
+      "loss": 0.9272,
+      "step": 12673
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3470795102578358e-05,
+      "loss": 1.1406,
+      "step": 12674
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3469823093384919e-05,
+      "loss": 1.3691,
+      "step": 12675
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3468851046919896e-05,
+      "loss": 1.0435,
+      "step": 12676
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3467878963193723e-05,
+      "loss": 1.1035,
+      "step": 12677
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.346690684221685e-05,
+      "loss": 1.084,
+      "step": 12678
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3465934683999709e-05,
+      "loss": 0.7974,
+      "step": 12679
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3464962488552753e-05,
+      "loss": 1.0547,
+      "step": 12680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3463990255886418e-05,
+      "loss": 1.1807,
+      "step": 12681
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3463017986011148e-05,
+      "loss": 0.7563,
+      "step": 12682
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.346204567893739e-05,
+      "loss": 1.1045,
+      "step": 12683
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3461073334675587e-05,
+      "loss": 0.8569,
+      "step": 12684
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3460100953236181e-05,
+      "loss": 0.96,
+      "step": 12685
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345912853462962e-05,
+      "loss": 1.0215,
+      "step": 12686
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3458156078866349e-05,
+      "loss": 1.3389,
+      "step": 12687
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3457183585956812e-05,
+      "loss": 0.9707,
+      "step": 12688
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3456211055911457e-05,
+      "loss": 0.9326,
+      "step": 12689
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345523848874073e-05,
+      "loss": 0.9912,
+      "step": 12690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345426588445508e-05,
+      "loss": 1.0215,
+      "step": 12691
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3453293243064947e-05,
+      "loss": 1.2891,
+      "step": 12692
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345232056458079e-05,
+      "loss": 1.041,
+      "step": 12693
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3451347849013046e-05,
+      "loss": 0.9883,
+      "step": 12694
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3450375096372172e-05,
+      "loss": 1.0576,
+      "step": 12695
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3449402306668613e-05,
+      "loss": 1.293,
+      "step": 12696
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3448429479912818e-05,
+      "loss": 0.8662,
+      "step": 12697
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3447456616115238e-05,
+      "loss": 1.2139,
+      "step": 12698
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3446483715286323e-05,
+      "loss": 0.9414,
+      "step": 12699
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3445510777436522e-05,
+      "loss": 0.9697,
+      "step": 12700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.344453780257629e-05,
+      "loss": 0.9902,
+      "step": 12701
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3443564790716072e-05,
+      "loss": 1.2637,
+      "step": 12702
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3442591741866329e-05,
+      "loss": 1.2959,
+      "step": 12703
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3441618656037505e-05,
+      "loss": 1.3086,
+      "step": 12704
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3440645533240054e-05,
+      "loss": 1.3564,
+      "step": 12705
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.343967237348443e-05,
+      "loss": 0.7988,
+      "step": 12706
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3438699176781087e-05,
+      "loss": 1.2139,
+      "step": 12707
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3437725943140479e-05,
+      "loss": 0.9316,
+      "step": 12708
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.343675267257306e-05,
+      "loss": 0.9287,
+      "step": 12709
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435779365089281e-05,
+      "loss": 1.0322,
+      "step": 12710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3434806020699602e-05,
+      "loss": 1.2549,
+      "step": 12711
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3433832639414476e-05,
+      "loss": 1.5469,
+      "step": 12712
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3432859221244358e-05,
+      "loss": 1.0537,
+      "step": 12713
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3431885766199704e-05,
+      "loss": 0.9141,
+      "step": 12714
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3430912274290973e-05,
+      "loss": 1.4453,
+      "step": 12715
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3429938745528619e-05,
+      "loss": 1.042,
+      "step": 12716
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3428965179923102e-05,
+      "loss": 1.1777,
+      "step": 12717
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3427991577484876e-05,
+      "loss": 0.9092,
+      "step": 12718
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3427017938224403e-05,
+      "loss": 0.7554,
+      "step": 12719
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3426044262152139e-05,
+      "loss": 0.8506,
+      "step": 12720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3425070549278543e-05,
+      "loss": 1.4238,
+      "step": 12721
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342409679961408e-05,
+      "loss": 1.2222,
+      "step": 12722
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3423123013169197e-05,
+      "loss": 0.6226,
+      "step": 12723
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3422149189954368e-05,
+      "loss": 0.9082,
+      "step": 12724
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3421175329980041e-05,
+      "loss": 1.1572,
+      "step": 12725
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342020143325669e-05,
+      "loss": 0.9121,
+      "step": 12726
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3419227499794764e-05,
+      "loss": 1.2051,
+      "step": 12727
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3418253529604733e-05,
+      "loss": 0.6641,
+      "step": 12728
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3417279522697053e-05,
+      "loss": 1.1699,
+      "step": 12729
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3416305479082192e-05,
+      "loss": 1.4531,
+      "step": 12730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3415331398770608e-05,
+      "loss": 1.4521,
+      "step": 12731
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3414357281772768e-05,
+      "loss": 1.0205,
+      "step": 12732
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3413383128099133e-05,
+      "loss": 1.1113,
+      "step": 12733
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3412408937760168e-05,
+      "loss": 1.5449,
+      "step": 12734
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341143471076634e-05,
+      "loss": 1.1992,
+      "step": 12735
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3410460447128109e-05,
+      "loss": 1.1074,
+      "step": 12736
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3409486146855944e-05,
+      "loss": 1.0488,
+      "step": 12737
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3408511809960305e-05,
+      "loss": 0.6924,
+      "step": 12738
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3407537436451666e-05,
+      "loss": 0.9136,
+      "step": 12739
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3406563026340487e-05,
+      "loss": 1.4805,
+      "step": 12740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340558857963724e-05,
+      "loss": 1.166,
+      "step": 12741
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3404614096352387e-05,
+      "loss": 1.1084,
+      "step": 12742
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3403639576496397e-05,
+      "loss": 0.9131,
+      "step": 12743
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3402665020079736e-05,
+      "loss": 1.1494,
+      "step": 12744
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340169042711288e-05,
+      "loss": 0.7056,
+      "step": 12745
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3400715797606293e-05,
+      "loss": 0.9092,
+      "step": 12746
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3399741131570441e-05,
+      "loss": 1.2705,
+      "step": 12747
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33987664290158e-05,
+      "loss": 0.998,
+      "step": 12748
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3397791689952833e-05,
+      "loss": 1.1875,
+      "step": 12749
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3396816914392017e-05,
+      "loss": 0.8105,
+      "step": 12750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3395842102343814e-05,
+      "loss": 0.9146,
+      "step": 12751
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3394867253818706e-05,
+      "loss": 1.0649,
+      "step": 12752
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3393892368827155e-05,
+      "loss": 1.0137,
+      "step": 12753
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3392917447379636e-05,
+      "loss": 1.0684,
+      "step": 12754
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3391942489486626e-05,
+      "loss": 0.916,
+      "step": 12755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.339096749515859e-05,
+      "loss": 0.8071,
+      "step": 12756
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3389992464406007e-05,
+      "loss": 0.9922,
+      "step": 12757
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3389017397239348e-05,
+      "loss": 1.2236,
+      "step": 12758
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3388042293669085e-05,
+      "loss": 0.7437,
+      "step": 12759
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3387067153705698e-05,
+      "loss": 1.0674,
+      "step": 12760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3386091977359655e-05,
+      "loss": 0.833,
+      "step": 12761
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3385116764641433e-05,
+      "loss": 0.9702,
+      "step": 12762
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3384141515561508e-05,
+      "loss": 1.0625,
+      "step": 12763
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3383166230130357e-05,
+      "loss": 1.1416,
+      "step": 12764
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3382190908358454e-05,
+      "loss": 1.0347,
+      "step": 12765
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.338121555025628e-05,
+      "loss": 0.7715,
+      "step": 12766
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3380240155834303e-05,
+      "loss": 1.0,
+      "step": 12767
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3379264725103011e-05,
+      "loss": 0.749,
+      "step": 12768
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3378289258072872e-05,
+      "loss": 0.8391,
+      "step": 12769
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3377313754754373e-05,
+      "loss": 0.9609,
+      "step": 12770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3376338215157986e-05,
+      "loss": 1.1445,
+      "step": 12771
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3375362639294194e-05,
+      "loss": 1.0293,
+      "step": 12772
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3374387027173472e-05,
+      "loss": 1.2637,
+      "step": 12773
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3373411378806302e-05,
+      "loss": 1.3965,
+      "step": 12774
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3372435694203165e-05,
+      "loss": 1.0605,
+      "step": 12775
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.337145997337454e-05,
+      "loss": 1.0791,
+      "step": 12776
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.337048421633091e-05,
+      "loss": 1.0957,
+      "step": 12777
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336950842308275e-05,
+      "loss": 0.8955,
+      "step": 12778
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3368532593640553e-05,
+      "loss": 1.1035,
+      "step": 12779
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336755672801479e-05,
+      "loss": 1.4961,
+      "step": 12780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336658082621595e-05,
+      "loss": 1.0742,
+      "step": 12781
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3365604888254511e-05,
+      "loss": 1.1777,
+      "step": 12782
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3364628914140963e-05,
+      "loss": 0.9893,
+      "step": 12783
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336365290388578e-05,
+      "loss": 1.0322,
+      "step": 12784
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3362676857499453e-05,
+      "loss": 0.9121,
+      "step": 12785
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3361700774992466e-05,
+      "loss": 0.7515,
+      "step": 12786
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3360724656375302e-05,
+      "loss": 0.9756,
+      "step": 12787
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3359748501658443e-05,
+      "loss": 1.0352,
+      "step": 12788
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335877231085238e-05,
+      "loss": 1.2949,
+      "step": 12789
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3357796083967598e-05,
+      "loss": 0.8022,
+      "step": 12790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335681982101458e-05,
+      "loss": 1.2402,
+      "step": 12791
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3355843522003818e-05,
+      "loss": 0.9082,
+      "step": 12792
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335486718694579e-05,
+      "loss": 0.8838,
+      "step": 12793
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3353890815850993e-05,
+      "loss": 1.1729,
+      "step": 12794
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335291440872991e-05,
+      "loss": 1.1699,
+      "step": 12795
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3351937965593031e-05,
+      "loss": 0.9902,
+      "step": 12796
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3350961486450843e-05,
+      "loss": 1.085,
+      "step": 12797
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3349984971313837e-05,
+      "loss": 1.0049,
+      "step": 12798
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33490084201925e-05,
+      "loss": 1.2339,
+      "step": 12799
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3348031833097319e-05,
+      "loss": 1.3652,
+      "step": 12800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3347055210038795e-05,
+      "loss": 1.1758,
+      "step": 12801
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3346078551027408e-05,
+      "loss": 1.0117,
+      "step": 12802
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3345101856073655e-05,
+      "loss": 1.124,
+      "step": 12803
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334412512518802e-05,
+      "loss": 1.1382,
+      "step": 12804
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3343148358381006e-05,
+      "loss": 1.6074,
+      "step": 12805
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3342171555663093e-05,
+      "loss": 1.0303,
+      "step": 12806
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3341194717044781e-05,
+      "loss": 1.085,
+      "step": 12807
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334021784253656e-05,
+      "loss": 1.3223,
+      "step": 12808
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3339240932148925e-05,
+      "loss": 1.0508,
+      "step": 12809
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333826398589237e-05,
+      "loss": 0.748,
+      "step": 12810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3337287003777384e-05,
+      "loss": 1.1709,
+      "step": 12811
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333630998581447e-05,
+      "loss": 0.8369,
+      "step": 12812
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3335332932014113e-05,
+      "loss": 0.9521,
+      "step": 12813
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3334355842386817e-05,
+      "loss": 1.0566,
+      "step": 12814
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3333378716943072e-05,
+      "loss": 1.4131,
+      "step": 12815
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3332401555693376e-05,
+      "loss": 0.7002,
+      "step": 12816
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3331424358648225e-05,
+      "loss": 1.1875,
+      "step": 12817
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3330447125818115e-05,
+      "loss": 1.2588,
+      "step": 12818
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3329469857213543e-05,
+      "loss": 1.0479,
+      "step": 12819
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3328492552845007e-05,
+      "loss": 0.8691,
+      "step": 12820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3327515212723006e-05,
+      "loss": 1.2754,
+      "step": 12821
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3326537836858038e-05,
+      "loss": 1.1191,
+      "step": 12822
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3325560425260601e-05,
+      "loss": 1.2817,
+      "step": 12823
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3324582977941192e-05,
+      "loss": 1.0586,
+      "step": 12824
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3323605494910315e-05,
+      "loss": 0.9062,
+      "step": 12825
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3322627976178462e-05,
+      "loss": 1.0059,
+      "step": 12826
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.332165042175614e-05,
+      "loss": 0.9917,
+      "step": 12827
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3320672831653849e-05,
+      "loss": 0.9224,
+      "step": 12828
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3319695205882088e-05,
+      "loss": 1.3389,
+      "step": 12829
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3318717544451359e-05,
+      "loss": 1.2959,
+      "step": 12830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3317739847372165e-05,
+      "loss": 1.1162,
+      "step": 12831
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3316762114655004e-05,
+      "loss": 1.0381,
+      "step": 12832
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3315784346310381e-05,
+      "loss": 1.3994,
+      "step": 12833
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33148065423488e-05,
+      "loss": 0.9541,
+      "step": 12834
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3313828702780764e-05,
+      "loss": 0.8711,
+      "step": 12835
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3312850827616776e-05,
+      "loss": 1.2642,
+      "step": 12836
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3311872916867336e-05,
+      "loss": 0.8789,
+      "step": 12837
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3310894970542958e-05,
+      "loss": 1.21,
+      "step": 12838
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3309916988654135e-05,
+      "loss": 0.9805,
+      "step": 12839
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3308938971211382e-05,
+      "loss": 1.374,
+      "step": 12840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3307960918225198e-05,
+      "loss": 0.9443,
+      "step": 12841
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3306982829706093e-05,
+      "loss": 1.0938,
+      "step": 12842
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330600470566457e-05,
+      "loss": 0.9746,
+      "step": 12843
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3305026546111137e-05,
+      "loss": 1.6172,
+      "step": 12844
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3304048351056305e-05,
+      "loss": 1.207,
+      "step": 12845
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3303070120510575e-05,
+      "loss": 1.103,
+      "step": 12846
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3302091854484458e-05,
+      "loss": 1.418,
+      "step": 12847
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330111355298846e-05,
+      "loss": 0.8511,
+      "step": 12848
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3300135216033096e-05,
+      "loss": 1.0947,
+      "step": 12849
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3299156843628865e-05,
+      "loss": 1.3477,
+      "step": 12850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3298178435786287e-05,
+      "loss": 1.0713,
+      "step": 12851
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.329719999251586e-05,
+      "loss": 1.1904,
+      "step": 12852
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3296221513828103e-05,
+      "loss": 1.1807,
+      "step": 12853
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3295242999733524e-05,
+      "loss": 1.1553,
+      "step": 12854
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.329426445024263e-05,
+      "loss": 1.3359,
+      "step": 12855
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3293285865365943e-05,
+      "loss": 0.9736,
+      "step": 12856
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3292307245113963e-05,
+      "loss": 1.0693,
+      "step": 12857
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.329132858949721e-05,
+      "loss": 1.146,
+      "step": 12858
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3290349898526189e-05,
+      "loss": 1.2461,
+      "step": 12859
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3289371172211418e-05,
+      "loss": 1.2256,
+      "step": 12860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3288392410563408e-05,
+      "loss": 0.9219,
+      "step": 12861
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3287413613592674e-05,
+      "loss": 0.8271,
+      "step": 12862
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.328643478130973e-05,
+      "loss": 1.1064,
+      "step": 12863
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3285455913725088e-05,
+      "loss": 1.5664,
+      "step": 12864
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3284477010849264e-05,
+      "loss": 1.1465,
+      "step": 12865
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3283498072692771e-05,
+      "loss": 0.8037,
+      "step": 12866
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3282519099266131e-05,
+      "loss": 0.9531,
+      "step": 12867
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3281540090579853e-05,
+      "loss": 0.8867,
+      "step": 12868
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3280561046644456e-05,
+      "loss": 0.9326,
+      "step": 12869
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3279581967470452e-05,
+      "loss": 0.7939,
+      "step": 12870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3278602853068368e-05,
+      "loss": 1.0488,
+      "step": 12871
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3277623703448712e-05,
+      "loss": 0.9785,
+      "step": 12872
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3276644518622003e-05,
+      "loss": 1.1465,
+      "step": 12873
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3275665298598763e-05,
+      "loss": 1.3252,
+      "step": 12874
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3274686043389509e-05,
+      "loss": 1.2988,
+      "step": 12875
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3273706753004758e-05,
+      "loss": 1.1934,
+      "step": 12876
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3272727427455027e-05,
+      "loss": 1.0098,
+      "step": 12877
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3271748066750845e-05,
+      "loss": 0.7305,
+      "step": 12878
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3270768670902721e-05,
+      "loss": 1.0459,
+      "step": 12879
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3269789239921184e-05,
+      "loss": 1.4883,
+      "step": 12880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3268809773816746e-05,
+      "loss": 0.9619,
+      "step": 12881
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3267830272599938e-05,
+      "loss": 1.0645,
+      "step": 12882
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3266850736281272e-05,
+      "loss": 0.6519,
+      "step": 12883
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3265871164871275e-05,
+      "loss": 1.2891,
+      "step": 12884
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3264891558380471e-05,
+      "loss": 1.1982,
+      "step": 12885
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3263911916819377e-05,
+      "loss": 1.1738,
+      "step": 12886
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326293224019852e-05,
+      "loss": 0.814,
+      "step": 12887
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3261952528528421e-05,
+      "loss": 1.1025,
+      "step": 12888
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3260972781819605e-05,
+      "loss": 1.0146,
+      "step": 12889
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3259993000082599e-05,
+      "loss": 0.9297,
+      "step": 12890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325901318332792e-05,
+      "loss": 0.7715,
+      "step": 12891
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.32580333315661e-05,
+      "loss": 1.415,
+      "step": 12892
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3257053444807662e-05,
+      "loss": 0.9131,
+      "step": 12893
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325607352306313e-05,
+      "loss": 1.0864,
+      "step": 12894
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325509356634303e-05,
+      "loss": 1.2744,
+      "step": 12895
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325411357465789e-05,
+      "loss": 1.3506,
+      "step": 12896
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3253133548018237e-05,
+      "loss": 0.5918,
+      "step": 12897
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3252153486434597e-05,
+      "loss": 0.9717,
+      "step": 12898
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3251173389917496e-05,
+      "loss": 0.9053,
+      "step": 12899
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3250193258477465e-05,
+      "loss": 1.0371,
+      "step": 12900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.324921309212503e-05,
+      "loss": 0.8613,
+      "step": 12901
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3248232890870723e-05,
+      "loss": 1.1133,
+      "step": 12902
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3247252654725067e-05,
+      "loss": 1.1172,
+      "step": 12903
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3246272383698596e-05,
+      "loss": 0.959,
+      "step": 12904
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.324529207780184e-05,
+      "loss": 1.165,
+      "step": 12905
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3244311737045324e-05,
+      "loss": 1.0293,
+      "step": 12906
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3243331361439583e-05,
+      "loss": 0.7109,
+      "step": 12907
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3242350950995146e-05,
+      "loss": 0.9473,
+      "step": 12908
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3241370505722545e-05,
+      "loss": 0.9053,
+      "step": 12909
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3240390025632315e-05,
+      "loss": 1.5586,
+      "step": 12910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3239409510734982e-05,
+      "loss": 1.1807,
+      "step": 12911
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3238428961041078e-05,
+      "loss": 0.8418,
+      "step": 12912
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3237448376561146e-05,
+      "loss": 0.9814,
+      "step": 12913
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3236467757305704e-05,
+      "loss": 1.0811,
+      "step": 12914
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3235487103285299e-05,
+      "loss": 1.1279,
+      "step": 12915
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3234506414510455e-05,
+      "loss": 1.1055,
+      "step": 12916
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.323352569099171e-05,
+      "loss": 1.04,
+      "step": 12917
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3232544932739601e-05,
+      "loss": 0.998,
+      "step": 12918
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3231564139764662e-05,
+      "loss": 1.0938,
+      "step": 12919
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3230583312077423e-05,
+      "loss": 0.7988,
+      "step": 12920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3229602449688423e-05,
+      "loss": 0.9941,
+      "step": 12921
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3228621552608203e-05,
+      "loss": 1.0967,
+      "step": 12922
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3227640620847293e-05,
+      "loss": 1.1367,
+      "step": 12923
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3226659654416235e-05,
+      "loss": 1.4277,
+      "step": 12924
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3225678653325558e-05,
+      "loss": 1.3574,
+      "step": 12925
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3224697617585809e-05,
+      "loss": 1.1094,
+      "step": 12926
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3223716547207516e-05,
+      "loss": 1.1348,
+      "step": 12927
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3222735442201231e-05,
+      "loss": 1.2266,
+      "step": 12928
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3221754302577479e-05,
+      "loss": 1.1611,
+      "step": 12929
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3220773128346806e-05,
+      "loss": 0.9707,
+      "step": 12930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321979191951975e-05,
+      "loss": 1.0244,
+      "step": 12931
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3218810676106852e-05,
+      "loss": 1.0879,
+      "step": 12932
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3217829398118652e-05,
+      "loss": 1.3555,
+      "step": 12933
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3216848085565688e-05,
+      "loss": 0.9883,
+      "step": 12934
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3215866738458503e-05,
+      "loss": 1.332,
+      "step": 12935
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3214885356807637e-05,
+      "loss": 1.1997,
+      "step": 12936
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321390394062363e-05,
+      "loss": 1.4131,
+      "step": 12937
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321292248991703e-05,
+      "loss": 0.8369,
+      "step": 12938
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3211941004698378e-05,
+      "loss": 1.1973,
+      "step": 12939
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321095948497821e-05,
+      "loss": 1.0264,
+      "step": 12940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3209977930767078e-05,
+      "loss": 0.9521,
+      "step": 12941
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3208996342075517e-05,
+      "loss": 1.3945,
+      "step": 12942
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3208014718914075e-05,
+      "loss": 1.0742,
+      "step": 12943
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3207033061293299e-05,
+      "loss": 1.1924,
+      "step": 12944
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3206051369223732e-05,
+      "loss": 1.1924,
+      "step": 12945
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3205069642715915e-05,
+      "loss": 1.0537,
+      "step": 12946
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.32040878817804e-05,
+      "loss": 0.9131,
+      "step": 12947
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3203106086427725e-05,
+      "loss": 0.79,
+      "step": 12948
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3202124256668443e-05,
+      "loss": 1.2295,
+      "step": 12949
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3201142392513095e-05,
+      "loss": 1.3008,
+      "step": 12950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3200160493972232e-05,
+      "loss": 1.123,
+      "step": 12951
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3199178561056398e-05,
+      "loss": 1.334,
+      "step": 12952
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3198196593776145e-05,
+      "loss": 0.9209,
+      "step": 12953
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3197214592142015e-05,
+      "loss": 1.125,
+      "step": 12954
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3196232556164566e-05,
+      "loss": 1.1152,
+      "step": 12955
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3195250485854331e-05,
+      "loss": 0.8647,
+      "step": 12956
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3194268381221875e-05,
+      "loss": 0.8877,
+      "step": 12957
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3193286242277736e-05,
+      "loss": 0.9106,
+      "step": 12958
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.319230406903247e-05,
+      "loss": 1.0264,
+      "step": 12959
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3191321861496627e-05,
+      "loss": 1.0449,
+      "step": 12960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3190339619680754e-05,
+      "loss": 1.0156,
+      "step": 12961
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3189357343595405e-05,
+      "loss": 1.1758,
+      "step": 12962
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3188375033251131e-05,
+      "loss": 1.1025,
+      "step": 12963
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3187392688658482e-05,
+      "loss": 1.0288,
+      "step": 12964
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3186410309828008e-05,
+      "loss": 1.2822,
+      "step": 12965
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318542789677027e-05,
+      "loss": 1.0225,
+      "step": 12966
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3184445449495808e-05,
+      "loss": 1.0518,
+      "step": 12967
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3183462968015186e-05,
+      "loss": 0.9639,
+      "step": 12968
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3182480452338953e-05,
+      "loss": 1.4648,
+      "step": 12969
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3181497902477665e-05,
+      "loss": 0.9023,
+      "step": 12970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3180515318441872e-05,
+      "loss": 1.1816,
+      "step": 12971
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3179532700242133e-05,
+      "loss": 0.8584,
+      "step": 12972
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3178550047888998e-05,
+      "loss": 0.6885,
+      "step": 12973
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3177567361393026e-05,
+      "loss": 0.9961,
+      "step": 12974
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3176584640764773e-05,
+      "loss": 0.8066,
+      "step": 12975
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3175601886014793e-05,
+      "loss": 1.1836,
+      "step": 12976
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3174619097153645e-05,
+      "loss": 1.0205,
+      "step": 12977
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3173636274191884e-05,
+      "loss": 0.9023,
+      "step": 12978
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3172653417140064e-05,
+      "loss": 1.1016,
+      "step": 12979
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3171670526008749e-05,
+      "loss": 1.0332,
+      "step": 12980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3170687600808491e-05,
+      "loss": 1.1221,
+      "step": 12981
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3169704641549852e-05,
+      "loss": 1.1523,
+      "step": 12982
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3168721648243392e-05,
+      "loss": 0.8496,
+      "step": 12983
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3167738620899663e-05,
+      "loss": 0.9746,
+      "step": 12984
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3166755559529234e-05,
+      "loss": 0.9893,
+      "step": 12985
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3165772464142654e-05,
+      "loss": 1.1484,
+      "step": 12986
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3164789334750492e-05,
+      "loss": 1.3613,
+      "step": 12987
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3163806171363302e-05,
+      "loss": 1.1465,
+      "step": 12988
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3162822973991647e-05,
+      "loss": 1.1895,
+      "step": 12989
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3161839742646092e-05,
+      "loss": 0.9775,
+      "step": 12990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3160856477337194e-05,
+      "loss": 0.9775,
+      "step": 12991
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3159873178075518e-05,
+      "loss": 1.3164,
+      "step": 12992
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3158889844871621e-05,
+      "loss": 1.541,
+      "step": 12993
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315790647773607e-05,
+      "loss": 1.0186,
+      "step": 12994
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3156923076679428e-05,
+      "loss": 1.127,
+      "step": 12995
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3155939641712258e-05,
+      "loss": 0.9414,
+      "step": 12996
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3154956172845121e-05,
+      "loss": 1.1074,
+      "step": 12997
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3153972670088584e-05,
+      "loss": 0.9971,
+      "step": 12998
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3152989133453211e-05,
+      "loss": 1.2285,
+      "step": 12999
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3152005562949566e-05,
+      "loss": 1.0352,
+      "step": 13000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3151021958588217e-05,
+      "loss": 1.3193,
+      "step": 13001
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3150038320379723e-05,
+      "loss": 1.1318,
+      "step": 13002
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3149054648334656e-05,
+      "loss": 1.2354,
+      "step": 13003
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314807094246358e-05,
+      "loss": 0.9365,
+      "step": 13004
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3147087202777063e-05,
+      "loss": 1.5293,
+      "step": 13005
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3146103429285668e-05,
+      "loss": 1.0576,
+      "step": 13006
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3145119621999967e-05,
+      "loss": 1.0273,
+      "step": 13007
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3144135780930524e-05,
+      "loss": 1.3613,
+      "step": 13008
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3143151906087912e-05,
+      "loss": 1.0723,
+      "step": 13009
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3142167997482693e-05,
+      "loss": 1.1914,
+      "step": 13010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3141184055125443e-05,
+      "loss": 0.8496,
+      "step": 13011
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3140200079026722e-05,
+      "loss": 1.1665,
+      "step": 13012
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3139216069197106e-05,
+      "loss": 0.9746,
+      "step": 13013
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3138232025647168e-05,
+      "loss": 1.0991,
+      "step": 13014
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313724794838747e-05,
+      "loss": 1.1797,
+      "step": 13015
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3136263837428588e-05,
+      "loss": 1.1211,
+      "step": 13016
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3135279692781094e-05,
+      "loss": 1.0947,
+      "step": 13017
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3134295514455552e-05,
+      "loss": 1.2432,
+      "step": 13018
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3133311302462542e-05,
+      "loss": 1.1465,
+      "step": 13019
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3132327056812628e-05,
+      "loss": 1.2236,
+      "step": 13020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313134277751639e-05,
+      "loss": 0.9521,
+      "step": 13021
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3130358464584399e-05,
+      "loss": 1.1602,
+      "step": 13022
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3129374118027225e-05,
+      "loss": 1.1826,
+      "step": 13023
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3128389737855443e-05,
+      "loss": 1.0107,
+      "step": 13024
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3127405324079628e-05,
+      "loss": 0.8877,
+      "step": 13025
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3126420876710352e-05,
+      "loss": 1.2295,
+      "step": 13026
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3125436395758192e-05,
+      "loss": 1.2168,
+      "step": 13027
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3124451881233722e-05,
+      "loss": 1.3086,
+      "step": 13028
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312346733314752e-05,
+      "loss": 0.9116,
+      "step": 13029
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3122482751510153e-05,
+      "loss": 1.0581,
+      "step": 13030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3121498136332207e-05,
+      "loss": 1.416,
+      "step": 13031
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3120513487624253e-05,
+      "loss": 0.7456,
+      "step": 13032
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311952880539687e-05,
+      "loss": 1.1104,
+      "step": 13033
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3118544089660635e-05,
+      "loss": 0.8853,
+      "step": 13034
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311755934042612e-05,
+      "loss": 1.3252,
+      "step": 13035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3116574557703912e-05,
+      "loss": 0.7209,
+      "step": 13036
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3115589741504584e-05,
+      "loss": 0.999,
+      "step": 13037
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311460489183871e-05,
+      "loss": 0.9873,
+      "step": 13038
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3113620008716881e-05,
+      "loss": 1.166,
+      "step": 13039
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3112635092149668e-05,
+      "loss": 1.1758,
+      "step": 13040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311165014214765e-05,
+      "loss": 1.1836,
+      "step": 13041
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3110665158721407e-05,
+      "loss": 0.6924,
+      "step": 13042
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3109680141881525e-05,
+      "loss": 1.0654,
+      "step": 13043
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3108695091638577e-05,
+      "loss": 0.835,
+      "step": 13044
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3107710008003154e-05,
+      "loss": 1.123,
+      "step": 13045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3106724890985824e-05,
+      "loss": 0.7969,
+      "step": 13046
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3105739740597183e-05,
+      "loss": 1.1748,
+      "step": 13047
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3104754556847802e-05,
+      "loss": 1.3691,
+      "step": 13048
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3103769339748271e-05,
+      "loss": 1.2236,
+      "step": 13049
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3102784089309166e-05,
+      "loss": 1.126,
+      "step": 13050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3101798805541076e-05,
+      "loss": 0.8379,
+      "step": 13051
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3100813488454582e-05,
+      "loss": 1.0781,
+      "step": 13052
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3099828138060266e-05,
+      "loss": 1.2031,
+      "step": 13053
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3098842754368719e-05,
+      "loss": 0.8525,
+      "step": 13054
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3097857337390519e-05,
+      "loss": 0.8232,
+      "step": 13055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3096871887136251e-05,
+      "loss": 0.9785,
+      "step": 13056
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3095886403616505e-05,
+      "loss": 1.2393,
+      "step": 13057
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3094900886841867e-05,
+      "loss": 1.1426,
+      "step": 13058
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3093915336822915e-05,
+      "loss": 1.3613,
+      "step": 13059
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3092929753570245e-05,
+      "loss": 0.8271,
+      "step": 13060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3091944137094435e-05,
+      "loss": 1.1582,
+      "step": 13061
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.309095848740608e-05,
+      "loss": 0.8096,
+      "step": 13062
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3089972804515764e-05,
+      "loss": 1.4199,
+      "step": 13063
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3088987088434076e-05,
+      "loss": 1.0156,
+      "step": 13064
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30880013391716e-05,
+      "loss": 1.0742,
+      "step": 13065
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3087015556738928e-05,
+      "loss": 1.1074,
+      "step": 13066
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308602974114665e-05,
+      "loss": 1.1992,
+      "step": 13067
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3085043892405357e-05,
+      "loss": 1.1309,
+      "step": 13068
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308405801052563e-05,
+      "loss": 1.0586,
+      "step": 13069
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3083072095518068e-05,
+      "loss": 1.0801,
+      "step": 13070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308208614739326e-05,
+      "loss": 1.0029,
+      "step": 13071
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3081100166161788e-05,
+      "loss": 1.1602,
+      "step": 13072
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3080114151834258e-05,
+      "loss": 1.6475,
+      "step": 13073
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3079128104421246e-05,
+      "loss": 1.5352,
+      "step": 13074
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3078142023933356e-05,
+      "loss": 0.8252,
+      "step": 13075
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3077155910381172e-05,
+      "loss": 1.6094,
+      "step": 13076
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3076169763775291e-05,
+      "loss": 1.3457,
+      "step": 13077
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3075183584126302e-05,
+      "loss": 1.125,
+      "step": 13078
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30741973714448e-05,
+      "loss": 0.8628,
+      "step": 13079
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3073211125741383e-05,
+      "loss": 1.0088,
+      "step": 13080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.307222484702664e-05,
+      "loss": 1.207,
+      "step": 13081
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3071238535311164e-05,
+      "loss": 1.3066,
+      "step": 13082
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3070252190605552e-05,
+      "loss": 0.9668,
+      "step": 13083
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3069265812920403e-05,
+      "loss": 1.2334,
+      "step": 13084
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3068279402266302e-05,
+      "loss": 1.168,
+      "step": 13085
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3067292958653854e-05,
+      "loss": 0.8906,
+      "step": 13086
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306630648209365e-05,
+      "loss": 1.1113,
+      "step": 13087
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3065319972596288e-05,
+      "loss": 1.1084,
+      "step": 13088
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3064333430172369e-05,
+      "loss": 0.6167,
+      "step": 13089
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306334685483248e-05,
+      "loss": 1.1016,
+      "step": 13090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3062360246587232e-05,
+      "loss": 1.0361,
+      "step": 13091
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306137360544721e-05,
+      "loss": 1.2344,
+      "step": 13092
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3060386931423017e-05,
+      "loss": 1.0029,
+      "step": 13093
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3059400224525253e-05,
+      "loss": 1.2852,
+      "step": 13094
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3058413484764514e-05,
+      "loss": 1.085,
+      "step": 13095
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3057426712151401e-05,
+      "loss": 1.46,
+      "step": 13096
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3056439906696516e-05,
+      "loss": 1.1826,
+      "step": 13097
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3055453068410455e-05,
+      "loss": 1.3174,
+      "step": 13098
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3054466197303818e-05,
+      "loss": 1.4219,
+      "step": 13099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305347929338721e-05,
+      "loss": 1.0215,
+      "step": 13100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3052492356671228e-05,
+      "loss": 0.7764,
+      "step": 13101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3051505387166477e-05,
+      "loss": 0.7019,
+      "step": 13102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305051838488355e-05,
+      "loss": 1.3564,
+      "step": 13103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.304953134983306e-05,
+      "loss": 1.1191,
+      "step": 13104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30485442820256e-05,
+      "loss": 1.3633,
+      "step": 13105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3047557181471784e-05,
+      "loss": 1.1299,
+      "step": 13106
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3046570048182203e-05,
+      "loss": 1.1084,
+      "step": 13107
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3045582882167469e-05,
+      "loss": 1.1035,
+      "step": 13108
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.304459568343818e-05,
+      "loss": 0.8965,
+      "step": 13109
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3043608452004943e-05,
+      "loss": 1.0518,
+      "step": 13110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3042621187878364e-05,
+      "loss": 1.1182,
+      "step": 13111
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3041633891069046e-05,
+      "loss": 1.5352,
+      "step": 13112
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3040646561587592e-05,
+      "loss": 1.4131,
+      "step": 13113
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303965919944461e-05,
+      "loss": 1.0117,
+      "step": 13114
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3038671804650708e-05,
+      "loss": 0.7598,
+      "step": 13115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3037684377216486e-05,
+      "loss": 1.0801,
+      "step": 13116
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303669691715256e-05,
+      "loss": 1.2051,
+      "step": 13117
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3035709424469525e-05,
+      "loss": 0.9756,
+      "step": 13118
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3034721899178e-05,
+      "loss": 1.2744,
+      "step": 13119
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3033734341288583e-05,
+      "loss": 1.4453,
+      "step": 13120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3032746750811889e-05,
+      "loss": 0.9795,
+      "step": 13121
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3031759127758524e-05,
+      "loss": 1.082,
+      "step": 13122
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3030771472139093e-05,
+      "loss": 1.2207,
+      "step": 13123
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3029783783964214e-05,
+      "loss": 1.0186,
+      "step": 13124
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3028796063244486e-05,
+      "loss": 1.1016,
+      "step": 13125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3027808309990524e-05,
+      "loss": 1.207,
+      "step": 13126
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.302682052421294e-05,
+      "loss": 0.8018,
+      "step": 13127
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3025832705922341e-05,
+      "loss": 0.7666,
+      "step": 13128
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3024844855129337e-05,
+      "loss": 1.1172,
+      "step": 13129
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3023856971844543e-05,
+      "loss": 0.7627,
+      "step": 13130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3022869056078567e-05,
+      "loss": 1.123,
+      "step": 13131
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3021881107842025e-05,
+      "loss": 0.9043,
+      "step": 13132
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3020893127145524e-05,
+      "loss": 1.2109,
+      "step": 13133
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3019905113999678e-05,
+      "loss": 1.0234,
+      "step": 13134
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3018917068415105e-05,
+      "loss": 1.0215,
+      "step": 13135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3017928990402409e-05,
+      "loss": 1.0791,
+      "step": 13136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3016940879972214e-05,
+      "loss": 1.2627,
+      "step": 13137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3015952737135124e-05,
+      "loss": 1.3691,
+      "step": 13138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301496456190176e-05,
+      "loss": 0.9531,
+      "step": 13139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3013976354282733e-05,
+      "loss": 0.8813,
+      "step": 13140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3012988114288663e-05,
+      "loss": 1.0488,
+      "step": 13141
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3011999841930159e-05,
+      "loss": 1.1787,
+      "step": 13142
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301101153721784e-05,
+      "loss": 1.2305,
+      "step": 13143
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301002320016232e-05,
+      "loss": 1.4551,
+      "step": 13144
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3009034830774216e-05,
+      "loss": 0.8706,
+      "step": 13145
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3008046429064152e-05,
+      "loss": 1.377,
+      "step": 13146
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.300705799504273e-05,
+      "loss": 1.2646,
+      "step": 13147
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3006069528720584e-05,
+      "loss": 1.0342,
+      "step": 13148
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3005081030108318e-05,
+      "loss": 1.1777,
+      "step": 13149
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3004092499216561e-05,
+      "loss": 1.0586,
+      "step": 13150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3003103936055921e-05,
+      "loss": 1.0029,
+      "step": 13151
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3002115340637025e-05,
+      "loss": 1.0771,
+      "step": 13152
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3001126712970487e-05,
+      "loss": 0.8018,
+      "step": 13153
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3000138053066931e-05,
+      "loss": 0.9805,
+      "step": 13154
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2999149360936973e-05,
+      "loss": 0.7397,
+      "step": 13155
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2998160636591234e-05,
+      "loss": 1.1709,
+      "step": 13156
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2997171880040338e-05,
+      "loss": 0.7041,
+      "step": 13157
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2996183091294901e-05,
+      "loss": 0.9004,
+      "step": 13158
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299519427036555e-05,
+      "loss": 0.9434,
+      "step": 13159
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2994205417262898e-05,
+      "loss": 0.8623,
+      "step": 13160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2993216531997574e-05,
+      "loss": 1.0728,
+      "step": 13161
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2992227614580196e-05,
+      "loss": 0.9404,
+      "step": 13162
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299123866502139e-05,
+      "loss": 1.0078,
+      "step": 13163
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2990249683331779e-05,
+      "loss": 0.833,
+      "step": 13164
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2989260669521984e-05,
+      "loss": 0.9873,
+      "step": 13165
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2988271623602628e-05,
+      "loss": 1.0938,
+      "step": 13166
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2987282545584337e-05,
+      "loss": 1.0547,
+      "step": 13167
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2986293435477735e-05,
+      "loss": 0.9629,
+      "step": 13168
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2985304293293445e-05,
+      "loss": 1.0889,
+      "step": 13169
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2984315119042095e-05,
+      "loss": 0.9404,
+      "step": 13170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2983325912734308e-05,
+      "loss": 0.9258,
+      "step": 13171
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2982336674380713e-05,
+      "loss": 1.0361,
+      "step": 13172
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2981347403991928e-05,
+      "loss": 0.835,
+      "step": 13173
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2980358101578592e-05,
+      "loss": 1.4795,
+      "step": 13174
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2979368767151317e-05,
+      "loss": 1.1475,
+      "step": 13175
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2978379400720744e-05,
+      "loss": 1.0205,
+      "step": 13176
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2977390002297492e-05,
+      "loss": 1.416,
+      "step": 13177
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2976400571892189e-05,
+      "loss": 1.5508,
+      "step": 13178
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2975411109515468e-05,
+      "loss": 0.7495,
+      "step": 13179
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297442161517795e-05,
+      "loss": 0.9893,
+      "step": 13180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2973432088890275e-05,
+      "loss": 0.8555,
+      "step": 13181
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2972442530663061e-05,
+      "loss": 0.9238,
+      "step": 13182
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2971452940506941e-05,
+      "loss": 0.916,
+      "step": 13183
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2970463318432547e-05,
+      "loss": 0.9414,
+      "step": 13184
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2969473664450509e-05,
+      "loss": 0.7388,
+      "step": 13185
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2968483978571453e-05,
+      "loss": 0.9521,
+      "step": 13186
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2967494260806018e-05,
+      "loss": 0.9111,
+      "step": 13187
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2966504511164825e-05,
+      "loss": 0.8691,
+      "step": 13188
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2965514729658514e-05,
+      "loss": 1.3477,
+      "step": 13189
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2964524916297716e-05,
+      "loss": 0.9463,
+      "step": 13190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2963535071093054e-05,
+      "loss": 0.8047,
+      "step": 13191
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2962545194055173e-05,
+      "loss": 1.2227,
+      "step": 13192
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2961555285194697e-05,
+      "loss": 1.2314,
+      "step": 13193
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2960565344522268e-05,
+      "loss": 1.3369,
+      "step": 13194
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2959575372048513e-05,
+      "loss": 0.9292,
+      "step": 13195
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2958585367784063e-05,
+      "loss": 0.854,
+      "step": 13196
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2957595331739558e-05,
+      "loss": 0.6787,
+      "step": 13197
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956605263925632e-05,
+      "loss": 0.9844,
+      "step": 13198
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.295561516435292e-05,
+      "loss": 1.0674,
+      "step": 13199
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2954625033032055e-05,
+      "loss": 1.1211,
+      "step": 13200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2953634869973673e-05,
+      "loss": 0.9512,
+      "step": 13201
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2952644675188413e-05,
+      "loss": 1.0713,
+      "step": 13202
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.295165444868691e-05,
+      "loss": 0.8682,
+      "step": 13203
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2950664190479796e-05,
+      "loss": 0.9971,
+      "step": 13204
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2949673900577715e-05,
+      "loss": 1.3184,
+      "step": 13205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29486835789913e-05,
+      "loss": 1.0283,
+      "step": 13206
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2947693225731191e-05,
+      "loss": 1.0684,
+      "step": 13207
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2946702840808024e-05,
+      "loss": 1.123,
+      "step": 13208
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2945712424232437e-05,
+      "loss": 1.2109,
+      "step": 13209
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2944721976015071e-05,
+      "loss": 1.1533,
+      "step": 13210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2943731496166565e-05,
+      "loss": 0.9258,
+      "step": 13211
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2942740984697556e-05,
+      "loss": 1.2275,
+      "step": 13212
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2941750441618686e-05,
+      "loss": 1.2363,
+      "step": 13213
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2940759866940593e-05,
+      "loss": 1.1621,
+      "step": 13214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293976926067392e-05,
+      "loss": 1.2559,
+      "step": 13215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2938778622829308e-05,
+      "loss": 0.7671,
+      "step": 13216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2937787953417393e-05,
+      "loss": 1.0,
+      "step": 13217
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2936797252448823e-05,
+      "loss": 0.937,
+      "step": 13218
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2935806519934233e-05,
+      "loss": 1.0234,
+      "step": 13219
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2934815755884272e-05,
+      "loss": 1.2305,
+      "step": 13220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2933824960309576e-05,
+      "loss": 1.2207,
+      "step": 13221
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2932834133220794e-05,
+      "loss": 1.1006,
+      "step": 13222
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293184327462856e-05,
+      "loss": 0.8555,
+      "step": 13223
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2930852384543528e-05,
+      "loss": 1.0205,
+      "step": 13224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2929861462976338e-05,
+      "loss": 1.4648,
+      "step": 13225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2928870509937633e-05,
+      "loss": 0.6953,
+      "step": 13226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2927879525438054e-05,
+      "loss": 0.7471,
+      "step": 13227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2926888509488253e-05,
+      "loss": 0.8057,
+      "step": 13228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2925897462098871e-05,
+      "loss": 1.2285,
+      "step": 13229
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2924906383280554e-05,
+      "loss": 0.7998,
+      "step": 13230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292391527304395e-05,
+      "loss": 0.9126,
+      "step": 13231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29229241313997e-05,
+      "loss": 1.0186,
+      "step": 13232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2921932958358452e-05,
+      "loss": 1.1416,
+      "step": 13233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292094175393086e-05,
+      "loss": 1.2354,
+      "step": 13234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291995051812756e-05,
+      "loss": 1.4209,
+      "step": 13235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291895925095921e-05,
+      "loss": 1.0977,
+      "step": 13236
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2917967952436447e-05,
+      "loss": 1.0498,
+      "step": 13237
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2916976622569932e-05,
+      "loss": 0.9033,
+      "step": 13238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2915985261370304e-05,
+      "loss": 0.9805,
+      "step": 13239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2914993868848211e-05,
+      "loss": 1.3086,
+      "step": 13240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2914002445014311e-05,
+      "loss": 1.0127,
+      "step": 13241
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2913010989879244e-05,
+      "loss": 1.1982,
+      "step": 13242
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2912019503453666e-05,
+      "loss": 1.084,
+      "step": 13243
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2911027985748224e-05,
+      "loss": 0.9971,
+      "step": 13244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291003643677357e-05,
+      "loss": 1.0859,
+      "step": 13245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2909044856540357e-05,
+      "loss": 0.958,
+      "step": 13246
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2908053245059231e-05,
+      "loss": 0.916,
+      "step": 13247
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2907061602340845e-05,
+      "loss": 1.4141,
+      "step": 13248
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2906069928395858e-05,
+      "loss": 1.2734,
+      "step": 13249
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2905078223234907e-05,
+      "loss": 1.0967,
+      "step": 13250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2904086486868664e-05,
+      "loss": 1.2988,
+      "step": 13251
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2903094719307765e-05,
+      "loss": 1.041,
+      "step": 13252
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2902102920562873e-05,
+      "loss": 0.9141,
+      "step": 13253
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2901111090644637e-05,
+      "loss": 1.124,
+      "step": 13254
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2900119229563712e-05,
+      "loss": 1.4395,
+      "step": 13255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2899127337330755e-05,
+      "loss": 0.9238,
+      "step": 13256
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2898135413956416e-05,
+      "loss": 1.0229,
+      "step": 13257
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2897143459451353e-05,
+      "loss": 1.4141,
+      "step": 13258
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2896151473826219e-05,
+      "loss": 1.0273,
+      "step": 13259
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2895159457091675e-05,
+      "loss": 1.2861,
+      "step": 13260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2894167409258367e-05,
+      "loss": 1.1387,
+      "step": 13261
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2893175330336963e-05,
+      "loss": 1.4609,
+      "step": 13262
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2892183220338108e-05,
+      "loss": 1.3086,
+      "step": 13263
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2891191079272467e-05,
+      "loss": 1.1611,
+      "step": 13264
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2890198907150694e-05,
+      "loss": 1.0166,
+      "step": 13265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2889206703983448e-05,
+      "loss": 0.7451,
+      "step": 13266
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2888214469781386e-05,
+      "loss": 0.8594,
+      "step": 13267
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2887222204555165e-05,
+      "loss": 1.2832,
+      "step": 13268
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2886229908315444e-05,
+      "loss": 1.3535,
+      "step": 13269
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2885237581072885e-05,
+      "loss": 1.0508,
+      "step": 13270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2884245222838144e-05,
+      "loss": 1.0713,
+      "step": 13271
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288325283362188e-05,
+      "loss": 1.043,
+      "step": 13272
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2882260413434758e-05,
+      "loss": 1.123,
+      "step": 13273
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288126796228743e-05,
+      "loss": 1.0791,
+      "step": 13274
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2880275480190565e-05,
+      "loss": 1.1204,
+      "step": 13275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2879282967154816e-05,
+      "loss": 1.1689,
+      "step": 13276
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2878290423190853e-05,
+      "loss": 0.9199,
+      "step": 13277
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.287729784830933e-05,
+      "loss": 1.0479,
+      "step": 13278
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2876305242520913e-05,
+      "loss": 1.0537,
+      "step": 13279
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2875312605836263e-05,
+      "loss": 1.0083,
+      "step": 13280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2874319938266038e-05,
+      "loss": 1.0947,
+      "step": 13281
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2873327239820912e-05,
+      "loss": 0.9331,
+      "step": 13282
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2872334510511536e-05,
+      "loss": 1.0596,
+      "step": 13283
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2871341750348582e-05,
+      "loss": 1.21,
+      "step": 13284
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2870348959342712e-05,
+      "loss": 1.0811,
+      "step": 13285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2869356137504586e-05,
+      "loss": 1.1406,
+      "step": 13286
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2868363284844874e-05,
+      "loss": 1.1426,
+      "step": 13287
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2867370401374237e-05,
+      "loss": 1.0791,
+      "step": 13288
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286637748710334e-05,
+      "loss": 1.3193,
+      "step": 13289
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2865384542042856e-05,
+      "loss": 0.8525,
+      "step": 13290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2864391566203443e-05,
+      "loss": 1.0234,
+      "step": 13291
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2863398559595765e-05,
+      "loss": 1.1846,
+      "step": 13292
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2862405522230499e-05,
+      "loss": 1.1631,
+      "step": 13293
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2861412454118298e-05,
+      "loss": 1.2188,
+      "step": 13294
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2860419355269845e-05,
+      "loss": 1.1279,
+      "step": 13295
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2859426225695795e-05,
+      "loss": 0.9424,
+      "step": 13296
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2858433065406821e-05,
+      "loss": 1.3066,
+      "step": 13297
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2857439874413589e-05,
+      "loss": 0.9072,
+      "step": 13298
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2856446652726771e-05,
+      "loss": 1.0703,
+      "step": 13299
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2855453400357032e-05,
+      "loss": 1.1147,
+      "step": 13300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2854460117315043e-05,
+      "loss": 1.21,
+      "step": 13301
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2853466803611474e-05,
+      "loss": 1.3232,
+      "step": 13302
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2852473459256994e-05,
+      "loss": 0.9004,
+      "step": 13303
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2851480084262273e-05,
+      "loss": 1.4121,
+      "step": 13304
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2850486678637983e-05,
+      "loss": 1.0498,
+      "step": 13305
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2849493242394796e-05,
+      "loss": 1.0547,
+      "step": 13306
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2848499775543373e-05,
+      "loss": 1.3818,
+      "step": 13307
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.28475062780944e-05,
+      "loss": 1.3086,
+      "step": 13308
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284651275005854e-05,
+      "loss": 0.7961,
+      "step": 13309
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2845519191446464e-05,
+      "loss": 1.1602,
+      "step": 13310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284452560226885e-05,
+      "loss": 1.1523,
+      "step": 13311
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284353198253637e-05,
+      "loss": 1.0322,
+      "step": 13312
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2842538332259694e-05,
+      "loss": 0.8174,
+      "step": 13313
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2841544651449498e-05,
+      "loss": 1.3745,
+      "step": 13314
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2840550940116452e-05,
+      "loss": 1.1455,
+      "step": 13315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2839557198271233e-05,
+      "loss": 1.1094,
+      "step": 13316
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2838563425924517e-05,
+      "loss": 0.9316,
+      "step": 13317
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2837569623086976e-05,
+      "loss": 0.6953,
+      "step": 13318
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2836575789769288e-05,
+      "loss": 0.8994,
+      "step": 13319
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2835581925982121e-05,
+      "loss": 1.2598,
+      "step": 13320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283458803173616e-05,
+      "loss": 1.2979,
+      "step": 13321
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2833594107042078e-05,
+      "loss": 0.917,
+      "step": 13322
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2832600151910546e-05,
+      "loss": 0.8447,
+      "step": 13323
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283160616635225e-05,
+      "loss": 0.9189,
+      "step": 13324
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2830612150377859e-05,
+      "loss": 0.6519,
+      "step": 13325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2829618103998059e-05,
+      "loss": 1.2695,
+      "step": 13326
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2828624027223515e-05,
+      "loss": 1.1299,
+      "step": 13327
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2827629920064917e-05,
+      "loss": 1.2754,
+      "step": 13328
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2826635782532938e-05,
+      "loss": 1.3135,
+      "step": 13329
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2825641614638254e-05,
+      "loss": 0.7949,
+      "step": 13330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2824647416391552e-05,
+      "loss": 0.9824,
+      "step": 13331
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2823653187803502e-05,
+      "loss": 1.0303,
+      "step": 13332
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.282265892888479e-05,
+      "loss": 1.2627,
+      "step": 13333
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2821664639646096e-05,
+      "loss": 1.1592,
+      "step": 13334
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2820670320098096e-05,
+      "loss": 0.8936,
+      "step": 13335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2819675970251476e-05,
+      "loss": 0.9678,
+      "step": 13336
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2818681590116912e-05,
+      "loss": 0.9727,
+      "step": 13337
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2817687179705087e-05,
+      "loss": 0.7036,
+      "step": 13338
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2816692739026685e-05,
+      "loss": 0.8262,
+      "step": 13339
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2815698268092387e-05,
+      "loss": 1.374,
+      "step": 13340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2814703766912871e-05,
+      "loss": 1.2695,
+      "step": 13341
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2813709235498824e-05,
+      "loss": 1.0859,
+      "step": 13342
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2812714673860927e-05,
+      "loss": 0.8667,
+      "step": 13343
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2811720082009865e-05,
+      "loss": 1.1689,
+      "step": 13344
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2810725459956316e-05,
+      "loss": 0.9658,
+      "step": 13345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2809730807710973e-05,
+      "loss": 1.0,
+      "step": 13346
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2808736125284513e-05,
+      "loss": 1.29,
+      "step": 13347
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2807741412687625e-05,
+      "loss": 1.1338,
+      "step": 13348
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2806746669930989e-05,
+      "loss": 1.1689,
+      "step": 13349
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2805751897025299e-05,
+      "loss": 0.9658,
+      "step": 13350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2804757093981227e-05,
+      "loss": 1.0391,
+      "step": 13351
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2803762260809471e-05,
+      "loss": 1.0273,
+      "step": 13352
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2802767397520712e-05,
+      "loss": 1.4131,
+      "step": 13353
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2801772504125636e-05,
+      "loss": 1.0669,
+      "step": 13354
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2800777580634929e-05,
+      "loss": 0.915,
+      "step": 13355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279978262705928e-05,
+      "loss": 0.9932,
+      "step": 13356
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2798787643409379e-05,
+      "loss": 0.9248,
+      "step": 13357
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2797792629695909e-05,
+      "loss": 1.0742,
+      "step": 13358
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2796797585929562e-05,
+      "loss": 0.9658,
+      "step": 13359
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279580251212102e-05,
+      "loss": 1.1875,
+      "step": 13360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279480740828098e-05,
+      "loss": 1.4375,
+      "step": 13361
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2793812274420127e-05,
+      "loss": 0.936,
+      "step": 13362
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2792817110549152e-05,
+      "loss": 1.123,
+      "step": 13363
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2791821916678736e-05,
+      "loss": 0.6484,
+      "step": 13364
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2790826692819587e-05,
+      "loss": 1.0547,
+      "step": 13365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2789831438982376e-05,
+      "loss": 1.3848,
+      "step": 13366
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2788836155177809e-05,
+      "loss": 0.9575,
+      "step": 13367
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2787840841416564e-05,
+      "loss": 1.1094,
+      "step": 13368
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2786845497709339e-05,
+      "loss": 1.165,
+      "step": 13369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2785850124066829e-05,
+      "loss": 1.4648,
+      "step": 13370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278485472049972e-05,
+      "loss": 0.8223,
+      "step": 13371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2783859287018707e-05,
+      "loss": 1.1592,
+      "step": 13372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2782863823634482e-05,
+      "loss": 1.2822,
+      "step": 13373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2781868330357738e-05,
+      "loss": 1.2734,
+      "step": 13374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2780872807199168e-05,
+      "loss": 1.1035,
+      "step": 13375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2779877254169466e-05,
+      "loss": 0.8687,
+      "step": 13376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2778881671279323e-05,
+      "loss": 1.0088,
+      "step": 13377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277788605853944e-05,
+      "loss": 0.9736,
+      "step": 13378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2776890415960503e-05,
+      "loss": 1.0283,
+      "step": 13379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2775894743553212e-05,
+      "loss": 0.8096,
+      "step": 13380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2774899041328266e-05,
+      "loss": 1.1875,
+      "step": 13381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2773903309296349e-05,
+      "loss": 1.4941,
+      "step": 13382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277290754746817e-05,
+      "loss": 1.2559,
+      "step": 13383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2771911755854413e-05,
+      "loss": 1.1318,
+      "step": 13384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2770915934465781e-05,
+      "loss": 1.0933,
+      "step": 13385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2769920083312972e-05,
+      "loss": 0.8789,
+      "step": 13386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2768924202406681e-05,
+      "loss": 1.041,
+      "step": 13387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2767928291757604e-05,
+      "loss": 1.2412,
+      "step": 13388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.276693235137644e-05,
+      "loss": 0.9805,
+      "step": 13389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2765936381273886e-05,
+      "loss": 1.377,
+      "step": 13390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2764940381460642e-05,
+      "loss": 1.1421,
+      "step": 13391
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2763944351947406e-05,
+      "loss": 0.7861,
+      "step": 13392
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2762948292744879e-05,
+      "loss": 1.1406,
+      "step": 13393
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2761952203863759e-05,
+      "loss": 1.0576,
+      "step": 13394
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2760956085314739e-05,
+      "loss": 0.9971,
+      "step": 13395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2759959937108533e-05,
+      "loss": 1.1152,
+      "step": 13396
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2758963759255824e-05,
+      "loss": 0.8677,
+      "step": 13397
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.275796755176733e-05,
+      "loss": 1.0327,
+      "step": 13398
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2756971314653742e-05,
+      "loss": 0.9863,
+      "step": 13399
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2755975047925759e-05,
+      "loss": 1.1807,
+      "step": 13400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.275497875159409e-05,
+      "loss": 0.8184,
+      "step": 13401
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.275398242566943e-05,
+      "loss": 1.126,
+      "step": 13402
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2752986070162488e-05,
+      "loss": 0.6157,
+      "step": 13403
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2751989685083959e-05,
+      "loss": 1.1738,
+      "step": 13404
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750993270444552e-05,
+      "loss": 0.6592,
+      "step": 13405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2749996826254969e-05,
+      "loss": 0.9302,
+      "step": 13406
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2749000352525913e-05,
+      "loss": 1.0098,
+      "step": 13407
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2748003849268083e-05,
+      "loss": 0.6929,
+      "step": 13408
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.274700731649219e-05,
+      "loss": 1.1406,
+      "step": 13409
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2746010754208933e-05,
+      "loss": 0.8574,
+      "step": 13410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2745014162429024e-05,
+      "loss": 0.9233,
+      "step": 13411
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.274401754116316e-05,
+      "loss": 0.9053,
+      "step": 13412
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2743020890422049e-05,
+      "loss": 1.3008,
+      "step": 13413
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.27420242102164e-05,
+      "loss": 0.96,
+      "step": 13414
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2741027500556915e-05,
+      "loss": 0.9668,
+      "step": 13415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.27400307614543e-05,
+      "loss": 1.3994,
+      "step": 13416
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2739033992919265e-05,
+      "loss": 0.9307,
+      "step": 13417
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2738037194962516e-05,
+      "loss": 0.8506,
+      "step": 13418
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2737040367594755e-05,
+      "loss": 0.9814,
+      "step": 13419
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2736043510826701e-05,
+      "loss": 0.8506,
+      "step": 13420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2735046624669048e-05,
+      "loss": 1.1289,
+      "step": 13421
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2734049709132515e-05,
+      "loss": 0.8755,
+      "step": 13422
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2733052764227805e-05,
+      "loss": 1.3311,
+      "step": 13423
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2732055789965627e-05,
+      "loss": 1.1689,
+      "step": 13424
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2731058786356693e-05,
+      "loss": 1.0938,
+      "step": 13425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273006175341171e-05,
+      "loss": 1.126,
+      "step": 13426
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272906469114139e-05,
+      "loss": 1.04,
+      "step": 13427
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2728067599556445e-05,
+      "loss": 1.1748,
+      "step": 13428
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2727070478667577e-05,
+      "loss": 0.834,
+      "step": 13429
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2726073328485502e-05,
+      "loss": 0.9053,
+      "step": 13430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2725076149020931e-05,
+      "loss": 0.999,
+      "step": 13431
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2724078940284579e-05,
+      "loss": 0.7881,
+      "step": 13432
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272308170228715e-05,
+      "loss": 1.248,
+      "step": 13433
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2722084435039362e-05,
+      "loss": 1.0898,
+      "step": 13434
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2721087138551922e-05,
+      "loss": 1.1582,
+      "step": 13435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272008981283555e-05,
+      "loss": 0.8838,
+      "step": 13436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2719092457900952e-05,
+      "loss": 1.1006,
+      "step": 13437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2718095073758847e-05,
+      "loss": 0.8853,
+      "step": 13438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271709766041994e-05,
+      "loss": 0.9639,
+      "step": 13439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2716100217894955e-05,
+      "loss": 1.4375,
+      "step": 13440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2715102746194598e-05,
+      "loss": 0.9404,
+      "step": 13441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2714105245329591e-05,
+      "loss": 0.8496,
+      "step": 13442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2713107715310641e-05,
+      "loss": 1.4238,
+      "step": 13443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271211015614847e-05,
+      "loss": 1.2793,
+      "step": 13444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2711112567853787e-05,
+      "loss": 0.8169,
+      "step": 13445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2710114950437314e-05,
+      "loss": 0.999,
+      "step": 13446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2709117303909761e-05,
+      "loss": 0.9883,
+      "step": 13447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270811962828185e-05,
+      "loss": 0.9155,
+      "step": 13448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2707121923564293e-05,
+      "loss": 1.3545,
+      "step": 13449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270612418976781e-05,
+      "loss": 0.7769,
+      "step": 13450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2705126426903119e-05,
+      "loss": 0.9268,
+      "step": 13451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2704128634980933e-05,
+      "loss": 1.084,
+      "step": 13452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2703130814011977e-05,
+      "loss": 1.0752,
+      "step": 13453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270213296400696e-05,
+      "loss": 0.9541,
+      "step": 13454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2701135084976609e-05,
+      "loss": 1.0898,
+      "step": 13455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2700137176931637e-05,
+      "loss": 0.7881,
+      "step": 13456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269913923988277e-05,
+      "loss": 0.9971,
+      "step": 13457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2698141273840719e-05,
+      "loss": 1.1631,
+      "step": 13458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2697143278816209e-05,
+      "loss": 0.9111,
+      "step": 13459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269614525481996e-05,
+      "loss": 1.0215,
+      "step": 13460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269514720186269e-05,
+      "loss": 0.8374,
+      "step": 13461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2694149119955122e-05,
+      "loss": 0.9746,
+      "step": 13462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2693151009107976e-05,
+      "loss": 1.0869,
+      "step": 13463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2692152869331974e-05,
+      "loss": 1.3984,
+      "step": 13464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2691154700637838e-05,
+      "loss": 0.9121,
+      "step": 13465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2690156503036288e-05,
+      "loss": 1.0,
+      "step": 13466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2689158276538045e-05,
+      "loss": 1.1992,
+      "step": 13467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2688160021153837e-05,
+      "loss": 1.4277,
+      "step": 13468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2687161736894386e-05,
+      "loss": 1.0811,
+      "step": 13469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2686163423770408e-05,
+      "loss": 1.3223,
+      "step": 13470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2685165081792636e-05,
+      "loss": 0.9766,
+      "step": 13471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2684166710971788e-05,
+      "loss": 0.8984,
+      "step": 13472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2683168311318593e-05,
+      "loss": 0.9893,
+      "step": 13473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2682169882843766e-05,
+      "loss": 1.1631,
+      "step": 13474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2681171425558044e-05,
+      "loss": 0.9512,
+      "step": 13475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2680172939472139e-05,
+      "loss": 1.1348,
+      "step": 13476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2679174424596789e-05,
+      "loss": 1.0957,
+      "step": 13477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2678175880942712e-05,
+      "loss": 1.2031,
+      "step": 13478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2677177308520636e-05,
+      "loss": 1.1436,
+      "step": 13479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2676178707341287e-05,
+      "loss": 0.9463,
+      "step": 13480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267518007741539e-05,
+      "loss": 0.8711,
+      "step": 13481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267418141875368e-05,
+      "loss": 0.917,
+      "step": 13482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2673182731366871e-05,
+      "loss": 1.2012,
+      "step": 13483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2672184015265702e-05,
+      "loss": 1.2607,
+      "step": 13484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2671185270460895e-05,
+      "loss": 1.1436,
+      "step": 13485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267018649696318e-05,
+      "loss": 1.0713,
+      "step": 13486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2669187694783285e-05,
+      "loss": 0.9902,
+      "step": 13487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2668188863931938e-05,
+      "loss": 1.0947,
+      "step": 13488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2667190004419871e-05,
+      "loss": 1.2988,
+      "step": 13489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.266619111625781e-05,
+      "loss": 1.0801,
+      "step": 13490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2665192199456486e-05,
+      "loss": 1.376,
+      "step": 13491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2664193254026627e-05,
+      "loss": 0.9883,
+      "step": 13492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2663194279978968e-05,
+      "loss": 1.0146,
+      "step": 13493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2662195277324237e-05,
+      "loss": 1.0234,
+      "step": 13494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2661196246073163e-05,
+      "loss": 0.7412,
+      "step": 13495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.266019718623648e-05,
+      "loss": 0.9395,
+      "step": 13496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2659198097824921e-05,
+      "loss": 1.6377,
+      "step": 13497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2658198980849213e-05,
+      "loss": 1.2832,
+      "step": 13498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2657199835320091e-05,
+      "loss": 0.9082,
+      "step": 13499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2656200661248287e-05,
+      "loss": 0.9609,
+      "step": 13500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2655201458644534e-05,
+      "loss": 0.8291,
+      "step": 13501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2654202227519565e-05,
+      "loss": 0.8257,
+      "step": 13502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2653202967884113e-05,
+      "loss": 0.9224,
+      "step": 13503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2652203679748913e-05,
+      "loss": 0.717,
+      "step": 13504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2651204363124696e-05,
+      "loss": 0.9697,
+      "step": 13505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2650205018022201e-05,
+      "loss": 1.333,
+      "step": 13506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2649205644452158e-05,
+      "loss": 1.1445,
+      "step": 13507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2648206242425305e-05,
+      "loss": 1.0361,
+      "step": 13508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2647206811952374e-05,
+      "loss": 1.0166,
+      "step": 13509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2646207353044107e-05,
+      "loss": 1.0752,
+      "step": 13510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264520786571123e-05,
+      "loss": 0.8652,
+      "step": 13511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2644208349964488e-05,
+      "loss": 1.3691,
+      "step": 13512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2643208805814613e-05,
+      "loss": 1.1416,
+      "step": 13513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264220923327234e-05,
+      "loss": 0.9858,
+      "step": 13514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2641209632348415e-05,
+      "loss": 0.9263,
+      "step": 13515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2640210003053561e-05,
+      "loss": 1.0273,
+      "step": 13516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2639210345398531e-05,
+      "loss": 1.0127,
+      "step": 13517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2638210659394053e-05,
+      "loss": 1.4805,
+      "step": 13518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2637210945050869e-05,
+      "loss": 1.2588,
+      "step": 13519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2636211202379713e-05,
+      "loss": 1.0503,
+      "step": 13520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.263521143139133e-05,
+      "loss": 1.4092,
+      "step": 13521
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2634211632096457e-05,
+      "loss": 0.9072,
+      "step": 13522
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2633211804505832e-05,
+      "loss": 1.3701,
+      "step": 13523
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2632211948630195e-05,
+      "loss": 1.2334,
+      "step": 13524
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.263121206448029e-05,
+      "loss": 0.8662,
+      "step": 13525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2630212152066854e-05,
+      "loss": 0.8779,
+      "step": 13526
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2629212211400625e-05,
+      "loss": 0.5811,
+      "step": 13527
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262821224249235e-05,
+      "loss": 1.4082,
+      "step": 13528
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2627212245352764e-05,
+      "loss": 1.0781,
+      "step": 13529
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2626212219992617e-05,
+      "loss": 1.04,
+      "step": 13530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2625212166422642e-05,
+      "loss": 1.1396,
+      "step": 13531
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2624212084653587e-05,
+      "loss": 1.2598,
+      "step": 13532
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2623211974696193e-05,
+      "loss": 1.0146,
+      "step": 13533
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.26222118365612e-05,
+      "loss": 0.9883,
+      "step": 13534
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2621211670259355e-05,
+      "loss": 0.7998,
+      "step": 13535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.26202114758014e-05,
+      "loss": 1.416,
+      "step": 13536
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2619211253198078e-05,
+      "loss": 1.2539,
+      "step": 13537
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2618211002460135e-05,
+      "loss": 1.0723,
+      "step": 13538
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2617210723598315e-05,
+      "loss": 0.7549,
+      "step": 13539
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2616210416623357e-05,
+      "loss": 1.3496,
+      "step": 13540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2615210081546016e-05,
+      "loss": 0.9902,
+      "step": 13541
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2614209718377026e-05,
+      "loss": 0.9229,
+      "step": 13542
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2613209327127144e-05,
+      "loss": 0.749,
+      "step": 13543
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2612208907807108e-05,
+      "loss": 0.8643,
+      "step": 13544
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2611208460427668e-05,
+      "loss": 1.2393,
+      "step": 13545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2610207984999567e-05,
+      "loss": 1.1982,
+      "step": 13546
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2609207481533554e-05,
+      "loss": 1.4053,
+      "step": 13547
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2608206950040379e-05,
+      "loss": 0.8125,
+      "step": 13548
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2607206390530781e-05,
+      "loss": 0.6772,
+      "step": 13549
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2606205803015515e-05,
+      "loss": 0.9873,
+      "step": 13550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2605205187505325e-05,
+      "loss": 0.8809,
+      "step": 13551
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2604204544010966e-05,
+      "loss": 0.6797,
+      "step": 13552
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2603203872543172e-05,
+      "loss": 1.1074,
+      "step": 13553
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.260220317311271e-05,
+      "loss": 0.5342,
+      "step": 13554
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2601202445730317e-05,
+      "loss": 1.2012,
+      "step": 13555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2600201690406746e-05,
+      "loss": 0.7256,
+      "step": 13556
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2599200907152747e-05,
+      "loss": 1.377,
+      "step": 13557
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2598200095979068e-05,
+      "loss": 1.459,
+      "step": 13558
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2597199256896464e-05,
+      "loss": 0.9277,
+      "step": 13559
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2596198389915677e-05,
+      "loss": 1.2803,
+      "step": 13560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.259519749504747e-05,
+      "loss": 0.7861,
+      "step": 13561
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2594196572302586e-05,
+      "loss": 1.3535,
+      "step": 13562
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2593195621691778e-05,
+      "loss": 1.0342,
+      "step": 13563
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2592194643225798e-05,
+      "loss": 0.9707,
+      "step": 13564
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2591193636915398e-05,
+      "loss": 1.541,
+      "step": 13565
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2590192602771332e-05,
+      "loss": 1.5312,
+      "step": 13566
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258919154080435e-05,
+      "loss": 0.8281,
+      "step": 13567
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2588190451025209e-05,
+      "loss": 1.0039,
+      "step": 13568
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2587189333444659e-05,
+      "loss": 1.252,
+      "step": 13569
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2586188188073456e-05,
+      "loss": 0.9697,
+      "step": 13570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258518701492235e-05,
+      "loss": 1.0742,
+      "step": 13571
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2584185814002099e-05,
+      "loss": 0.9443,
+      "step": 13572
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2583184585323454e-05,
+      "loss": 0.8516,
+      "step": 13573
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2582183328897177e-05,
+      "loss": 1.2344,
+      "step": 13574
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2581182044734015e-05,
+      "loss": 1.3252,
+      "step": 13575
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258018073284473e-05,
+      "loss": 0.8853,
+      "step": 13576
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2579179393240071e-05,
+      "loss": 0.9482,
+      "step": 13577
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2578178025930798e-05,
+      "loss": 0.6719,
+      "step": 13578
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2577176630927668e-05,
+      "loss": 0.79,
+      "step": 13579
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2576175208241434e-05,
+      "loss": 0.75,
+      "step": 13580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2575173757882859e-05,
+      "loss": 1.1875,
+      "step": 13581
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2574172279862695e-05,
+      "loss": 0.8828,
+      "step": 13582
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2573170774191703e-05,
+      "loss": 0.9629,
+      "step": 13583
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2572169240880635e-05,
+      "loss": 0.9409,
+      "step": 13584
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2571167679940258e-05,
+      "loss": 1.0215,
+      "step": 13585
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257016609138132e-05,
+      "loss": 1.0645,
+      "step": 13586
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.256916447521459e-05,
+      "loss": 1.3398,
+      "step": 13587
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2568162831450821e-05,
+      "loss": 0.9775,
+      "step": 13588
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2567161160100771e-05,
+      "loss": 0.9229,
+      "step": 13589
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2566159461175204e-05,
+      "loss": 1.1084,
+      "step": 13590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.256515773468488e-05,
+      "loss": 0.8848,
+      "step": 13591
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2564155980640552e-05,
+      "loss": 0.9683,
+      "step": 13592
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.256315419905299e-05,
+      "loss": 0.9248,
+      "step": 13593
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2562152389932947e-05,
+      "loss": 0.9199,
+      "step": 13594
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.256115055329119e-05,
+      "loss": 0.9023,
+      "step": 13595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2560148689138476e-05,
+      "loss": 1.2197,
+      "step": 13596
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2559146797485571e-05,
+      "loss": 1.1548,
+      "step": 13597
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2558144878343235e-05,
+      "loss": 1.0557,
+      "step": 13598
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2557142931722226e-05,
+      "loss": 1.3008,
+      "step": 13599
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2556140957633313e-05,
+      "loss": 0.9925,
+      "step": 13600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2555138956087257e-05,
+      "loss": 1.0957,
+      "step": 13601
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2554136927094817e-05,
+      "loss": 1.2217,
+      "step": 13602
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.255313487066676e-05,
+      "loss": 0.9611,
+      "step": 13603
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2552132786813849e-05,
+      "loss": 1.0479,
+      "step": 13604
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2551130675546853e-05,
+      "loss": 0.8428,
+      "step": 13605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.255012853687653e-05,
+      "loss": 0.9404,
+      "step": 13606
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2549126370813642e-05,
+      "loss": 0.811,
+      "step": 13607
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2548124177368962e-05,
+      "loss": 1.208,
+      "step": 13608
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254712195655325e-05,
+      "loss": 1.0527,
+      "step": 13609
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2546119708377273e-05,
+      "loss": 1.0215,
+      "step": 13610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2545117432851801e-05,
+      "loss": 0.7061,
+      "step": 13611
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254411512998759e-05,
+      "loss": 1.0264,
+      "step": 13612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2543112799795414e-05,
+      "loss": 1.0249,
+      "step": 13613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2542110442286041e-05,
+      "loss": 0.8486,
+      "step": 13614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2541108057470231e-05,
+      "loss": 0.9629,
+      "step": 13615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2540105645358757e-05,
+      "loss": 1.1885,
+      "step": 13616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2539103205962385e-05,
+      "loss": 1.5234,
+      "step": 13617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2538100739291882e-05,
+      "loss": 1.2275,
+      "step": 13618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2537098245358016e-05,
+      "loss": 1.0654,
+      "step": 13619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2536095724171555e-05,
+      "loss": 0.9385,
+      "step": 13620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2535093175743272e-05,
+      "loss": 1.0283,
+      "step": 13621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253409060008393e-05,
+      "loss": 0.9619,
+      "step": 13622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.25330879972043e-05,
+      "loss": 0.8799,
+      "step": 13623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2532085367115157e-05,
+      "loss": 1.0391,
+      "step": 13624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2531082709827262e-05,
+      "loss": 0.8438,
+      "step": 13625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2530080025351392e-05,
+      "loss": 0.7568,
+      "step": 13626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2529077313698319e-05,
+      "loss": 1.063,
+      "step": 13627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2528074574878803e-05,
+      "loss": 1.0332,
+      "step": 13628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2527071808903628e-05,
+      "loss": 1.4023,
+      "step": 13629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2526069015783553e-05,
+      "loss": 0.9854,
+      "step": 13630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2525066195529363e-05,
+      "loss": 1.165,
+      "step": 13631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252406334815182e-05,
+      "loss": 1.3789,
+      "step": 13632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.25230604736617e-05,
+      "loss": 1.2217,
+      "step": 13633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2522057572069772e-05,
+      "loss": 1.0381,
+      "step": 13634
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2521054643386815e-05,
+      "loss": 1.2217,
+      "step": 13635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2520051687623596e-05,
+      "loss": 0.8418,
+      "step": 13636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.251904870479089e-05,
+      "loss": 1.3164,
+      "step": 13637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2518045694899475e-05,
+      "loss": 0.7676,
+      "step": 13638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2517042657960118e-05,
+      "loss": 1.0283,
+      "step": 13639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2516039593983603e-05,
+      "loss": 1.3154,
+      "step": 13640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2515036502980693e-05,
+      "loss": 1.1089,
+      "step": 13641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2514033384962173e-05,
+      "loss": 1.2529,
+      "step": 13642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2513030239938809e-05,
+      "loss": 0.8809,
+      "step": 13643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2512027067921382e-05,
+      "loss": 1.3076,
+      "step": 13644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2511023868920667e-05,
+      "loss": 1.3789,
+      "step": 13645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.251002064294744e-05,
+      "loss": 0.8774,
+      "step": 13646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2509017390012477e-05,
+      "loss": 1.1885,
+      "step": 13647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2508014110126551e-05,
+      "loss": 1.2881,
+      "step": 13648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2507010803300448e-05,
+      "loss": 1.0449,
+      "step": 13649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2506007469544936e-05,
+      "loss": 1.1621,
+      "step": 13650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2505004108870797e-05,
+      "loss": 0.9189,
+      "step": 13651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2504000721288806e-05,
+      "loss": 1.1006,
+      "step": 13652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2502997306809745e-05,
+      "loss": 0.7007,
+      "step": 13653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2501993865444388e-05,
+      "loss": 1.1465,
+      "step": 13654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2500990397203521e-05,
+      "loss": 0.6746,
+      "step": 13655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249998690209791e-05,
+      "loss": 0.9121,
+      "step": 13656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2498983380138345e-05,
+      "loss": 0.981,
+      "step": 13657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2497979831335604e-05,
+      "loss": 1.0767,
+      "step": 13658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2496976255700461e-05,
+      "loss": 1.3037,
+      "step": 13659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2495972653243705e-05,
+      "loss": 1.2002,
+      "step": 13660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2494969023976106e-05,
+      "loss": 1.3789,
+      "step": 13661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2493965367908454e-05,
+      "loss": 1.4482,
+      "step": 13662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2492961685051525e-05,
+      "loss": 1.1172,
+      "step": 13663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24919579754161e-05,
+      "loss": 1.0303,
+      "step": 13664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2490954239012962e-05,
+      "loss": 0.9741,
+      "step": 13665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2489950475852892e-05,
+      "loss": 1.3281,
+      "step": 13666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2488946685946672e-05,
+      "loss": 0.6807,
+      "step": 13667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2487942869305085e-05,
+      "loss": 1.0342,
+      "step": 13668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2486939025938916e-05,
+      "loss": 0.8677,
+      "step": 13669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2485935155858942e-05,
+      "loss": 0.8794,
+      "step": 13670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.248493125907595e-05,
+      "loss": 1.1484,
+      "step": 13671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2483927335600723e-05,
+      "loss": 1.4385,
+      "step": 13672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2482923385444049e-05,
+      "loss": 0.7363,
+      "step": 13673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24819194086167e-05,
+      "loss": 0.8379,
+      "step": 13674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2480915405129474e-05,
+      "loss": 0.7717,
+      "step": 13675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247991137499315e-05,
+      "loss": 1.2002,
+      "step": 13676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2478907318218509e-05,
+      "loss": 1.041,
+      "step": 13677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2477903234816342e-05,
+      "loss": 0.8535,
+      "step": 13678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2476899124797433e-05,
+      "loss": 1.1416,
+      "step": 13679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2475894988172567e-05,
+      "loss": 0.9209,
+      "step": 13680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247489082495253e-05,
+      "loss": 0.8027,
+      "step": 13681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2473886635148109e-05,
+      "loss": 1.0166,
+      "step": 13682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2472882418770088e-05,
+      "loss": 1.0459,
+      "step": 13683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2471878175829262e-05,
+      "loss": 1.0693,
+      "step": 13684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2470873906336406e-05,
+      "loss": 1.0127,
+      "step": 13685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2469869610302322e-05,
+      "loss": 1.1758,
+      "step": 13686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2468865287737781e-05,
+      "loss": 1.1504,
+      "step": 13687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2467860938653586e-05,
+      "loss": 1.0586,
+      "step": 13688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2466856563060517e-05,
+      "loss": 1.2822,
+      "step": 13689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2465852160969364e-05,
+      "loss": 1.2227,
+      "step": 13690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2464847732390919e-05,
+      "loss": 1.1074,
+      "step": 13691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2463843277335968e-05,
+      "loss": 1.1221,
+      "step": 13692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24628387958153e-05,
+      "loss": 1.0508,
+      "step": 13693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2461834287839706e-05,
+      "loss": 0.9863,
+      "step": 13694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246082975341998e-05,
+      "loss": 0.9941,
+      "step": 13695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2459825192566905e-05,
+      "loss": 1.5977,
+      "step": 13696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2458820605291277e-05,
+      "loss": 0.939,
+      "step": 13697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2457815991603886e-05,
+      "loss": 1.0093,
+      "step": 13698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2456811351515521e-05,
+      "loss": 1.4473,
+      "step": 13699
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2455806685036973e-05,
+      "loss": 1.5586,
+      "step": 13700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2454801992179042e-05,
+      "loss": 0.8975,
+      "step": 13701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2453797272952506e-05,
+      "loss": 1.2471,
+      "step": 13702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245279252736817e-05,
+      "loss": 0.6616,
+      "step": 13703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2451787755436821e-05,
+      "loss": 1.5049,
+      "step": 13704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245078295716925e-05,
+      "loss": 0.8721,
+      "step": 13705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2449778132576256e-05,
+      "loss": 0.8691,
+      "step": 13706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244877328166863e-05,
+      "loss": 0.8115,
+      "step": 13707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244776840445716e-05,
+      "loss": 1.0181,
+      "step": 13708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2446763500952647e-05,
+      "loss": 1.0322,
+      "step": 13709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2445758571165884e-05,
+      "loss": 1.0225,
+      "step": 13710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2444753615107665e-05,
+      "loss": 1.4453,
+      "step": 13711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2443748632788786e-05,
+      "loss": 1.2188,
+      "step": 13712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2442743624220039e-05,
+      "loss": 1.0234,
+      "step": 13713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2441738589412222e-05,
+      "loss": 0.8691,
+      "step": 13714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244073352837613e-05,
+      "loss": 0.9941,
+      "step": 13715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2439728441122559e-05,
+      "loss": 1.3047,
+      "step": 13716
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2438723327662308e-05,
+      "loss": 0.9482,
+      "step": 13717
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2437718188006165e-05,
+      "loss": 0.7905,
+      "step": 13718
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.243671302216494e-05,
+      "loss": 0.7993,
+      "step": 13719
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2435707830149417e-05,
+      "loss": 0.8228,
+      "step": 13720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2434702611970402e-05,
+      "loss": 1.249,
+      "step": 13721
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2433697367638689e-05,
+      "loss": 1.0498,
+      "step": 13722
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2432692097165076e-05,
+      "loss": 1.1885,
+      "step": 13723
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2431686800560363e-05,
+      "loss": 0.96,
+      "step": 13724
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2430681477835349e-05,
+      "loss": 0.9629,
+      "step": 13725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242967612900083e-05,
+      "loss": 1.0332,
+      "step": 13726
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2428670754067608e-05,
+      "loss": 1.25,
+      "step": 13727
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242766535304648e-05,
+      "loss": 1.0107,
+      "step": 13728
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2426659925948248e-05,
+      "loss": 0.7822,
+      "step": 13729
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2425654472783712e-05,
+      "loss": 1.1699,
+      "step": 13730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2424648993563666e-05,
+      "loss": 1.3662,
+      "step": 13731
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242364348829892e-05,
+      "loss": 1.2246,
+      "step": 13732
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2422637957000268e-05,
+      "loss": 1.0264,
+      "step": 13733
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2421632399678514e-05,
+      "loss": 1.1494,
+      "step": 13734
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242062681634446e-05,
+      "loss": 1.4482,
+      "step": 13735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2419621207008902e-05,
+      "loss": 0.9717,
+      "step": 13736
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2418615571682649e-05,
+      "loss": 0.7725,
+      "step": 13737
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2417609910376498e-05,
+      "loss": 1.1074,
+      "step": 13738
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2416604223101255e-05,
+      "loss": 1.0361,
+      "step": 13739
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2415598509867721e-05,
+      "loss": 1.0664,
+      "step": 13740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24145927706867e-05,
+      "loss": 1.2354,
+      "step": 13741
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2413587005568995e-05,
+      "loss": 1.374,
+      "step": 13742
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241258121452541e-05,
+      "loss": 0.9766,
+      "step": 13743
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2411575397566743e-05,
+      "loss": 0.8857,
+      "step": 13744
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2410569554703808e-05,
+      "loss": 1.1836,
+      "step": 13745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2409563685947404e-05,
+      "loss": 0.8662,
+      "step": 13746
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2408557791308336e-05,
+      "loss": 0.9751,
+      "step": 13747
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240755187079741e-05,
+      "loss": 1.127,
+      "step": 13748
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2406545924425426e-05,
+      "loss": 0.8926,
+      "step": 13749
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2405539952203199e-05,
+      "loss": 0.8721,
+      "step": 13750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2404533954141529e-05,
+      "loss": 0.7349,
+      "step": 13751
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240352793025122e-05,
+      "loss": 1.0859,
+      "step": 13752
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2402521880543083e-05,
+      "loss": 1.043,
+      "step": 13753
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2401515805027924e-05,
+      "loss": 0.8125,
+      "step": 13754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2400509703716549e-05,
+      "loss": 1.0352,
+      "step": 13755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2399503576619765e-05,
+      "loss": 1.207,
+      "step": 13756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2398497423748374e-05,
+      "loss": 0.6138,
+      "step": 13757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2397491245113193e-05,
+      "loss": 1.2256,
+      "step": 13758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2396485040725025e-05,
+      "loss": 1.2881,
+      "step": 13759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.239547881059468e-05,
+      "loss": 1.0684,
+      "step": 13760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.239447255473297e-05,
+      "loss": 1.3477,
+      "step": 13761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2393466273150693e-05,
+      "loss": 0.8486,
+      "step": 13762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2392459965858673e-05,
+      "loss": 1.3008,
+      "step": 13763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2391453632867703e-05,
+      "loss": 1.0596,
+      "step": 13764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2390447274188606e-05,
+      "loss": 1.1094,
+      "step": 13765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2389440889832184e-05,
+      "loss": 0.8672,
+      "step": 13766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2388434479809251e-05,
+      "loss": 0.9961,
+      "step": 13767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2387428044130616e-05,
+      "loss": 1.168,
+      "step": 13768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.238642158280709e-05,
+      "loss": 0.7979,
+      "step": 13769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2385415095849487e-05,
+      "loss": 0.9404,
+      "step": 13770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2384408583268615e-05,
+      "loss": 1.0449,
+      "step": 13771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2383402045075283e-05,
+      "loss": 1.1074,
+      "step": 13772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2382395481280309e-05,
+      "loss": 0.8447,
+      "step": 13773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2381388891894506e-05,
+      "loss": 1.2422,
+      "step": 13774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2380382276928676e-05,
+      "loss": 0.8096,
+      "step": 13775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2379375636393643e-05,
+      "loss": 1.2305,
+      "step": 13776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2378368970300211e-05,
+      "loss": 1.1108,
+      "step": 13777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2377362278659202e-05,
+      "loss": 1.1045,
+      "step": 13778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2376355561481423e-05,
+      "loss": 0.8467,
+      "step": 13779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2375348818777691e-05,
+      "loss": 0.8691,
+      "step": 13780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2374342050558818e-05,
+      "loss": 1.2012,
+      "step": 13781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2373335256835619e-05,
+      "loss": 1.416,
+      "step": 13782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237232843761891e-05,
+      "loss": 1.1045,
+      "step": 13783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2371321592919506e-05,
+      "loss": 0.9238,
+      "step": 13784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2370314722748219e-05,
+      "loss": 0.9917,
+      "step": 13785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2369307827115866e-05,
+      "loss": 1.4453,
+      "step": 13786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2368300906033268e-05,
+      "loss": 1.4717,
+      "step": 13787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236729395951123e-05,
+      "loss": 0.7183,
+      "step": 13788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2366286987560579e-05,
+      "loss": 1.1875,
+      "step": 13789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2365279990192123e-05,
+      "loss": 0.6641,
+      "step": 13790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2364272967416686e-05,
+      "loss": 0.8613,
+      "step": 13791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236326591924508e-05,
+      "loss": 1.1582,
+      "step": 13792
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2362258845688122e-05,
+      "loss": 1.0957,
+      "step": 13793
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2361251746756637e-05,
+      "loss": 0.9209,
+      "step": 13794
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2360244622461434e-05,
+      "loss": 0.9941,
+      "step": 13795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2359237472813337e-05,
+      "loss": 0.7998,
+      "step": 13796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235823029782316e-05,
+      "loss": 1.2832,
+      "step": 13797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2357223097501728e-05,
+      "loss": 0.9111,
+      "step": 13798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2356215871859854e-05,
+      "loss": 1.375,
+      "step": 13799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235520862090836e-05,
+      "loss": 1.0674,
+      "step": 13800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235420134465806e-05,
+      "loss": 1.0498,
+      "step": 13801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2353194043119783e-05,
+      "loss": 1.1758,
+      "step": 13802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2352186716304342e-05,
+      "loss": 1.1465,
+      "step": 13803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351179364222564e-05,
+      "loss": 1.0586,
+      "step": 13804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2350171986885263e-05,
+      "loss": 1.1807,
+      "step": 13805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2349164584303261e-05,
+      "loss": 1.1152,
+      "step": 13806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2348157156487384e-05,
+      "loss": 1.2705,
+      "step": 13807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2347149703448447e-05,
+      "loss": 1.0615,
+      "step": 13808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2346142225197277e-05,
+      "loss": 1.0498,
+      "step": 13809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2345134721744693e-05,
+      "loss": 1.0635,
+      "step": 13810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2344127193101518e-05,
+      "loss": 0.9326,
+      "step": 13811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2343119639278573e-05,
+      "loss": 0.8477,
+      "step": 13812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2342112060286683e-05,
+      "loss": 0.9512,
+      "step": 13813
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2341104456136672e-05,
+      "loss": 1.4785,
+      "step": 13814
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2340096826839361e-05,
+      "loss": 1.0732,
+      "step": 13815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2339089172405574e-05,
+      "loss": 0.8037,
+      "step": 13816
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2338081492846135e-05,
+      "loss": 0.8711,
+      "step": 13817
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2337073788171869e-05,
+      "loss": 1.2354,
+      "step": 13818
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2336066058393597e-05,
+      "loss": 0.9678,
+      "step": 13819
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2335058303522149e-05,
+      "loss": 1.1133,
+      "step": 13820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2334050523568343e-05,
+      "loss": 1.502,
+      "step": 13821
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2333042718543013e-05,
+      "loss": 1.5918,
+      "step": 13822
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2332034888456978e-05,
+      "loss": 0.998,
+      "step": 13823
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2331027033321068e-05,
+      "loss": 1.2637,
+      "step": 13824
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2330019153146104e-05,
+      "loss": 0.7383,
+      "step": 13825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2329011247942915e-05,
+      "loss": 0.916,
+      "step": 13826
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2328003317722327e-05,
+      "loss": 1.1455,
+      "step": 13827
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2326995362495168e-05,
+      "loss": 1.0771,
+      "step": 13828
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2325987382272266e-05,
+      "loss": 0.8008,
+      "step": 13829
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2324979377064445e-05,
+      "loss": 0.8857,
+      "step": 13830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2323971346882536e-05,
+      "loss": 1.228,
+      "step": 13831
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2322963291737364e-05,
+      "loss": 1.1299,
+      "step": 13832
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232195521163976e-05,
+      "loss": 0.9561,
+      "step": 13833
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232094710660055e-05,
+      "loss": 1.0835,
+      "step": 13834
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2319938976630563e-05,
+      "loss": 0.8408,
+      "step": 13835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2318930821740628e-05,
+      "loss": 1.0713,
+      "step": 13836
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2317922641941578e-05,
+      "loss": 0.7183,
+      "step": 13837
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2316914437244236e-05,
+      "loss": 0.6841,
+      "step": 13838
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2315906207659434e-05,
+      "loss": 0.6357,
+      "step": 13839
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2314897953198007e-05,
+      "loss": 1.0879,
+      "step": 13840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2313889673870779e-05,
+      "loss": 1.0171,
+      "step": 13841
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2312881369688584e-05,
+      "loss": 1.0439,
+      "step": 13842
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.231187304066225e-05,
+      "loss": 1.1572,
+      "step": 13843
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2310864686802616e-05,
+      "loss": 1.1641,
+      "step": 13844
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2309856308120498e-05,
+      "loss": 1.0156,
+      "step": 13845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2308847904626744e-05,
+      "loss": 1.2061,
+      "step": 13846
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2307839476332176e-05,
+      "loss": 1.0664,
+      "step": 13847
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2306831023247628e-05,
+      "loss": 1.0146,
+      "step": 13848
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2305822545383937e-05,
+      "loss": 1.2617,
+      "step": 13849
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2304814042751927e-05,
+      "loss": 1.0254,
+      "step": 13850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.230380551536244e-05,
+      "loss": 1.2695,
+      "step": 13851
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2302796963226304e-05,
+      "loss": 1.1816,
+      "step": 13852
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2301788386354357e-05,
+      "loss": 1.1328,
+      "step": 13853
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2300779784757428e-05,
+      "loss": 0.9736,
+      "step": 13854
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2299771158446352e-05,
+      "loss": 0.9419,
+      "step": 13855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2298762507431963e-05,
+      "loss": 1.2578,
+      "step": 13856
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2297753831725096e-05,
+      "loss": 1.1426,
+      "step": 13857
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2296745131336589e-05,
+      "loss": 0.9473,
+      "step": 13858
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2295736406277273e-05,
+      "loss": 1.1396,
+      "step": 13859
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2294727656557985e-05,
+      "loss": 1.0752,
+      "step": 13860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2293718882189561e-05,
+      "loss": 1.0127,
+      "step": 13861
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2292710083182837e-05,
+      "loss": 1.3994,
+      "step": 13862
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2291701259548648e-05,
+      "loss": 0.6602,
+      "step": 13863
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.229069241129783e-05,
+      "loss": 0.7051,
+      "step": 13864
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.228968353844122e-05,
+      "loss": 1.0205,
+      "step": 13865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2288674640989657e-05,
+      "loss": 0.9092,
+      "step": 13866
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2287665718953978e-05,
+      "loss": 0.6294,
+      "step": 13867
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2286656772345018e-05,
+      "loss": 0.6958,
+      "step": 13868
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2285647801173614e-05,
+      "loss": 1.3369,
+      "step": 13869
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2284638805450609e-05,
+      "loss": 1.0156,
+      "step": 13870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2283629785186836e-05,
+      "loss": 1.4727,
+      "step": 13871
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2282620740393137e-05,
+      "loss": 0.8896,
+      "step": 13872
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.228161167108035e-05,
+      "loss": 1.2227,
+      "step": 13873
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2280602577259312e-05,
+      "loss": 1.0977,
+      "step": 13874
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2279593458940866e-05,
+      "loss": 1.2412,
+      "step": 13875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2278584316135846e-05,
+      "loss": 1.0703,
+      "step": 13876
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2277575148855101e-05,
+      "loss": 1.1211,
+      "step": 13877
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227656595710946e-05,
+      "loss": 1.2344,
+      "step": 13878
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2275556740909773e-05,
+      "loss": 0.8496,
+      "step": 13879
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2274547500266875e-05,
+      "loss": 1.1514,
+      "step": 13880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227353823519161e-05,
+      "loss": 1.293,
+      "step": 13881
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2272528945694814e-05,
+      "loss": 1.3018,
+      "step": 13882
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2271519631787335e-05,
+      "loss": 0.8936,
+      "step": 13883
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2270510293480009e-05,
+      "loss": 1.1787,
+      "step": 13884
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2269500930783682e-05,
+      "loss": 1.2881,
+      "step": 13885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2268491543709195e-05,
+      "loss": 1.1553,
+      "step": 13886
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2267482132267391e-05,
+      "loss": 1.1318,
+      "step": 13887
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2266472696469114e-05,
+      "loss": 0.4912,
+      "step": 13888
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2265463236325201e-05,
+      "loss": 1.0645,
+      "step": 13889
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2264453751846504e-05,
+      "loss": 0.6958,
+      "step": 13890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2263444243043857e-05,
+      "loss": 0.8662,
+      "step": 13891
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2262434709928111e-05,
+      "loss": 0.8184,
+      "step": 13892
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2261425152510108e-05,
+      "loss": 1.3555,
+      "step": 13893
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2260415570800688e-05,
+      "loss": 1.1211,
+      "step": 13894
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2259405964810704e-05,
+      "loss": 0.5371,
+      "step": 13895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2258396334550992e-05,
+      "loss": 1.2266,
+      "step": 13896
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2257386680032406e-05,
+      "loss": 1.1152,
+      "step": 13897
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2256377001265785e-05,
+      "loss": 0.9565,
+      "step": 13898
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2255367298261976e-05,
+      "loss": 1.0312,
+      "step": 13899
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2254357571031825e-05,
+      "loss": 1.2002,
+      "step": 13900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2253347819586178e-05,
+      "loss": 1.1621,
+      "step": 13901
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2252338043935885e-05,
+      "loss": 1.4224,
+      "step": 13902
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2251328244091787e-05,
+      "loss": 0.9209,
+      "step": 13903
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2250318420064733e-05,
+      "loss": 1.083,
+      "step": 13904
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.224930857186557e-05,
+      "loss": 0.8218,
+      "step": 13905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.224829869950515e-05,
+      "loss": 0.7363,
+      "step": 13906
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2247288802994311e-05,
+      "loss": 1.1484,
+      "step": 13907
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224627888234391e-05,
+      "loss": 1.5547,
+      "step": 13908
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224526893756479e-05,
+      "loss": 0.9814,
+      "step": 13909
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.22442589686678e-05,
+      "loss": 0.7734,
+      "step": 13910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2243248975663794e-05,
+      "loss": 1.2891,
+      "step": 13911
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2242238958563615e-05,
+      "loss": 1.3232,
+      "step": 13912
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2241228917378113e-05,
+      "loss": 0.7607,
+      "step": 13913
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224021885211814e-05,
+      "loss": 1.2979,
+      "step": 13914
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2239208762794543e-05,
+      "loss": 1.3574,
+      "step": 13915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2238198649418174e-05,
+      "loss": 0.9863,
+      "step": 13916
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2237188511999882e-05,
+      "loss": 0.9629,
+      "step": 13917
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2236178350550517e-05,
+      "loss": 1.7227,
+      "step": 13918
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2235168165080938e-05,
+      "loss": 1.0581,
+      "step": 13919
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223415795560198e-05,
+      "loss": 0.8872,
+      "step": 13920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223314772212451e-05,
+      "loss": 0.8281,
+      "step": 13921
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2232137464659368e-05,
+      "loss": 1.0381,
+      "step": 13922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2231127183217415e-05,
+      "loss": 1.0566,
+      "step": 13923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2230116877809497e-05,
+      "loss": 1.5059,
+      "step": 13924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2229106548446469e-05,
+      "loss": 1.2617,
+      "step": 13925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222809619513918e-05,
+      "loss": 1.2637,
+      "step": 13926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2227085817898488e-05,
+      "loss": 1.1152,
+      "step": 13927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2226075416735241e-05,
+      "loss": 0.7466,
+      "step": 13928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2225064991660298e-05,
+      "loss": 1.1592,
+      "step": 13929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2224054542684507e-05,
+      "loss": 0.7544,
+      "step": 13930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2223044069818729e-05,
+      "loss": 0.8555,
+      "step": 13931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2222033573073808e-05,
+      "loss": 0.8906,
+      "step": 13932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2221023052460608e-05,
+      "loss": 0.8252,
+      "step": 13933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222001250798998e-05,
+      "loss": 1.1523,
+      "step": 13934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2219001939672777e-05,
+      "loss": 0.9111,
+      "step": 13935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2217991347519857e-05,
+      "loss": 1.502,
+      "step": 13936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2216980731542074e-05,
+      "loss": 0.9937,
+      "step": 13937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221597009175028e-05,
+      "loss": 1.1113,
+      "step": 13938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221495942815534e-05,
+      "loss": 1.2158,
+      "step": 13939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2213948740768102e-05,
+      "loss": 0.8799,
+      "step": 13940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2212938029599428e-05,
+      "loss": 0.9004,
+      "step": 13941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2211927294660172e-05,
+      "loss": 0.7666,
+      "step": 13942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2210916535961187e-05,
+      "loss": 0.7461,
+      "step": 13943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2209905753513337e-05,
+      "loss": 0.771,
+      "step": 13944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2208894947327479e-05,
+      "loss": 1.0376,
+      "step": 13945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2207884117414465e-05,
+      "loss": 0.8984,
+      "step": 13946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2206873263785156e-05,
+      "loss": 0.9717,
+      "step": 13947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2205862386450414e-05,
+      "loss": 1.0254,
+      "step": 13948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2204851485421092e-05,
+      "loss": 1.1719,
+      "step": 13949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2203840560708053e-05,
+      "loss": 1.1367,
+      "step": 13950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2202829612322151e-05,
+      "loss": 0.6997,
+      "step": 13951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.220181864027425e-05,
+      "loss": 0.6372,
+      "step": 13952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2200807644575206e-05,
+      "loss": 1.0625,
+      "step": 13953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2199796625235883e-05,
+      "loss": 0.9307,
+      "step": 13954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2198785582267136e-05,
+      "loss": 0.9434,
+      "step": 13955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2197774515679827e-05,
+      "loss": 1.1504,
+      "step": 13956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219676342548482e-05,
+      "loss": 0.9248,
+      "step": 13957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219575231169297e-05,
+      "loss": 1.2334,
+      "step": 13958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2194741174315143e-05,
+      "loss": 1.166,
+      "step": 13959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2193730013362197e-05,
+      "loss": 0.7925,
+      "step": 13960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2192718828844994e-05,
+      "loss": 1.0933,
+      "step": 13961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2191707620774397e-05,
+      "loss": 0.8545,
+      "step": 13962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2190696389161272e-05,
+      "loss": 0.8887,
+      "step": 13963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2189685134016469e-05,
+      "loss": 0.8865,
+      "step": 13964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2188673855350865e-05,
+      "loss": 1.043,
+      "step": 13965
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2187662553175313e-05,
+      "loss": 0.8516,
+      "step": 13966
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218665122750068e-05,
+      "loss": 1.0996,
+      "step": 13967
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2185639878337827e-05,
+      "loss": 0.9663,
+      "step": 13968
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218462850569762e-05,
+      "loss": 0.8911,
+      "step": 13969
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2183617109590925e-05,
+      "loss": 0.9683,
+      "step": 13970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2182605690028597e-05,
+      "loss": 0.9575,
+      "step": 13971
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2181594247021508e-05,
+      "loss": 0.8525,
+      "step": 13972
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2180582780580523e-05,
+      "loss": 0.8564,
+      "step": 13973
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2179571290716504e-05,
+      "loss": 1.1885,
+      "step": 13974
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2178559777440316e-05,
+      "loss": 1.3906,
+      "step": 13975
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2177548240762825e-05,
+      "loss": 0.9297,
+      "step": 13976
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2176536680694895e-05,
+      "loss": 0.8711,
+      "step": 13977
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2175525097247396e-05,
+      "loss": 1.0615,
+      "step": 13978
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2174513490431188e-05,
+      "loss": 1.084,
+      "step": 13979
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2173501860257145e-05,
+      "loss": 1.1396,
+      "step": 13980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2172490206736124e-05,
+      "loss": 1.2119,
+      "step": 13981
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2171478529879002e-05,
+      "loss": 0.5281,
+      "step": 13982
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.217046682969664e-05,
+      "loss": 0.9648,
+      "step": 13983
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2169455106199901e-05,
+      "loss": 0.9644,
+      "step": 13984
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2168443359399664e-05,
+      "loss": 1.041,
+      "step": 13985
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2167431589306791e-05,
+      "loss": 1.0039,
+      "step": 13986
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2166419795932147e-05,
+      "loss": 0.9854,
+      "step": 13987
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2165407979286603e-05,
+      "loss": 1.168,
+      "step": 13988
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2164396139381029e-05,
+      "loss": 0.8213,
+      "step": 13989
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2163384276226293e-05,
+      "loss": 0.9434,
+      "step": 13990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2162372389833266e-05,
+      "loss": 1.2441,
+      "step": 13991
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.216136048021281e-05,
+      "loss": 0.9653,
+      "step": 13992
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2160348547375802e-05,
+      "loss": 0.8857,
+      "step": 13993
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.215933659133311e-05,
+      "loss": 1.02,
+      "step": 13994
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2158324612095602e-05,
+      "loss": 0.5781,
+      "step": 13995
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2157312609674152e-05,
+      "loss": 1.1865,
+      "step": 13996
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2156300584079625e-05,
+      "loss": 1.1689,
+      "step": 13997
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2155288535322897e-05,
+      "loss": 0.7666,
+      "step": 13998
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2154276463414838e-05,
+      "loss": 0.9788,
+      "step": 13999
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2153264368366317e-05,
+      "loss": 1.3311,
+      "step": 14000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2152252250188207e-05,
+      "loss": 0.9775,
+      "step": 14001
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2151240108891382e-05,
+      "loss": 1.1943,
+      "step": 14002
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2150227944486712e-05,
+      "loss": 1.2354,
+      "step": 14003
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2149215756985066e-05,
+      "loss": 0.9034,
+      "step": 14004
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2148203546397322e-05,
+      "loss": 0.9175,
+      "step": 14005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214719131273435e-05,
+      "loss": 1.0273,
+      "step": 14006
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2146179056007023e-05,
+      "loss": 0.9978,
+      "step": 14007
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2145166776226216e-05,
+      "loss": 1.1797,
+      "step": 14008
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2144154473402803e-05,
+      "loss": 0.9375,
+      "step": 14009
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2143142147547653e-05,
+      "loss": 1.4766,
+      "step": 14010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2142129798671646e-05,
+      "loss": 1.1719,
+      "step": 14011
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2141117426785653e-05,
+      "loss": 1.0635,
+      "step": 14012
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214010503190055e-05,
+      "loss": 1.2441,
+      "step": 14013
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2139092614027208e-05,
+      "loss": 1.1494,
+      "step": 14014
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2138080173176508e-05,
+      "loss": 1.1646,
+      "step": 14015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2137067709359318e-05,
+      "loss": 1.207,
+      "step": 14016
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2136055222586522e-05,
+      "loss": 0.9219,
+      "step": 14017
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.213504271286899e-05,
+      "loss": 0.9326,
+      "step": 14018
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2134030180217598e-05,
+      "loss": 0.9844,
+      "step": 14019
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2133017624643226e-05,
+      "loss": 0.9053,
+      "step": 14020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2132005046156747e-05,
+      "loss": 1.1387,
+      "step": 14021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2130992444769041e-05,
+      "loss": 1.0908,
+      "step": 14022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212997982049098e-05,
+      "loss": 1.0596,
+      "step": 14023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2128967173333448e-05,
+      "loss": 1.0371,
+      "step": 14024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2127954503307315e-05,
+      "loss": 1.0605,
+      "step": 14025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2126941810423465e-05,
+      "loss": 1.3047,
+      "step": 14026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2125929094692772e-05,
+      "loss": 1.1865,
+      "step": 14027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2124916356126116e-05,
+      "loss": 0.748,
+      "step": 14028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2123903594734376e-05,
+      "loss": 0.9395,
+      "step": 14029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2122890810528429e-05,
+      "loss": 0.8057,
+      "step": 14030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2121878003519156e-05,
+      "loss": 1.1387,
+      "step": 14031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2120865173717434e-05,
+      "loss": 1.0938,
+      "step": 14032
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2119852321134145e-05,
+      "loss": 1.1152,
+      "step": 14033
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2118839445780164e-05,
+      "loss": 0.7812,
+      "step": 14034
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2117826547666378e-05,
+      "loss": 0.9551,
+      "step": 14035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.211681362680366e-05,
+      "loss": 1.1162,
+      "step": 14036
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2115800683202896e-05,
+      "loss": 1.1895,
+      "step": 14037
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2114787716874963e-05,
+      "loss": 1.29,
+      "step": 14038
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2113774727830746e-05,
+      "loss": 0.9219,
+      "step": 14039
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2112761716081124e-05,
+      "loss": 1.0156,
+      "step": 14040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2111748681636972e-05,
+      "loss": 1.0156,
+      "step": 14041
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2110735624509184e-05,
+      "loss": 1.21,
+      "step": 14042
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2109722544708634e-05,
+      "loss": 1.1416,
+      "step": 14043
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2108709442246203e-05,
+      "loss": 1.2129,
+      "step": 14044
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2107696317132778e-05,
+      "loss": 0.7476,
+      "step": 14045
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210668316937924e-05,
+      "loss": 1.0488,
+      "step": 14046
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2105669998996467e-05,
+      "loss": 1.0967,
+      "step": 14047
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2104656805995352e-05,
+      "loss": 0.9385,
+      "step": 14048
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2103643590386772e-05,
+      "loss": 0.6924,
+      "step": 14049
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210263035218161e-05,
+      "loss": 1.0518,
+      "step": 14050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2101617091390751e-05,
+      "loss": 0.8291,
+      "step": 14051
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2100603808025082e-05,
+      "loss": 0.8788,
+      "step": 14052
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2099590502095484e-05,
+      "loss": 0.9307,
+      "step": 14053
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209857717361284e-05,
+      "loss": 1.208,
+      "step": 14054
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209756382258804e-05,
+      "loss": 1.208,
+      "step": 14055
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2096550449031968e-05,
+      "loss": 1.0391,
+      "step": 14056
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2095537052955505e-05,
+      "loss": 0.7568,
+      "step": 14057
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2094523634369538e-05,
+      "loss": 1.4805,
+      "step": 14058
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2093510193284954e-05,
+      "loss": 1.0762,
+      "step": 14059
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2092496729712639e-05,
+      "loss": 1.3926,
+      "step": 14060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209148324366348e-05,
+      "loss": 0.8579,
+      "step": 14061
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2090469735148363e-05,
+      "loss": 0.96,
+      "step": 14062
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208945620417817e-05,
+      "loss": 1.0039,
+      "step": 14063
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2088442650763798e-05,
+      "loss": 1.1064,
+      "step": 14064
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2087429074916125e-05,
+      "loss": 0.8262,
+      "step": 14065
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2086415476646047e-05,
+      "loss": 0.6177,
+      "step": 14066
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2085401855964443e-05,
+      "loss": 0.7695,
+      "step": 14067
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2084388212882205e-05,
+      "loss": 1.3857,
+      "step": 14068
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208337454741022e-05,
+      "loss": 0.7822,
+      "step": 14069
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2082360859559379e-05,
+      "loss": 0.916,
+      "step": 14070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208134714934057e-05,
+      "loss": 1.1631,
+      "step": 14071
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2080333416764678e-05,
+      "loss": 0.8267,
+      "step": 14072
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2079319661842597e-05,
+      "loss": 1.0049,
+      "step": 14073
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2078305884585212e-05,
+      "loss": 1.0498,
+      "step": 14074
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2077292085003417e-05,
+      "loss": 1.0498,
+      "step": 14075
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20762782631081e-05,
+      "loss": 1.1016,
+      "step": 14076
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.207526441891015e-05,
+      "loss": 1.3926,
+      "step": 14077
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2074250552420459e-05,
+      "loss": 1.0488,
+      "step": 14078
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.207323666364992e-05,
+      "loss": 0.957,
+      "step": 14079
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2072222752609415e-05,
+      "loss": 1.0732,
+      "step": 14080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2071208819309846e-05,
+      "loss": 1.1846,
+      "step": 14081
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2070194863762094e-05,
+      "loss": 1.1328,
+      "step": 14082
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2069180885977059e-05,
+      "loss": 1.2051,
+      "step": 14083
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206816688596563e-05,
+      "loss": 0.9995,
+      "step": 14084
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2067152863738696e-05,
+      "loss": 1.2969,
+      "step": 14085
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2066138819307155e-05,
+      "loss": 0.9961,
+      "step": 14086
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2065124752681894e-05,
+      "loss": 1.0303,
+      "step": 14087
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206411066387381e-05,
+      "loss": 1.2383,
+      "step": 14088
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206309655289379e-05,
+      "loss": 1.1934,
+      "step": 14089
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2062082419752737e-05,
+      "loss": 1.2979,
+      "step": 14090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2061068264461534e-05,
+      "loss": 0.8359,
+      "step": 14091
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2060054087031081e-05,
+      "loss": 1.1338,
+      "step": 14092
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.205903988747227e-05,
+      "loss": 0.8994,
+      "step": 14093
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2058025665795996e-05,
+      "loss": 1.0957,
+      "step": 14094
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2057011422013151e-05,
+      "loss": 1.1348,
+      "step": 14095
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2055997156134633e-05,
+      "loss": 1.1133,
+      "step": 14096
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2054982868171339e-05,
+      "loss": 0.8433,
+      "step": 14097
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2053968558134152e-05,
+      "loss": 1.25,
+      "step": 14098
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2052954226033983e-05,
+      "loss": 1.1367,
+      "step": 14099
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2051939871881716e-05,
+      "loss": 1.0107,
+      "step": 14100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2050925495688257e-05,
+      "loss": 1.6035,
+      "step": 14101
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2049911097464493e-05,
+      "loss": 1.1104,
+      "step": 14102
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2048896677221322e-05,
+      "loss": 0.9326,
+      "step": 14103
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2047882234969643e-05,
+      "loss": 1.3418,
+      "step": 14104
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2046867770720352e-05,
+      "loss": 0.8135,
+      "step": 14105
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2045853284484346e-05,
+      "loss": 1.1348,
+      "step": 14106
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2044838776272522e-05,
+      "loss": 1.0283,
+      "step": 14107
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2043824246095779e-05,
+      "loss": 0.9658,
+      "step": 14108
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2042809693965012e-05,
+      "loss": 0.7148,
+      "step": 14109
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2041795119891124e-05,
+      "loss": 1.1064,
+      "step": 14110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2040780523885003e-05,
+      "loss": 1.3398,
+      "step": 14111
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2039765905957559e-05,
+      "loss": 0.9951,
+      "step": 14112
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2038751266119681e-05,
+      "loss": 0.9541,
+      "step": 14113
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2037736604382279e-05,
+      "loss": 1.2979,
+      "step": 14114
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2036721920756242e-05,
+      "loss": 1.0967,
+      "step": 14115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2035707215252473e-05,
+      "loss": 1.2695,
+      "step": 14116
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2034692487881872e-05,
+      "loss": 1.041,
+      "step": 14117
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2033677738655336e-05,
+      "loss": 0.6934,
+      "step": 14118
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2032662967583771e-05,
+      "loss": 1.2041,
+      "step": 14119
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2031648174678072e-05,
+      "loss": 1.1875,
+      "step": 14120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2030633359949143e-05,
+      "loss": 1.2354,
+      "step": 14121
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202961852340788e-05,
+      "loss": 1.0986,
+      "step": 14122
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202860366506519e-05,
+      "loss": 0.9443,
+      "step": 14123
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202758878493197e-05,
+      "loss": 1.0322,
+      "step": 14124
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2026573883019124e-05,
+      "loss": 0.7021,
+      "step": 14125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2025558959337548e-05,
+      "loss": 1.4297,
+      "step": 14126
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2024544013898156e-05,
+      "loss": 0.9199,
+      "step": 14127
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202352904671184e-05,
+      "loss": 0.7368,
+      "step": 14128
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2022514057789501e-05,
+      "loss": 1.1343,
+      "step": 14129
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202149904714205e-05,
+      "loss": 1.0527,
+      "step": 14130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2020484014780385e-05,
+      "loss": 1.0527,
+      "step": 14131
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2019468960715407e-05,
+      "loss": 0.6802,
+      "step": 14132
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2018453884958026e-05,
+      "loss": 1.155,
+      "step": 14133
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2017438787519139e-05,
+      "loss": 0.9229,
+      "step": 14134
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2016423668409654e-05,
+      "loss": 1.5703,
+      "step": 14135
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201540852764047e-05,
+      "loss": 0.8281,
+      "step": 14136
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2014393365222498e-05,
+      "loss": 0.8643,
+      "step": 14137
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201337818116664e-05,
+      "loss": 0.9819,
+      "step": 14138
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2012362975483798e-05,
+      "loss": 0.8262,
+      "step": 14139
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2011347748184878e-05,
+      "loss": 1.2959,
+      "step": 14140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2010332499280791e-05,
+      "loss": 1.0361,
+      "step": 14141
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2009317228782434e-05,
+      "loss": 1.2549,
+      "step": 14142
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008301936700717e-05,
+      "loss": 0.8359,
+      "step": 14143
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2007286623046544e-05,
+      "loss": 0.9229,
+      "step": 14144
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2006271287830824e-05,
+      "loss": 1.165,
+      "step": 14145
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200525593106446e-05,
+      "loss": 0.6528,
+      "step": 14146
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004240552758364e-05,
+      "loss": 1.376,
+      "step": 14147
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2003225152923434e-05,
+      "loss": 0.9844,
+      "step": 14148
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002209731570586e-05,
+      "loss": 0.9277,
+      "step": 14149
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001194288710722e-05,
+      "loss": 1.085,
+      "step": 14150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2000178824354752e-05,
+      "loss": 1.0068,
+      "step": 14151
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1999163338513582e-05,
+      "loss": 1.0693,
+      "step": 14152
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998147831198118e-05,
+      "loss": 0.9688,
+      "step": 14153
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1997132302419277e-05,
+      "loss": 1.1865,
+      "step": 14154
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996116752187957e-05,
+      "loss": 1.2539,
+      "step": 14155
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1995101180515073e-05,
+      "loss": 1.1475,
+      "step": 14156
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199408558741153e-05,
+      "loss": 1.1514,
+      "step": 14157
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199306997288824e-05,
+      "loss": 0.7603,
+      "step": 14158
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1992054336956114e-05,
+      "loss": 1.0713,
+      "step": 14159
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991038679626055e-05,
+      "loss": 0.6511,
+      "step": 14160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199002300090898e-05,
+      "loss": 1.0195,
+      "step": 14161
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989007300815793e-05,
+      "loss": 1.1934,
+      "step": 14162
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198799157935741e-05,
+      "loss": 0.988,
+      "step": 14163
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1986975836544737e-05,
+      "loss": 1.0625,
+      "step": 14164
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985960072388687e-05,
+      "loss": 1.2109,
+      "step": 14165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198494428690017e-05,
+      "loss": 1.2227,
+      "step": 14166
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.19839284800901e-05,
+      "loss": 1.2725,
+      "step": 14167
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198291265196938e-05,
+      "loss": 1.3486,
+      "step": 14168
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1981896802548935e-05,
+      "loss": 1.0483,
+      "step": 14169
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980880931839664e-05,
+      "loss": 1.2676,
+      "step": 14170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197986503985249e-05,
+      "loss": 1.1543,
+      "step": 14171
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978849126598317e-05,
+      "loss": 0.7051,
+      "step": 14172
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197783319208806e-05,
+      "loss": 0.8174,
+      "step": 14173
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976817236332635e-05,
+      "loss": 0.8486,
+      "step": 14174
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197580125934295e-05,
+      "loss": 0.9189,
+      "step": 14175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1974785261129922e-05,
+      "loss": 0.6865,
+      "step": 14176
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1973769241704462e-05,
+      "loss": 1.0684,
+      "step": 14177
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972753201077486e-05,
+      "loss": 1.2021,
+      "step": 14178
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1971737139259906e-05,
+      "loss": 1.083,
+      "step": 14179
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1970721056262642e-05,
+      "loss": 1.0381,
+      "step": 14180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1969704952096597e-05,
+      "loss": 1.082,
+      "step": 14181
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1968688826772695e-05,
+      "loss": 1.1201,
+      "step": 14182
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1967672680301848e-05,
+      "loss": 0.9473,
+      "step": 14183
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196665651269497e-05,
+      "loss": 0.9111,
+      "step": 14184
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196564032396298e-05,
+      "loss": 1.1162,
+      "step": 14185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964624114116784e-05,
+      "loss": 0.9648,
+      "step": 14186
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963607883167313e-05,
+      "loss": 1.2373,
+      "step": 14187
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1962591631125469e-05,
+      "loss": 1.0225,
+      "step": 14188
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1961575358002177e-05,
+      "loss": 1.071,
+      "step": 14189
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1960559063808349e-05,
+      "loss": 1.4473,
+      "step": 14190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1959542748554902e-05,
+      "loss": 0.9414,
+      "step": 14191
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1958526412252755e-05,
+      "loss": 1.001,
+      "step": 14192
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1957510054912825e-05,
+      "loss": 1.1804,
+      "step": 14193
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1956493676546026e-05,
+      "loss": 1.3516,
+      "step": 14194
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1955477277163278e-05,
+      "loss": 0.8408,
+      "step": 14195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1954460856775501e-05,
+      "loss": 0.9478,
+      "step": 14196
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1953444415393607e-05,
+      "loss": 1.4014,
+      "step": 14197
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1952427953028522e-05,
+      "loss": 1.0713,
+      "step": 14198
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1951411469691157e-05,
+      "loss": 0.7295,
+      "step": 14199
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1950394965392437e-05,
+      "loss": 1.1348,
+      "step": 14200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1949378440143274e-05,
+      "loss": 1.1523,
+      "step": 14201
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1948361893954597e-05,
+      "loss": 1.3105,
+      "step": 14202
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1947345326837313e-05,
+      "loss": 1.0283,
+      "step": 14203
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1946328738802353e-05,
+      "loss": 1.165,
+      "step": 14204
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.194531212986063e-05,
+      "loss": 0.8408,
+      "step": 14205
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1944295500023066e-05,
+      "loss": 1.4316,
+      "step": 14206
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1943278849300581e-05,
+      "loss": 1.0732,
+      "step": 14207
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1942262177704096e-05,
+      "loss": 0.7261,
+      "step": 14208
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1941245485244533e-05,
+      "loss": 1.3672,
+      "step": 14209
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1940228771932808e-05,
+      "loss": 1.0156,
+      "step": 14210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.193921203777985e-05,
+      "loss": 1.0156,
+      "step": 14211
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1938195282796573e-05,
+      "loss": 0.8975,
+      "step": 14212
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1937178506993903e-05,
+      "loss": 0.9331,
+      "step": 14213
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1936161710382756e-05,
+      "loss": 0.8955,
+      "step": 14214
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1935144892974067e-05,
+      "loss": 1.2793,
+      "step": 14215
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1934128054778744e-05,
+      "loss": 0.7646,
+      "step": 14216
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1933111195807716e-05,
+      "loss": 1.1685,
+      "step": 14217
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1932094316071905e-05,
+      "loss": 1.0674,
+      "step": 14218
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.193107741558223e-05,
+      "loss": 1.166,
+      "step": 14219
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1930060494349624e-05,
+      "loss": 1.0986,
+      "step": 14220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929043552385004e-05,
+      "loss": 1.1436,
+      "step": 14221
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1928026589699292e-05,
+      "loss": 0.7886,
+      "step": 14222
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1927009606303416e-05,
+      "loss": 0.9453,
+      "step": 14223
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925992602208298e-05,
+      "loss": 1.2109,
+      "step": 14224
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192497557742486e-05,
+      "loss": 1.0176,
+      "step": 14225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1923958531964033e-05,
+      "loss": 0.9307,
+      "step": 14226
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1922941465836733e-05,
+      "loss": 1.2197,
+      "step": 14227
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192192437905389e-05,
+      "loss": 0.8135,
+      "step": 14228
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1920907271626435e-05,
+      "loss": 0.9834,
+      "step": 14229
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919890143565278e-05,
+      "loss": 0.8149,
+      "step": 14230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1918872994881361e-05,
+      "loss": 1.2842,
+      "step": 14231
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917855825585599e-05,
+      "loss": 0.8237,
+      "step": 14232
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1916838635688923e-05,
+      "loss": 0.9277,
+      "step": 14233
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1915821425202258e-05,
+      "loss": 1.1357,
+      "step": 14234
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191480419413653e-05,
+      "loss": 0.9258,
+      "step": 14235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1913786942502667e-05,
+      "loss": 0.9248,
+      "step": 14236
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1912769670311595e-05,
+      "loss": 1.3613,
+      "step": 14237
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1911752377574239e-05,
+      "loss": 0.876,
+      "step": 14238
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191073506430153e-05,
+      "loss": 1.0254,
+      "step": 14239
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1909717730504396e-05,
+      "loss": 1.1401,
+      "step": 14240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1908700376193763e-05,
+      "loss": 0.9766,
+      "step": 14241
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1907683001380562e-05,
+      "loss": 0.9463,
+      "step": 14242
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1906665606075711e-05,
+      "loss": 1.2393,
+      "step": 14243
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.190564819029015e-05,
+      "loss": 1.3262,
+      "step": 14244
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1904630754034803e-05,
+      "loss": 1.1885,
+      "step": 14245
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1903613297320602e-05,
+      "loss": 1.0879,
+      "step": 14246
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1902595820158474e-05,
+      "loss": 1.3145,
+      "step": 14247
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1901578322559345e-05,
+      "loss": 0.9785,
+      "step": 14248
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.190056080453415e-05,
+      "loss": 1.021,
+      "step": 14249
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1899543266093814e-05,
+      "loss": 1.1436,
+      "step": 14250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1898525707249271e-05,
+      "loss": 0.9351,
+      "step": 14251
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.189750812801145e-05,
+      "loss": 1.6055,
+      "step": 14252
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.189649052839128e-05,
+      "loss": 1.0049,
+      "step": 14253
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1895472908399695e-05,
+      "loss": 0.9971,
+      "step": 14254
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1894455268047626e-05,
+      "loss": 1.1123,
+      "step": 14255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1893437607345996e-05,
+      "loss": 0.7031,
+      "step": 14256
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.189241992630575e-05,
+      "loss": 1.4395,
+      "step": 14257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1891402224937805e-05,
+      "loss": 1.3438,
+      "step": 14258
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1890384503253103e-05,
+      "loss": 1.5322,
+      "step": 14259
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1889366761262573e-05,
+      "loss": 1.2725,
+      "step": 14260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1888348998977147e-05,
+      "loss": 1.1162,
+      "step": 14261
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1887331216407755e-05,
+      "loss": 1.3643,
+      "step": 14262
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1886313413565333e-05,
+      "loss": 0.8945,
+      "step": 14263
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1885295590460815e-05,
+      "loss": 1.2969,
+      "step": 14264
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188427774710513e-05,
+      "loss": 0.7354,
+      "step": 14265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1883259883509213e-05,
+      "loss": 0.9482,
+      "step": 14266
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1882241999683999e-05,
+      "loss": 1.3359,
+      "step": 14267
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188122409564042e-05,
+      "loss": 0.835,
+      "step": 14268
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188020617138941e-05,
+      "loss": 0.6631,
+      "step": 14269
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1879188226941907e-05,
+      "loss": 0.623,
+      "step": 14270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1878170262308837e-05,
+      "loss": 1.2217,
+      "step": 14271
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1877152277501142e-05,
+      "loss": 0.9268,
+      "step": 14272
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1876134272529755e-05,
+      "loss": 0.9937,
+      "step": 14273
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187511624740561e-05,
+      "loss": 1.4219,
+      "step": 14274
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1874098202139646e-05,
+      "loss": 1.2676,
+      "step": 14275
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187308013674279e-05,
+      "loss": 1.002,
+      "step": 14276
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1872062051225985e-05,
+      "loss": 1.1758,
+      "step": 14277
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1871043945600167e-05,
+      "loss": 1.1221,
+      "step": 14278
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1870025819876268e-05,
+      "loss": 1.0459,
+      "step": 14279
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1869007674065225e-05,
+      "loss": 0.7446,
+      "step": 14280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186798950817798e-05,
+      "loss": 0.8232,
+      "step": 14281
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1866971322225462e-05,
+      "loss": 1.3008,
+      "step": 14282
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1865953116218613e-05,
+      "loss": 1.0835,
+      "step": 14283
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186493489016837e-05,
+      "loss": 0.8633,
+      "step": 14284
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1863916644085666e-05,
+      "loss": 0.5144,
+      "step": 14285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1862898377981447e-05,
+      "loss": 0.9258,
+      "step": 14286
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186188009186664e-05,
+      "loss": 1.4336,
+      "step": 14287
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1860861785752194e-05,
+      "loss": 0.915,
+      "step": 14288
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1859843459649036e-05,
+      "loss": 0.7002,
+      "step": 14289
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1858825113568118e-05,
+      "loss": 1.3535,
+      "step": 14290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1857806747520369e-05,
+      "loss": 0.8794,
+      "step": 14291
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1856788361516727e-05,
+      "loss": 0.8711,
+      "step": 14292
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1855769955568137e-05,
+      "loss": 0.8369,
+      "step": 14293
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1854751529685533e-05,
+      "loss": 1.2393,
+      "step": 14294
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1853733083879862e-05,
+      "loss": 1.1318,
+      "step": 14295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1852714618162056e-05,
+      "loss": 1.3057,
+      "step": 14296
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1851696132543058e-05,
+      "loss": 0.7466,
+      "step": 14297
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1850677627033807e-05,
+      "loss": 0.8291,
+      "step": 14298
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1849659101645249e-05,
+      "loss": 1.2549,
+      "step": 14299
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1848640556388316e-05,
+      "loss": 0.9414,
+      "step": 14300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1847621991273956e-05,
+      "loss": 1.2266,
+      "step": 14301
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1846603406313105e-05,
+      "loss": 0.873,
+      "step": 14302
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1845584801516707e-05,
+      "loss": 1.1777,
+      "step": 14303
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1844566176895704e-05,
+      "loss": 0.8706,
+      "step": 14304
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1843547532461038e-05,
+      "loss": 0.9917,
+      "step": 14305
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1842528868223645e-05,
+      "loss": 1.0879,
+      "step": 14306
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1841510184194476e-05,
+      "loss": 1.333,
+      "step": 14307
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1840491480384467e-05,
+      "loss": 1.1055,
+      "step": 14308
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1839472756804562e-05,
+      "loss": 1.0264,
+      "step": 14309
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1838454013465706e-05,
+      "loss": 1.0024,
+      "step": 14310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183743525037884e-05,
+      "loss": 1.0713,
+      "step": 14311
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1836416467554905e-05,
+      "loss": 1.1309,
+      "step": 14312
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183539766500485e-05,
+      "loss": 1.1851,
+      "step": 14313
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1834378842739616e-05,
+      "loss": 1.0264,
+      "step": 14314
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1833360000770142e-05,
+      "loss": 1.1592,
+      "step": 14315
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1832341139107381e-05,
+      "loss": 1.1074,
+      "step": 14316
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1831322257762271e-05,
+      "loss": 0.8223,
+      "step": 14317
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1830303356745757e-05,
+      "loss": 1.2812,
+      "step": 14318
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1829284436068789e-05,
+      "loss": 1.0742,
+      "step": 14319
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1828265495742301e-05,
+      "loss": 1.0391,
+      "step": 14320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1827246535777248e-05,
+      "loss": 1.6758,
+      "step": 14321
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1826227556184574e-05,
+      "loss": 1.0283,
+      "step": 14322
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.182520855697522e-05,
+      "loss": 0.9893,
+      "step": 14323
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1824189538160134e-05,
+      "loss": 1.1914,
+      "step": 14324
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1823170499750262e-05,
+      "loss": 1.042,
+      "step": 14325
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1822151441756553e-05,
+      "loss": 1.2168,
+      "step": 14326
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1821132364189947e-05,
+      "loss": 1.0137,
+      "step": 14327
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1820113267061397e-05,
+      "loss": 1.082,
+      "step": 14328
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1819094150381845e-05,
+      "loss": 1.2207,
+      "step": 14329
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1818075014162243e-05,
+      "loss": 0.8789,
+      "step": 14330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181705585841353e-05,
+      "loss": 1.0078,
+      "step": 14331
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1816036683146665e-05,
+      "loss": 1.0537,
+      "step": 14332
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1815017488372585e-05,
+      "loss": 1.252,
+      "step": 14333
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1813998274102242e-05,
+      "loss": 1.0303,
+      "step": 14334
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1812979040346586e-05,
+      "loss": 1.1245,
+      "step": 14335
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1811959787116563e-05,
+      "loss": 1.0332,
+      "step": 14336
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1810940514423121e-05,
+      "loss": 1.0215,
+      "step": 14337
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180992122227721e-05,
+      "loss": 1.1328,
+      "step": 14338
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1808901910689776e-05,
+      "loss": 0.7271,
+      "step": 14339
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1807882579671773e-05,
+      "loss": 1.2627,
+      "step": 14340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1806863229234147e-05,
+      "loss": 1.2812,
+      "step": 14341
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1805843859387847e-05,
+      "loss": 1.5713,
+      "step": 14342
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1804824470143825e-05,
+      "loss": 1.1094,
+      "step": 14343
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180380506151303e-05,
+      "loss": 1.1553,
+      "step": 14344
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1802785633506413e-05,
+      "loss": 1.1514,
+      "step": 14345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1801766186134917e-05,
+      "loss": 0.6748,
+      "step": 14346
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1800746719409507e-05,
+      "loss": 0.9893,
+      "step": 14347
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1799727233341121e-05,
+      "loss": 1.3369,
+      "step": 14348
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1798707727940715e-05,
+      "loss": 1.125,
+      "step": 14349
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179768820321924e-05,
+      "loss": 1.2422,
+      "step": 14350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1796668659187644e-05,
+      "loss": 1.0049,
+      "step": 14351
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1795649095856884e-05,
+      "loss": 0.7822,
+      "step": 14352
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179462951323791e-05,
+      "loss": 0.8037,
+      "step": 14353
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1793609911341672e-05,
+      "loss": 1.1992,
+      "step": 14354
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1792590290179124e-05,
+      "loss": 1.1133,
+      "step": 14355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1791570649761219e-05,
+      "loss": 1.3086,
+      "step": 14356
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1790550990098908e-05,
+      "loss": 1.1377,
+      "step": 14357
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1789531311203147e-05,
+      "loss": 1.0273,
+      "step": 14358
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1788511613084879e-05,
+      "loss": 1.2676,
+      "step": 14359
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.178749189575507e-05,
+      "loss": 1.4219,
+      "step": 14360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1786472159224667e-05,
+      "loss": 0.6968,
+      "step": 14361
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1785452403504625e-05,
+      "loss": 1.0459,
+      "step": 14362
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1784432628605896e-05,
+      "loss": 1.126,
+      "step": 14363
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1783412834539434e-05,
+      "loss": 1.1914,
+      "step": 14364
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1782393021316199e-05,
+      "loss": 0.8027,
+      "step": 14365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1781373188947141e-05,
+      "loss": 1.3232,
+      "step": 14366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1780353337443211e-05,
+      "loss": 0.7583,
+      "step": 14367
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1779333466815371e-05,
+      "loss": 1.1465,
+      "step": 14368
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1778313577074571e-05,
+      "loss": 1.1538,
+      "step": 14369
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177729366823177e-05,
+      "loss": 1.2695,
+      "step": 14370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1776273740297923e-05,
+      "loss": 0.9502,
+      "step": 14371
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177525379328398e-05,
+      "loss": 1.1992,
+      "step": 14372
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1774233827200904e-05,
+      "loss": 1.3037,
+      "step": 14373
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1773213842059647e-05,
+      "loss": 1.0469,
+      "step": 14374
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177219383787117e-05,
+      "loss": 0.999,
+      "step": 14375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1771173814646424e-05,
+      "loss": 0.8682,
+      "step": 14376
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1770153772396367e-05,
+      "loss": 0.8442,
+      "step": 14377
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176913371113196e-05,
+      "loss": 1.1045,
+      "step": 14378
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1768113630864153e-05,
+      "loss": 0.8978,
+      "step": 14379
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1767093531603907e-05,
+      "loss": 1.417,
+      "step": 14380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1766073413362185e-05,
+      "loss": 1.207,
+      "step": 14381
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1765053276149936e-05,
+      "loss": 0.7585,
+      "step": 14382
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1764033119978122e-05,
+      "loss": 1.1465,
+      "step": 14383
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1763012944857699e-05,
+      "loss": 1.0693,
+      "step": 14384
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1761992750799628e-05,
+      "loss": 1.0684,
+      "step": 14385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176097253781487e-05,
+      "loss": 1.0547,
+      "step": 14386
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1759952305914376e-05,
+      "loss": 0.9165,
+      "step": 14387
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1758932055109112e-05,
+      "loss": 1.126,
+      "step": 14388
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1757911785410035e-05,
+      "loss": 0.8008,
+      "step": 14389
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17568914968281e-05,
+      "loss": 1.2021,
+      "step": 14390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1755871189374274e-05,
+      "loss": 0.9949,
+      "step": 14391
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1754850863059513e-05,
+      "loss": 0.9502,
+      "step": 14392
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1753830517894778e-05,
+      "loss": 1.2383,
+      "step": 14393
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1752810153891026e-05,
+      "loss": 0.8516,
+      "step": 14394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175178977105922e-05,
+      "loss": 0.8223,
+      "step": 14395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175076936941032e-05,
+      "loss": 1.0391,
+      "step": 14396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1749748948955288e-05,
+      "loss": 0.8916,
+      "step": 14397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1748728509705083e-05,
+      "loss": 1.2363,
+      "step": 14398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.174770805167067e-05,
+      "loss": 0.9473,
+      "step": 14399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1746687574863004e-05,
+      "loss": 0.9746,
+      "step": 14400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1745667079293053e-05,
+      "loss": 1.041,
+      "step": 14401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1744646564971777e-05,
+      "loss": 1.1758,
+      "step": 14402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1743626031910133e-05,
+      "loss": 1.0645,
+      "step": 14403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1742605480119093e-05,
+      "loss": 0.8818,
+      "step": 14404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1741584909609608e-05,
+      "loss": 1.1328,
+      "step": 14405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.174056432039265e-05,
+      "loss": 0.9971,
+      "step": 14406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1739543712479178e-05,
+      "loss": 0.8145,
+      "step": 14407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1738523085880151e-05,
+      "loss": 1.1533,
+      "step": 14408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173750244060654e-05,
+      "loss": 0.8936,
+      "step": 14409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1736481776669307e-05,
+      "loss": 0.7607,
+      "step": 14410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1735461094079407e-05,
+      "loss": 1.1738,
+      "step": 14411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1734440392847815e-05,
+      "loss": 1.3418,
+      "step": 14412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1733419672985487e-05,
+      "loss": 1.1611,
+      "step": 14413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1732398934503393e-05,
+      "loss": 1.1465,
+      "step": 14414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1731378177412497e-05,
+      "loss": 1.1426,
+      "step": 14415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1730357401723754e-05,
+      "loss": 0.8594,
+      "step": 14416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1729336607448138e-05,
+      "loss": 1.1396,
+      "step": 14417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1728315794596614e-05,
+      "loss": 1.3799,
+      "step": 14418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1727294963180147e-05,
+      "loss": 0.9072,
+      "step": 14419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17262741132097e-05,
+      "loss": 0.8154,
+      "step": 14420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1725253244696235e-05,
+      "loss": 1.4297,
+      "step": 14421
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1724232357650729e-05,
+      "loss": 0.9805,
+      "step": 14422
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1723211452084139e-05,
+      "loss": 0.9404,
+      "step": 14423
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1722190528007429e-05,
+      "loss": 1.1582,
+      "step": 14424
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1721169585431572e-05,
+      "loss": 0.7412,
+      "step": 14425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1720148624367533e-05,
+      "loss": 1.5547,
+      "step": 14426
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1719127644826277e-05,
+      "loss": 0.9893,
+      "step": 14427
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1718106646818774e-05,
+      "loss": 1.2515,
+      "step": 14428
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1717085630355989e-05,
+      "loss": 1.1934,
+      "step": 14429
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1716064595448891e-05,
+      "loss": 1.2031,
+      "step": 14430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1715043542108443e-05,
+      "loss": 1.3545,
+      "step": 14431
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1714022470345619e-05,
+      "loss": 1.2861,
+      "step": 14432
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1713001380171383e-05,
+      "loss": 1.207,
+      "step": 14433
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17119802715967e-05,
+      "loss": 1.2178,
+      "step": 14434
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171095914463255e-05,
+      "loss": 0.7637,
+      "step": 14435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1709937999289888e-05,
+      "loss": 1.083,
+      "step": 14436
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1708916835579696e-05,
+      "loss": 0.9561,
+      "step": 14437
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170789565351293e-05,
+      "loss": 0.8047,
+      "step": 14438
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1706874453100567e-05,
+      "loss": 0.959,
+      "step": 14439
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1705853234353575e-05,
+      "loss": 1.0576,
+      "step": 14440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1704831997282921e-05,
+      "loss": 0.6057,
+      "step": 14441
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1703810741899577e-05,
+      "loss": 1.0586,
+      "step": 14442
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1702789468214515e-05,
+      "loss": 1.167,
+      "step": 14443
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17017681762387e-05,
+      "loss": 1.0322,
+      "step": 14444
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1700746865983108e-05,
+      "loss": 1.2051,
+      "step": 14445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1699725537458708e-05,
+      "loss": 0.895,
+      "step": 14446
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1698704190676464e-05,
+      "loss": 0.7773,
+      "step": 14447
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1697682825647358e-05,
+      "loss": 0.9219,
+      "step": 14448
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1696661442382351e-05,
+      "loss": 1.1396,
+      "step": 14449
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1695640040892424e-05,
+      "loss": 1.1484,
+      "step": 14450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169461862118854e-05,
+      "loss": 0.8994,
+      "step": 14451
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1693597183281678e-05,
+      "loss": 1.3516,
+      "step": 14452
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1692575727182802e-05,
+      "loss": 0.9736,
+      "step": 14453
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1691554252902888e-05,
+      "loss": 1.0801,
+      "step": 14454
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169053276045291e-05,
+      "loss": 1.2705,
+      "step": 14455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1689511249843839e-05,
+      "loss": 1.3047,
+      "step": 14456
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168848972108665e-05,
+      "loss": 0.7539,
+      "step": 14457
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1687468174192309e-05,
+      "loss": 0.6323,
+      "step": 14458
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1686446609171797e-05,
+      "loss": 1.2422,
+      "step": 14459
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1685425026036082e-05,
+      "loss": 0.958,
+      "step": 14460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1684403424796143e-05,
+      "loss": 1.0337,
+      "step": 14461
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1683381805462946e-05,
+      "loss": 0.9742,
+      "step": 14462
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168236016804747e-05,
+      "loss": 0.9805,
+      "step": 14463
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1681338512560692e-05,
+      "loss": 1.1777,
+      "step": 14464
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1680316839013578e-05,
+      "loss": 1.0928,
+      "step": 14465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1679295147417109e-05,
+      "loss": 1.1143,
+      "step": 14466
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1678273437782254e-05,
+      "loss": 1.1006,
+      "step": 14467
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1677251710119997e-05,
+      "loss": 0.7241,
+      "step": 14468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1676229964441302e-05,
+      "loss": 0.9307,
+      "step": 14469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167520820075715e-05,
+      "loss": 1.2568,
+      "step": 14470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167418641907852e-05,
+      "loss": 1.1021,
+      "step": 14471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167316461941638e-05,
+      "loss": 1.0479,
+      "step": 14472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1672142801781709e-05,
+      "loss": 0.8252,
+      "step": 14473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1671120966185486e-05,
+      "loss": 1.1738,
+      "step": 14474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1670099112638683e-05,
+      "loss": 0.875,
+      "step": 14475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1669077241152278e-05,
+      "loss": 0.8833,
+      "step": 14476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1668055351737251e-05,
+      "loss": 0.821,
+      "step": 14477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1667033444404568e-05,
+      "loss": 1.3125,
+      "step": 14478
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.166601151916522e-05,
+      "loss": 1.3906,
+      "step": 14479
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1664989576030173e-05,
+      "loss": 0.9365,
+      "step": 14480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1663967615010411e-05,
+      "loss": 0.6343,
+      "step": 14481
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.166294563611691e-05,
+      "loss": 0.8472,
+      "step": 14482
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1661923639360644e-05,
+      "loss": 0.9238,
+      "step": 14483
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1660901624752596e-05,
+      "loss": 0.6865,
+      "step": 14484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.165987959230374e-05,
+      "loss": 1.332,
+      "step": 14485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1658857542025057e-05,
+      "loss": 0.9541,
+      "step": 14486
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1657835473927525e-05,
+      "loss": 0.8457,
+      "step": 14487
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1656813388022124e-05,
+      "loss": 1.043,
+      "step": 14488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.165579128431983e-05,
+      "loss": 1.2285,
+      "step": 14489
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1654769162831627e-05,
+      "loss": 1.2402,
+      "step": 14490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1653747023568485e-05,
+      "loss": 0.9062,
+      "step": 14491
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1652724866541393e-05,
+      "loss": 0.9985,
+      "step": 14492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1651702691761325e-05,
+      "loss": 1.0361,
+      "step": 14493
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1650680499239265e-05,
+      "loss": 1.1094,
+      "step": 14494
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1649658288986188e-05,
+      "loss": 0.9297,
+      "step": 14495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1648636061013076e-05,
+      "loss": 0.7256,
+      "step": 14496
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1647613815330914e-05,
+      "loss": 0.9629,
+      "step": 14497
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1646591551950677e-05,
+      "loss": 1.0186,
+      "step": 14498
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1645569270883345e-05,
+      "loss": 1.4746,
+      "step": 14499
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1644546972139906e-05,
+      "loss": 0.668,
+      "step": 14500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1643524655731336e-05,
+      "loss": 1.1714,
+      "step": 14501
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1642502321668613e-05,
+      "loss": 1.2383,
+      "step": 14502
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164147996996273e-05,
+      "loss": 0.8081,
+      "step": 14503
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1640457600624653e-05,
+      "loss": 1.0811,
+      "step": 14504
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163943521366538e-05,
+      "loss": 0.8184,
+      "step": 14505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1638412809095878e-05,
+      "loss": 0.9014,
+      "step": 14506
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1637390386927141e-05,
+      "loss": 1.2158,
+      "step": 14507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1636367947170149e-05,
+      "loss": 0.9775,
+      "step": 14508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1635345489835879e-05,
+      "loss": 0.7988,
+      "step": 14509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1634323014935323e-05,
+      "loss": 0.7861,
+      "step": 14510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1633300522479452e-05,
+      "loss": 1.3145,
+      "step": 14511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163227801247926e-05,
+      "loss": 1.1504,
+      "step": 14512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1631255484945727e-05,
+      "loss": 0.752,
+      "step": 14513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1630232939889835e-05,
+      "loss": 0.9258,
+      "step": 14514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162921037732257e-05,
+      "loss": 1.084,
+      "step": 14515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1628187797254916e-05,
+      "loss": 0.9678,
+      "step": 14516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1627165199697852e-05,
+      "loss": 1.6855,
+      "step": 14517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162614258466237e-05,
+      "loss": 1.0625,
+      "step": 14518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162511995215945e-05,
+      "loss": 1.2061,
+      "step": 14519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1624097302200078e-05,
+      "loss": 1.4014,
+      "step": 14520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162307463479524e-05,
+      "loss": 1.2744,
+      "step": 14521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1622051949955915e-05,
+      "loss": 0.959,
+      "step": 14522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1621029247693098e-05,
+      "loss": 0.9482,
+      "step": 14523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1620006528017766e-05,
+      "loss": 0.8701,
+      "step": 14524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1618983790940915e-05,
+      "loss": 1.0444,
+      "step": 14525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161796103647352e-05,
+      "loss": 1.0918,
+      "step": 14526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1616938264626572e-05,
+      "loss": 0.7812,
+      "step": 14527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1615915475411054e-05,
+      "loss": 1.1426,
+      "step": 14528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1614892668837957e-05,
+      "loss": 1.0986,
+      "step": 14529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1613869844918266e-05,
+      "loss": 0.9268,
+      "step": 14530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1612847003662965e-05,
+      "loss": 1.0107,
+      "step": 14531
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1611824145083046e-05,
+      "loss": 1.0693,
+      "step": 14532
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1610801269189493e-05,
+      "loss": 1.4941,
+      "step": 14533
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1609778375993297e-05,
+      "loss": 1.0322,
+      "step": 14534
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1608755465505437e-05,
+      "loss": 1.0762,
+      "step": 14535
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1607732537736911e-05,
+      "loss": 1.1006,
+      "step": 14536
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1606709592698697e-05,
+      "loss": 0.8518,
+      "step": 14537
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1605686630401794e-05,
+      "loss": 1.1758,
+      "step": 14538
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160466365085718e-05,
+      "loss": 1.0166,
+      "step": 14539
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160364065407585e-05,
+      "loss": 1.1768,
+      "step": 14540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160261764006879e-05,
+      "loss": 0.9053,
+      "step": 14541
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160159460884699e-05,
+      "loss": 0.9277,
+      "step": 14542
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160057156042144e-05,
+      "loss": 0.8057,
+      "step": 14543
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1599548494803125e-05,
+      "loss": 1.4551,
+      "step": 14544
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1598525412003039e-05,
+      "loss": 0.9004,
+      "step": 14545
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159750231203217e-05,
+      "loss": 0.751,
+      "step": 14546
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1596479194901506e-05,
+      "loss": 0.7847,
+      "step": 14547
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1595456060622039e-05,
+      "loss": 1.0615,
+      "step": 14548
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159443290920476e-05,
+      "loss": 1.1328,
+      "step": 14549
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1593409740660655e-05,
+      "loss": 1.0469,
+      "step": 14550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1592386555000721e-05,
+      "loss": 1.2129,
+      "step": 14551
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1591363352235945e-05,
+      "loss": 0.9336,
+      "step": 14552
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1590340132377313e-05,
+      "loss": 1.0234,
+      "step": 14553
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1589316895435827e-05,
+      "loss": 1.377,
+      "step": 14554
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1588293641422468e-05,
+      "loss": 1.0059,
+      "step": 14555
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1587270370348235e-05,
+      "loss": 1.082,
+      "step": 14556
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1586247082224115e-05,
+      "loss": 1.4316,
+      "step": 14557
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1585223777061104e-05,
+      "loss": 0.8071,
+      "step": 14558
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1584200454870188e-05,
+      "loss": 0.9365,
+      "step": 14559
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1583177115662366e-05,
+      "loss": 0.8242,
+      "step": 14560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.158215375944862e-05,
+      "loss": 0.8691,
+      "step": 14561
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1581130386239952e-05,
+      "loss": 0.8936,
+      "step": 14562
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1580106996047353e-05,
+      "loss": 1.3281,
+      "step": 14563
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1579083588881814e-05,
+      "loss": 0.7749,
+      "step": 14564
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1578060164754333e-05,
+      "loss": 0.9854,
+      "step": 14565
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1577036723675891e-05,
+      "loss": 0.7954,
+      "step": 14566
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1576013265657497e-05,
+      "loss": 1.1025,
+      "step": 14567
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1574989790710132e-05,
+      "loss": 1.2197,
+      "step": 14568
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.15739662988448e-05,
+      "loss": 0.9102,
+      "step": 14569
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1572942790072486e-05,
+      "loss": 1.2031,
+      "step": 14570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1571919264404188e-05,
+      "loss": 1.0391,
+      "step": 14571
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1570895721850901e-05,
+      "loss": 1.1787,
+      "step": 14572
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1569872162423619e-05,
+      "loss": 1.04,
+      "step": 14573
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1568848586133336e-05,
+      "loss": 0.874,
+      "step": 14574
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1567824992991051e-05,
+      "loss": 1.2334,
+      "step": 14575
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1566801383007752e-05,
+      "loss": 0.6392,
+      "step": 14576
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1565777756194438e-05,
+      "loss": 1.1006,
+      "step": 14577
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1564754112562107e-05,
+      "loss": 0.9814,
+      "step": 14578
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1563730452121748e-05,
+      "loss": 0.959,
+      "step": 14579
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1562706774884365e-05,
+      "loss": 1.2529,
+      "step": 14580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1561683080860947e-05,
+      "loss": 1.2891,
+      "step": 14581
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1560659370062493e-05,
+      "loss": 0.8789,
+      "step": 14582
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.15596356425e-05,
+      "loss": 1.2151,
+      "step": 14583
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1558611898184462e-05,
+      "loss": 0.7788,
+      "step": 14584
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1557588137126878e-05,
+      "loss": 0.9258,
+      "step": 14585
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1556564359338244e-05,
+      "loss": 0.9961,
+      "step": 14586
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.155554056482956e-05,
+      "loss": 1.0947,
+      "step": 14587
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1554516753611818e-05,
+      "loss": 1.031,
+      "step": 14588
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1553492925696016e-05,
+      "loss": 1.0391,
+      "step": 14589
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1552469081093159e-05,
+      "loss": 1.2246,
+      "step": 14590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1551445219814237e-05,
+      "loss": 1.2314,
+      "step": 14591
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1550421341870246e-05,
+      "loss": 1.3125,
+      "step": 14592
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1549397447272195e-05,
+      "loss": 0.8701,
+      "step": 14593
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.154837353603107e-05,
+      "loss": 1.1416,
+      "step": 14594
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1547349608157878e-05,
+      "loss": 0.9346,
+      "step": 14595
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1546325663663616e-05,
+      "loss": 0.9521,
+      "step": 14596
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.154530170255928e-05,
+      "loss": 1.2783,
+      "step": 14597
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.154427772485587e-05,
+      "loss": 1.002,
+      "step": 14598
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1543253730564384e-05,
+      "loss": 1.3076,
+      "step": 14599
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1542229719695828e-05,
+      "loss": 0.9541,
+      "step": 14600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1541205692261194e-05,
+      "loss": 0.7783,
+      "step": 14601
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1540181648271487e-05,
+      "loss": 1.0068,
+      "step": 14602
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1539157587737702e-05,
+      "loss": 0.7271,
+      "step": 14603
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1538133510670842e-05,
+      "loss": 1.1523,
+      "step": 14604
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1537109417081906e-05,
+      "loss": 1.293,
+      "step": 14605
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.15360853069819e-05,
+      "loss": 1.1328,
+      "step": 14606
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1535061180381813e-05,
+      "loss": 1.1279,
+      "step": 14607
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1534037037292653e-05,
+      "loss": 0.9746,
+      "step": 14608
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1533012877725427e-05,
+      "loss": 1.0205,
+      "step": 14609
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1531988701691124e-05,
+      "loss": 0.8564,
+      "step": 14610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1530964509200755e-05,
+      "loss": 1.373,
+      "step": 14611
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1529940300265311e-05,
+      "loss": 1.1113,
+      "step": 14612
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1528916074895808e-05,
+      "loss": 0.8496,
+      "step": 14613
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1527891833103237e-05,
+      "loss": 0.9668,
+      "step": 14614
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1526867574898606e-05,
+      "loss": 1.0942,
+      "step": 14615
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1525843300292912e-05,
+      "loss": 1.2461,
+      "step": 14616
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152481900929716e-05,
+      "loss": 0.8604,
+      "step": 14617
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1523794701922351e-05,
+      "loss": 1.2305,
+      "step": 14618
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1522770378179488e-05,
+      "loss": 1.1387,
+      "step": 14619
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152174603807958e-05,
+      "loss": 1.0737,
+      "step": 14620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152072168163362e-05,
+      "loss": 1.7227,
+      "step": 14621
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1519697308852622e-05,
+      "loss": 0.9121,
+      "step": 14622
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1518672919747577e-05,
+      "loss": 1.1699,
+      "step": 14623
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1517648514329503e-05,
+      "loss": 1.1836,
+      "step": 14624
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151662409260939e-05,
+      "loss": 0.7974,
+      "step": 14625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151559965459825e-05,
+      "loss": 0.9082,
+      "step": 14626
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1514575200307088e-05,
+      "loss": 1.3711,
+      "step": 14627
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1513550729746903e-05,
+      "loss": 1.1289,
+      "step": 14628
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1512526242928703e-05,
+      "loss": 0.8691,
+      "step": 14629
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151150173986349e-05,
+      "loss": 0.9453,
+      "step": 14630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1510477220562273e-05,
+      "loss": 1.2969,
+      "step": 14631
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1509452685036054e-05,
+      "loss": 1.335,
+      "step": 14632
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1508428133295837e-05,
+      "loss": 0.9434,
+      "step": 14633
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1507403565352631e-05,
+      "loss": 0.9941,
+      "step": 14634
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1506378981217444e-05,
+      "loss": 1.3125,
+      "step": 14635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.150535438090127e-05,
+      "loss": 1.4609,
+      "step": 14636
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1504329764415128e-05,
+      "loss": 1.1484,
+      "step": 14637
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1503305131770013e-05,
+      "loss": 1.3311,
+      "step": 14638
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1502280482976942e-05,
+      "loss": 0.7119,
+      "step": 14639
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1501255818046912e-05,
+      "loss": 0.8584,
+      "step": 14640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1500231136990939e-05,
+      "loss": 1.5283,
+      "step": 14641
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149920643982002e-05,
+      "loss": 0.9033,
+      "step": 14642
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1498181726545164e-05,
+      "loss": 0.9023,
+      "step": 14643
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1497156997177386e-05,
+      "loss": 1.0029,
+      "step": 14644
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1496132251727684e-05,
+      "loss": 0.8291,
+      "step": 14645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1495107490207071e-05,
+      "loss": 1.5801,
+      "step": 14646
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1494082712626553e-05,
+      "loss": 0.3589,
+      "step": 14647
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1493057918997137e-05,
+      "loss": 1.4258,
+      "step": 14648
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149203310932983e-05,
+      "loss": 0.8965,
+      "step": 14649
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1491008283635648e-05,
+      "loss": 1.4268,
+      "step": 14650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1489983441925583e-05,
+      "loss": 1.5547,
+      "step": 14651
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1488958584210658e-05,
+      "loss": 1.249,
+      "step": 14652
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1487933710501883e-05,
+      "loss": 1.2607,
+      "step": 14653
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1486908820810253e-05,
+      "loss": 0.9707,
+      "step": 14654
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148588391514679e-05,
+      "loss": 0.6011,
+      "step": 14655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1484858993522492e-05,
+      "loss": 0.8569,
+      "step": 14656
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148383405594838e-05,
+      "loss": 1.0342,
+      "step": 14657
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1482809102435458e-05,
+      "loss": 0.9375,
+      "step": 14658
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148178413299473e-05,
+      "loss": 0.8496,
+      "step": 14659
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1480759147637217e-05,
+      "loss": 1.1865,
+      "step": 14660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1479734146373921e-05,
+      "loss": 0.7463,
+      "step": 14661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1478709129215857e-05,
+      "loss": 1.1177,
+      "step": 14662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1477684096174032e-05,
+      "loss": 0.8955,
+      "step": 14663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1476659047259455e-05,
+      "loss": 0.7988,
+      "step": 14664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1475633982483144e-05,
+      "loss": 0.8779,
+      "step": 14665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1474608901856103e-05,
+      "loss": 1.0381,
+      "step": 14666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1473583805389343e-05,
+      "loss": 0.9707,
+      "step": 14667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.147255869309388e-05,
+      "loss": 1.2046,
+      "step": 14668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1471533564980719e-05,
+      "loss": 1.0576,
+      "step": 14669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1470508421060883e-05,
+      "loss": 0.9053,
+      "step": 14670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1469483261345371e-05,
+      "loss": 0.7993,
+      "step": 14671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.14684580858452e-05,
+      "loss": 1.3799,
+      "step": 14672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1467432894571383e-05,
+      "loss": 1.1553,
+      "step": 14673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146640768753493e-05,
+      "loss": 1.1025,
+      "step": 14674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1465382464746857e-05,
+      "loss": 0.7583,
+      "step": 14675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1464357226218172e-05,
+      "loss": 1.3008,
+      "step": 14676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146333197195989e-05,
+      "loss": 0.6455,
+      "step": 14677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1462306701983026e-05,
+      "loss": 1.1934,
+      "step": 14678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146128141629859e-05,
+      "loss": 1.4111,
+      "step": 14679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1460256114917596e-05,
+      "loss": 0.9277,
+      "step": 14680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1459230797851059e-05,
+      "loss": 1.1562,
+      "step": 14681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1458205465109988e-05,
+      "loss": 1.1387,
+      "step": 14682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1457180116705404e-05,
+      "loss": 0.9062,
+      "step": 14683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1456154752648316e-05,
+      "loss": 0.9336,
+      "step": 14684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1455129372949737e-05,
+      "loss": 1.0752,
+      "step": 14685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1454103977620685e-05,
+      "loss": 1.1885,
+      "step": 14686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1453078566672174e-05,
+      "loss": 0.9414,
+      "step": 14687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1452053140115214e-05,
+      "loss": 1.2217,
+      "step": 14688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1451027697960825e-05,
+      "loss": 1.0762,
+      "step": 14689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145000224022002e-05,
+      "loss": 1.2275,
+      "step": 14690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1448976766903814e-05,
+      "loss": 1.0283,
+      "step": 14691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1447951278023222e-05,
+      "loss": 1.2227,
+      "step": 14692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1446925773589259e-05,
+      "loss": 1.2734,
+      "step": 14693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1445900253612945e-05,
+      "loss": 1.041,
+      "step": 14694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1444874718105288e-05,
+      "loss": 1.0869,
+      "step": 14695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144384916707731e-05,
+      "loss": 1.1807,
+      "step": 14696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1442823600540025e-05,
+      "loss": 0.9404,
+      "step": 14697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1441798018504447e-05,
+      "loss": 0.9326,
+      "step": 14698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1440772420981598e-05,
+      "loss": 1.2266,
+      "step": 14699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143974680798249e-05,
+      "loss": 1.4512,
+      "step": 14700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1438721179518144e-05,
+      "loss": 1.1641,
+      "step": 14701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1437695535599572e-05,
+      "loss": 0.7256,
+      "step": 14702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1436669876237793e-05,
+      "loss": 1.0176,
+      "step": 14703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1435644201443825e-05,
+      "loss": 1.1006,
+      "step": 14704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1434618511228685e-05,
+      "loss": 1.2383,
+      "step": 14705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143359280560339e-05,
+      "loss": 0.8535,
+      "step": 14706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1432567084578958e-05,
+      "loss": 0.9287,
+      "step": 14707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143154134816641e-05,
+      "loss": 1.0342,
+      "step": 14708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143051559637676e-05,
+      "loss": 1.1016,
+      "step": 14709
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1429489829221025e-05,
+      "loss": 1.042,
+      "step": 14710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1428464046710229e-05,
+      "loss": 1.4648,
+      "step": 14711
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1427438248855389e-05,
+      "loss": 1.1123,
+      "step": 14712
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1426412435667518e-05,
+      "loss": 0.8525,
+      "step": 14713
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1425386607157642e-05,
+      "loss": 0.8789,
+      "step": 14714
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1424360763336777e-05,
+      "loss": 0.9697,
+      "step": 14715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1423334904215944e-05,
+      "loss": 1.0684,
+      "step": 14716
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1422309029806158e-05,
+      "loss": 1.0044,
+      "step": 14717
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1421283140118442e-05,
+      "loss": 1.0332,
+      "step": 14718
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1420257235163816e-05,
+      "loss": 0.7256,
+      "step": 14719
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1419231314953301e-05,
+      "loss": 0.7568,
+      "step": 14720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1418205379497913e-05,
+      "loss": 1.2939,
+      "step": 14721
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1417179428808675e-05,
+      "loss": 0.8652,
+      "step": 14722
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1416153462896606e-05,
+      "loss": 0.9004,
+      "step": 14723
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141512748177273e-05,
+      "loss": 1.0146,
+      "step": 14724
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1414101485448066e-05,
+      "loss": 1.2344,
+      "step": 14725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1413075473933629e-05,
+      "loss": 1.7949,
+      "step": 14726
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141204944724045e-05,
+      "loss": 1.2393,
+      "step": 14727
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1411023405379543e-05,
+      "loss": 0.8691,
+      "step": 14728
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1409997348361931e-05,
+      "loss": 0.9482,
+      "step": 14729
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1408971276198638e-05,
+      "loss": 1.1523,
+      "step": 14730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1407945188900685e-05,
+      "loss": 0.6582,
+      "step": 14731
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1406919086479089e-05,
+      "loss": 1.209,
+      "step": 14732
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1405892968944878e-05,
+      "loss": 1.1367,
+      "step": 14733
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1404866836309072e-05,
+      "loss": 0.9893,
+      "step": 14734
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1403840688582692e-05,
+      "loss": 1.1914,
+      "step": 14735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1402814525776763e-05,
+      "loss": 0.9375,
+      "step": 14736
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1401788347902305e-05,
+      "loss": 1.3682,
+      "step": 14737
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1400762154970346e-05,
+      "loss": 0.8926,
+      "step": 14738
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1399735946991901e-05,
+      "loss": 1.2109,
+      "step": 14739
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1398709723978002e-05,
+      "loss": 1.0459,
+      "step": 14740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1397683485939664e-05,
+      "loss": 1.2197,
+      "step": 14741
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1396657232887917e-05,
+      "loss": 1.3047,
+      "step": 14742
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1395630964833782e-05,
+      "loss": 0.9688,
+      "step": 14743
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1394604681788276e-05,
+      "loss": 1.5273,
+      "step": 14744
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1393578383762438e-05,
+      "loss": 0.606,
+      "step": 14745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1392552070767282e-05,
+      "loss": 1.1416,
+      "step": 14746
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1391525742813831e-05,
+      "loss": 1.0518,
+      "step": 14747
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1390499399913116e-05,
+      "loss": 1.0898,
+      "step": 14748
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1389473042076156e-05,
+      "loss": 0.8975,
+      "step": 14749
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138844666931398e-05,
+      "loss": 1.291,
+      "step": 14750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1387420281637606e-05,
+      "loss": 0.9658,
+      "step": 14751
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1386393879058068e-05,
+      "loss": 0.9619,
+      "step": 14752
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1385367461586387e-05,
+      "loss": 1.1611,
+      "step": 14753
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1384341029233588e-05,
+      "loss": 0.8936,
+      "step": 14754
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1383314582010697e-05,
+      "loss": 1.0342,
+      "step": 14755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1382288119928742e-05,
+      "loss": 1.2451,
+      "step": 14756
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1381261642998742e-05,
+      "loss": 0.8477,
+      "step": 14757
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1380235151231733e-05,
+      "loss": 0.9873,
+      "step": 14758
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1379208644638736e-05,
+      "loss": 1.1816,
+      "step": 14759
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1378182123230773e-05,
+      "loss": 1.0488,
+      "step": 14760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1377155587018878e-05,
+      "loss": 0.8003,
+      "step": 14761
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1376129036014073e-05,
+      "loss": 1.3066,
+      "step": 14762
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1375102470227386e-05,
+      "loss": 0.8828,
+      "step": 14763
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1374075889669845e-05,
+      "loss": 0.9507,
+      "step": 14764
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1373049294352475e-05,
+      "loss": 1.1582,
+      "step": 14765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1372022684286308e-05,
+      "loss": 0.9219,
+      "step": 14766
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1370996059482366e-05,
+      "loss": 0.835,
+      "step": 14767
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1369969419951681e-05,
+      "loss": 0.6455,
+      "step": 14768
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1368942765705279e-05,
+      "loss": 0.8237,
+      "step": 14769
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1367916096754185e-05,
+      "loss": 0.9844,
+      "step": 14770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1366889413109435e-05,
+      "loss": 1.0547,
+      "step": 14771
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1365862714782046e-05,
+      "loss": 0.877,
+      "step": 14772
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1364836001783057e-05,
+      "loss": 1.3086,
+      "step": 14773
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.136380927412349e-05,
+      "loss": 1.1196,
+      "step": 14774
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1362782531814376e-05,
+      "loss": 1.1982,
+      "step": 14775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1361755774866743e-05,
+      "loss": 0.8076,
+      "step": 14776
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1360729003291621e-05,
+      "loss": 0.7939,
+      "step": 14777
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1359702217100041e-05,
+      "loss": 1.125,
+      "step": 14778
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.135867541630303e-05,
+      "loss": 1.0083,
+      "step": 14779
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1357648600911615e-05,
+      "loss": 0.9668,
+      "step": 14780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.135662177093683e-05,
+      "loss": 1.1416,
+      "step": 14781
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1355594926389708e-05,
+      "loss": 1.3047,
+      "step": 14782
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1354568067281266e-05,
+      "loss": 1.0781,
+      "step": 14783
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1353541193622551e-05,
+      "loss": 1.0615,
+      "step": 14784
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1352514305424578e-05,
+      "loss": 0.8262,
+      "step": 14785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1351487402698388e-05,
+      "loss": 0.8789,
+      "step": 14786
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1350460485455009e-05,
+      "loss": 0.8389,
+      "step": 14787
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1349433553705465e-05,
+      "loss": 1.1924,
+      "step": 14788
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1348406607460797e-05,
+      "loss": 1.1855,
+      "step": 14789
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1347379646732032e-05,
+      "loss": 1.1641,
+      "step": 14790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1346352671530201e-05,
+      "loss": 1.1602,
+      "step": 14791
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1345325681866335e-05,
+      "loss": 1.0439,
+      "step": 14792
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1344298677751465e-05,
+      "loss": 0.8794,
+      "step": 14793
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1343271659196624e-05,
+      "loss": 1.2139,
+      "step": 14794
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1342244626212845e-05,
+      "loss": 1.001,
+      "step": 14795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1341217578811154e-05,
+      "loss": 1.0869,
+      "step": 14796
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1340190517002592e-05,
+      "loss": 1.0244,
+      "step": 14797
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1339163440798187e-05,
+      "loss": 1.0127,
+      "step": 14798
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1338136350208968e-05,
+      "loss": 0.9805,
+      "step": 14799
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1337109245245976e-05,
+      "loss": 1.3301,
+      "step": 14800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1336082125920234e-05,
+      "loss": 0.9033,
+      "step": 14801
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1335054992242784e-05,
+      "loss": 0.9131,
+      "step": 14802
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1334027844224653e-05,
+      "loss": 1.0996,
+      "step": 14803
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1333000681876877e-05,
+      "loss": 0.9385,
+      "step": 14804
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1331973505210488e-05,
+      "loss": 0.8398,
+      "step": 14805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1330946314236519e-05,
+      "loss": 0.8984,
+      "step": 14806
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1329919108966008e-05,
+      "loss": 0.8506,
+      "step": 14807
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1328891889409983e-05,
+      "loss": 0.8477,
+      "step": 14808
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1327864655579481e-05,
+      "loss": 1.0137,
+      "step": 14809
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1326837407485537e-05,
+      "loss": 1.0244,
+      "step": 14810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1325810145139185e-05,
+      "loss": 0.998,
+      "step": 14811
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1324782868551456e-05,
+      "loss": 0.6514,
+      "step": 14812
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1323755577733392e-05,
+      "loss": 1.041,
+      "step": 14813
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1322728272696017e-05,
+      "loss": 1.0488,
+      "step": 14814
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.132170095345038e-05,
+      "loss": 1.3799,
+      "step": 14815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13206736200075e-05,
+      "loss": 0.9199,
+      "step": 14816
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1319646272378427e-05,
+      "loss": 0.9463,
+      "step": 14817
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1318618910574186e-05,
+      "loss": 1.0303,
+      "step": 14818
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1317591534605819e-05,
+      "loss": 1.21,
+      "step": 14819
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1316564144484357e-05,
+      "loss": 1.0303,
+      "step": 14820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1315536740220838e-05,
+      "loss": 1.0283,
+      "step": 14821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1314509321826299e-05,
+      "loss": 1.0596,
+      "step": 14822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1313481889311774e-05,
+      "loss": 0.917,
+      "step": 14823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1312454442688303e-05,
+      "loss": 1.127,
+      "step": 14824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1311426981966918e-05,
+      "loss": 1.0205,
+      "step": 14825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131039950715866e-05,
+      "loss": 0.9297,
+      "step": 14826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1309372018274559e-05,
+      "loss": 0.8359,
+      "step": 14827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130834451532566e-05,
+      "loss": 0.9121,
+      "step": 14828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1307316998322992e-05,
+      "loss": 0.8447,
+      "step": 14829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13062894672776e-05,
+      "loss": 1.166,
+      "step": 14830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130526192220052e-05,
+      "loss": 1.0049,
+      "step": 14831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1304234363102781e-05,
+      "loss": 0.8652,
+      "step": 14832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1303206789995431e-05,
+      "loss": 0.8643,
+      "step": 14833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1302179202889505e-05,
+      "loss": 1.2168,
+      "step": 14834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1301151601796037e-05,
+      "loss": 1.2207,
+      "step": 14835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1300123986726068e-05,
+      "loss": 1.0342,
+      "step": 14836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1299096357690637e-05,
+      "loss": 1.0986,
+      "step": 14837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1298068714700783e-05,
+      "loss": 0.7842,
+      "step": 14838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1297041057767545e-05,
+      "loss": 0.8369,
+      "step": 14839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1296013386901954e-05,
+      "loss": 0.7949,
+      "step": 14840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.129498570211506e-05,
+      "loss": 0.9893,
+      "step": 14841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1293958003417894e-05,
+      "loss": 0.9209,
+      "step": 14842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1292930290821497e-05,
+      "loss": 1.1748,
+      "step": 14843
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1291902564336912e-05,
+      "loss": 0.7397,
+      "step": 14844
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1290874823975173e-05,
+      "loss": 1.335,
+      "step": 14845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1289847069747326e-05,
+      "loss": 1.0977,
+      "step": 14846
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1288819301664404e-05,
+      "loss": 1.3418,
+      "step": 14847
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1287791519737454e-05,
+      "loss": 0.9507,
+      "step": 14848
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1286763723977509e-05,
+      "loss": 0.8799,
+      "step": 14849
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1285735914395613e-05,
+      "loss": 1.2354,
+      "step": 14850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1284708091002805e-05,
+      "loss": 1.0488,
+      "step": 14851
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1283680253810128e-05,
+      "loss": 0.8926,
+      "step": 14852
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.128265240282862e-05,
+      "loss": 0.9463,
+      "step": 14853
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1281624538069322e-05,
+      "loss": 1.2344,
+      "step": 14854
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1280596659543274e-05,
+      "loss": 0.856,
+      "step": 14855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1279568767261523e-05,
+      "loss": 1.207,
+      "step": 14856
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1278540861235106e-05,
+      "loss": 1.1812,
+      "step": 14857
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1277512941475059e-05,
+      "loss": 1.2178,
+      "step": 14858
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1276485007992432e-05,
+      "loss": 1.0254,
+      "step": 14859
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1275457060798263e-05,
+      "loss": 1.1299,
+      "step": 14860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1274429099903597e-05,
+      "loss": 1.147,
+      "step": 14861
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1273401125319472e-05,
+      "loss": 1.1201,
+      "step": 14862
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1272373137056931e-05,
+      "loss": 0.9814,
+      "step": 14863
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1271345135127017e-05,
+      "loss": 1.2617,
+      "step": 14864
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1270317119540772e-05,
+      "loss": 0.9199,
+      "step": 14865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1269289090309239e-05,
+      "loss": 1.166,
+      "step": 14866
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126826104744346e-05,
+      "loss": 1.1357,
+      "step": 14867
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1267232990954479e-05,
+      "loss": 0.998,
+      "step": 14868
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1266204920853339e-05,
+      "loss": 0.9756,
+      "step": 14869
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1265176837151084e-05,
+      "loss": 1.0547,
+      "step": 14870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1264148739858752e-05,
+      "loss": 1.2441,
+      "step": 14871
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1263120628987394e-05,
+      "loss": 1.0752,
+      "step": 14872
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1262092504548045e-05,
+      "loss": 1.1191,
+      "step": 14873
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126106436655176e-05,
+      "loss": 0.853,
+      "step": 14874
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1260036215009574e-05,
+      "loss": 0.9932,
+      "step": 14875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1259008049932533e-05,
+      "loss": 0.8359,
+      "step": 14876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1257979871331682e-05,
+      "loss": 1.2158,
+      "step": 14877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1256951679218065e-05,
+      "loss": 1.4404,
+      "step": 14878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1255923473602727e-05,
+      "loss": 1.0195,
+      "step": 14879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1254895254496712e-05,
+      "loss": 0.916,
+      "step": 14880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1253867021911064e-05,
+      "loss": 0.5415,
+      "step": 14881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1252838775856829e-05,
+      "loss": 0.9141,
+      "step": 14882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1251810516345056e-05,
+      "loss": 1.1172,
+      "step": 14883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1250782243386779e-05,
+      "loss": 0.6353,
+      "step": 14884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1249753956993057e-05,
+      "loss": 0.6814,
+      "step": 14885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1248725657174923e-05,
+      "loss": 1.2627,
+      "step": 14886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1247697343943433e-05,
+      "loss": 0.9443,
+      "step": 14887
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1246669017309629e-05,
+      "loss": 1.0615,
+      "step": 14888
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124564067728455e-05,
+      "loss": 1.0068,
+      "step": 14889
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1244612323879256e-05,
+      "loss": 1.334,
+      "step": 14890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124358395710478e-05,
+      "loss": 1.1309,
+      "step": 14891
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1242555576972176e-05,
+      "loss": 0.7256,
+      "step": 14892
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124152718349249e-05,
+      "loss": 1.082,
+      "step": 14893
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1240498776676764e-05,
+      "loss": 1.021,
+      "step": 14894
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1239470356536048e-05,
+      "loss": 1.0713,
+      "step": 14895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.123844192308139e-05,
+      "loss": 1.2793,
+      "step": 14896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1237413476323835e-05,
+      "loss": 0.9648,
+      "step": 14897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1236385016274432e-05,
+      "loss": 1.0391,
+      "step": 14898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1235356542944225e-05,
+      "loss": 0.8945,
+      "step": 14899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1234328056344264e-05,
+      "loss": 1.209,
+      "step": 14900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.12332995564856e-05,
+      "loss": 0.8594,
+      "step": 14901
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1232271043379273e-05,
+      "loss": 1.0312,
+      "step": 14902
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1231242517036338e-05,
+      "loss": 0.8047,
+      "step": 14903
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1230213977467837e-05,
+      "loss": 1.5273,
+      "step": 14904
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1229185424684825e-05,
+      "loss": 0.7666,
+      "step": 14905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1228156858698344e-05,
+      "loss": 0.7642,
+      "step": 14906
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1227128279519446e-05,
+      "loss": 0.7803,
+      "step": 14907
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1226099687159178e-05,
+      "loss": 1.2793,
+      "step": 14908
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.122507108162859e-05,
+      "loss": 1.04,
+      "step": 14909
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1224042462938732e-05,
+      "loss": 1.1055,
+      "step": 14910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1223013831100649e-05,
+      "loss": 0.9023,
+      "step": 14911
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1221985186125393e-05,
+      "loss": 0.9453,
+      "step": 14912
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1220956528024015e-05,
+      "loss": 0.8701,
+      "step": 14913
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1219927856807561e-05,
+      "loss": 1.0708,
+      "step": 14914
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1218899172487081e-05,
+      "loss": 0.73,
+      "step": 14915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121787047507363e-05,
+      "loss": 0.7744,
+      "step": 14916
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121684176457825e-05,
+      "loss": 1.123,
+      "step": 14917
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1215813041011998e-05,
+      "loss": 0.9814,
+      "step": 14918
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1214784304385917e-05,
+      "loss": 1.103,
+      "step": 14919
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1213755554711063e-05,
+      "loss": 0.8926,
+      "step": 14920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1212726791998486e-05,
+      "loss": 1.1318,
+      "step": 14921
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1211698016259234e-05,
+      "loss": 1.1133,
+      "step": 14922
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121066922750436e-05,
+      "loss": 1.3125,
+      "step": 14923
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1209640425744912e-05,
+      "loss": 1.0469,
+      "step": 14924
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1208611610991946e-05,
+      "loss": 0.9512,
+      "step": 14925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1207582783256508e-05,
+      "loss": 1.0317,
+      "step": 14926
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1206553942549654e-05,
+      "loss": 1.2246,
+      "step": 14927
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1205525088882428e-05,
+      "loss": 0.9253,
+      "step": 14928
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1204496222265892e-05,
+      "loss": 1.0293,
+      "step": 14929
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1203467342711087e-05,
+      "loss": 0.9834,
+      "step": 14930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1202438450229074e-05,
+      "loss": 0.7529,
+      "step": 14931
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1201409544830901e-05,
+      "loss": 1.1655,
+      "step": 14932
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1200380626527615e-05,
+      "loss": 1.3848,
+      "step": 14933
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.119935169533028e-05,
+      "loss": 1.1025,
+      "step": 14934
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1198322751249937e-05,
+      "loss": 1.291,
+      "step": 14935
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1197293794297648e-05,
+      "loss": 1.2861,
+      "step": 14936
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1196264824484457e-05,
+      "loss": 1.2461,
+      "step": 14937
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1195235841821423e-05,
+      "loss": 0.6528,
+      "step": 14938
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1194206846319595e-05,
+      "loss": 1.1016,
+      "step": 14939
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.119317783799003e-05,
+      "loss": 0.8574,
+      "step": 14940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1192148816843777e-05,
+      "loss": 0.9863,
+      "step": 14941
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1191119782891893e-05,
+      "loss": 0.8982,
+      "step": 14942
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1190090736145429e-05,
+      "loss": 1.6172,
+      "step": 14943
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1189061676615442e-05,
+      "loss": 0.9102,
+      "step": 14944
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1188032604312984e-05,
+      "loss": 1.627,
+      "step": 14945
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1187003519249104e-05,
+      "loss": 1.0039,
+      "step": 14946
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1185974421434865e-05,
+      "loss": 1.3027,
+      "step": 14947
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1184945310881312e-05,
+      "loss": 1.0938,
+      "step": 14948
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1183916187599509e-05,
+      "loss": 0.9214,
+      "step": 14949
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1182887051600502e-05,
+      "loss": 0.8799,
+      "step": 14950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.118185790289535e-05,
+      "loss": 1.5938,
+      "step": 14951
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1180828741495108e-05,
+      "loss": 1.2588,
+      "step": 14952
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1179799567410829e-05,
+      "loss": 1.1338,
+      "step": 14953
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1178770380653567e-05,
+      "loss": 0.8057,
+      "step": 14954
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117774118123438e-05,
+      "loss": 0.7705,
+      "step": 14955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1176711969164322e-05,
+      "loss": 0.9854,
+      "step": 14956
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117568274445445e-05,
+      "loss": 1.4365,
+      "step": 14957
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117465350711582e-05,
+      "loss": 1.2471,
+      "step": 14958
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117362425715948e-05,
+      "loss": 1.251,
+      "step": 14959
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1172594994596497e-05,
+      "loss": 0.7432,
+      "step": 14960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1171565719437917e-05,
+      "loss": 0.9258,
+      "step": 14961
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1170536431694805e-05,
+      "loss": 0.8665,
+      "step": 14962
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1169507131378213e-05,
+      "loss": 1.2031,
+      "step": 14963
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1168477818499194e-05,
+      "loss": 1.2637,
+      "step": 14964
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116744849306881e-05,
+      "loss": 1.4434,
+      "step": 14965
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1166419155098113e-05,
+      "loss": 1.4355,
+      "step": 14966
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1165389804598165e-05,
+      "loss": 1.125,
+      "step": 14967
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116436044158002e-05,
+      "loss": 0.751,
+      "step": 14968
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1163331066054735e-05,
+      "loss": 1.1221,
+      "step": 14969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1162301678033367e-05,
+      "loss": 0.8711,
+      "step": 14970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1161272277526973e-05,
+      "loss": 1.4834,
+      "step": 14971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1160242864546614e-05,
+      "loss": 1.21,
+      "step": 14972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1159213439103342e-05,
+      "loss": 0.832,
+      "step": 14973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1158184001208217e-05,
+      "loss": 1.1309,
+      "step": 14974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11571545508723e-05,
+      "loss": 0.7549,
+      "step": 14975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1156125088106646e-05,
+      "loss": 0.7627,
+      "step": 14976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1155095612922311e-05,
+      "loss": 0.8564,
+      "step": 14977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1154066125330358e-05,
+      "loss": 1.4297,
+      "step": 14978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115303662534184e-05,
+      "loss": 0.9492,
+      "step": 14979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1152007112967822e-05,
+      "loss": 1.0547,
+      "step": 14980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1150977588219358e-05,
+      "loss": 1.0156,
+      "step": 14981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114994805110751e-05,
+      "loss": 1.167,
+      "step": 14982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114891850164333e-05,
+      "loss": 1.1787,
+      "step": 14983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1147888939837886e-05,
+      "loss": 0.7686,
+      "step": 14984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1146859365702232e-05,
+      "loss": 0.9189,
+      "step": 14985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114582977924743e-05,
+      "loss": 1.4639,
+      "step": 14986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1144800180484536e-05,
+      "loss": 1.1631,
+      "step": 14987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1143770569424612e-05,
+      "loss": 1.3018,
+      "step": 14988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114274094607872e-05,
+      "loss": 1.0498,
+      "step": 14989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114171131045791e-05,
+      "loss": 1.3223,
+      "step": 14990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1140681662573255e-05,
+      "loss": 1.1729,
+      "step": 14991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1139652002435806e-05,
+      "loss": 0.6538,
+      "step": 14992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1138622330056629e-05,
+      "loss": 1.0659,
+      "step": 14993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1137592645446781e-05,
+      "loss": 0.8154,
+      "step": 14994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1136562948617324e-05,
+      "loss": 0.9297,
+      "step": 14995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1135533239579318e-05,
+      "loss": 0.792,
+      "step": 14996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1134503518343823e-05,
+      "loss": 0.915,
+      "step": 14997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1133473784921897e-05,
+      "loss": 0.7573,
+      "step": 14998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1132444039324608e-05,
+      "loss": 0.9863,
+      "step": 14999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1131414281563013e-05,
+      "loss": 0.9668,
+      "step": 15000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1130384511648175e-05,
+      "loss": 0.8682,
+      "step": 15001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1129354729591158e-05,
+      "loss": 0.8555,
+      "step": 15002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1128324935403013e-05,
+      "loss": 1.0127,
+      "step": 15003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1127295129094813e-05,
+      "loss": 0.6387,
+      "step": 15004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112626531067761e-05,
+      "loss": 1.2188,
+      "step": 15005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112523548016248e-05,
+      "loss": 0.5586,
+      "step": 15006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1124205637560471e-05,
+      "loss": 1.043,
+      "step": 15007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112317578288265e-05,
+      "loss": 1.165,
+      "step": 15008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1122145916140081e-05,
+      "loss": 0.999,
+      "step": 15009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1121116037343826e-05,
+      "loss": 0.8184,
+      "step": 15010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1120086146504947e-05,
+      "loss": 1.0283,
+      "step": 15011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1119056243634507e-05,
+      "loss": 1.0493,
+      "step": 15012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1118026328743567e-05,
+      "loss": 0.8789,
+      "step": 15013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1116996401843194e-05,
+      "loss": 1.043,
+      "step": 15014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1115966462944448e-05,
+      "loss": 0.9375,
+      "step": 15015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1114936512058392e-05,
+      "loss": 1.3018,
+      "step": 15016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1113906549196092e-05,
+      "loss": 1.2363,
+      "step": 15017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1112876574368606e-05,
+      "loss": 1.0645,
+      "step": 15018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1111846587587005e-05,
+      "loss": 1.332,
+      "step": 15019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1110816588862349e-05,
+      "loss": 1.0615,
+      "step": 15020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1109786578205701e-05,
+      "loss": 1.1055,
+      "step": 15021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1108756555628127e-05,
+      "loss": 0.9824,
+      "step": 15022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1107726521140685e-05,
+      "loss": 1.1113,
+      "step": 15023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1106696474754451e-05,
+      "loss": 1.1709,
+      "step": 15024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.110566641648048e-05,
+      "loss": 0.959,
+      "step": 15025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1104636346329838e-05,
+      "loss": 0.9014,
+      "step": 15026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1103606264313593e-05,
+      "loss": 1.2871,
+      "step": 15027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1102576170442808e-05,
+      "loss": 1.4902,
+      "step": 15028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1101546064728546e-05,
+      "loss": 1.1836,
+      "step": 15029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1100515947181875e-05,
+      "loss": 1.1211,
+      "step": 15030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1099485817813858e-05,
+      "loss": 0.9766,
+      "step": 15031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109845567663556e-05,
+      "loss": 0.7742,
+      "step": 15032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109742552365805e-05,
+      "loss": 1.0508,
+      "step": 15033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1096395358892389e-05,
+      "loss": 0.9053,
+      "step": 15034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1095365182349647e-05,
+      "loss": 0.7429,
+      "step": 15035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1094334994040883e-05,
+      "loss": 0.8496,
+      "step": 15036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109330479397717e-05,
+      "loss": 1.1875,
+      "step": 15037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1092274582169573e-05,
+      "loss": 0.8291,
+      "step": 15038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1091244358629154e-05,
+      "loss": 0.9561,
+      "step": 15039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1090214123366985e-05,
+      "loss": 1.4727,
+      "step": 15040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1089183876394125e-05,
+      "loss": 0.7866,
+      "step": 15041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1088153617721646e-05,
+      "loss": 1.0977,
+      "step": 15042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1087123347360616e-05,
+      "loss": 1.2852,
+      "step": 15043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1086093065322096e-05,
+      "loss": 1.1309,
+      "step": 15044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1085062771617158e-05,
+      "loss": 1.1279,
+      "step": 15045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1084032466256867e-05,
+      "loss": 0.6611,
+      "step": 15046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1083002149252292e-05,
+      "loss": 0.5784,
+      "step": 15047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1081971820614497e-05,
+      "loss": 0.9775,
+      "step": 15048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1080941480354547e-05,
+      "loss": 1.3535,
+      "step": 15049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107991112848352e-05,
+      "loss": 0.8892,
+      "step": 15050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1078880765012477e-05,
+      "loss": 1.0039,
+      "step": 15051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1077850389952482e-05,
+      "loss": 1.084,
+      "step": 15052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107682000331461e-05,
+      "loss": 0.6914,
+      "step": 15053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1075789605109926e-05,
+      "loss": 0.897,
+      "step": 15054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1074759195349497e-05,
+      "loss": 1.1338,
+      "step": 15055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1073728774044393e-05,
+      "loss": 1.0161,
+      "step": 15056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107269834120568e-05,
+      "loss": 1.0801,
+      "step": 15057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1071667896844432e-05,
+      "loss": 1.0337,
+      "step": 15058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107063744097171e-05,
+      "loss": 1.2539,
+      "step": 15059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106960697359859e-05,
+      "loss": 1.3711,
+      "step": 15060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106857649473614e-05,
+      "loss": 1.1289,
+      "step": 15061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106754600439542e-05,
+      "loss": 0.7666,
+      "step": 15062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1066515502587512e-05,
+      "loss": 1.2217,
+      "step": 15063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1065484989323477e-05,
+      "loss": 0.833,
+      "step": 15064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1064454464614387e-05,
+      "loss": 1.1191,
+      "step": 15065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106342392847131e-05,
+      "loss": 1.166,
+      "step": 15066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106239338090532e-05,
+      "loss": 0.6392,
+      "step": 15067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106136282192748e-05,
+      "loss": 1.2354,
+      "step": 15068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1060332251548864e-05,
+      "loss": 0.719,
+      "step": 15069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1059301669780543e-05,
+      "loss": 1.333,
+      "step": 15070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1058271076633584e-05,
+      "loss": 1.0566,
+      "step": 15071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1057240472119062e-05,
+      "loss": 1.1641,
+      "step": 15072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1056209856248041e-05,
+      "loss": 0.7578,
+      "step": 15073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1055179229031598e-05,
+      "loss": 1.0098,
+      "step": 15074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1054148590480797e-05,
+      "loss": 1.0723,
+      "step": 15075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1053117940606717e-05,
+      "loss": 1.1191,
+      "step": 15076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1052087279420418e-05,
+      "loss": 1.2959,
+      "step": 15077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1051056606932979e-05,
+      "loss": 1.2393,
+      "step": 15078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1050025923155472e-05,
+      "loss": 1.1758,
+      "step": 15079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1048995228098963e-05,
+      "loss": 1.3955,
+      "step": 15080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1047964521774528e-05,
+      "loss": 1.1182,
+      "step": 15081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1046933804193234e-05,
+      "loss": 1.1318,
+      "step": 15082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1045903075366156e-05,
+      "loss": 1.1143,
+      "step": 15083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1044872335304362e-05,
+      "loss": 1.1479,
+      "step": 15084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1043841584018929e-05,
+      "loss": 0.8286,
+      "step": 15085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1042810821520924e-05,
+      "loss": 0.9736,
+      "step": 15086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1041780047821424e-05,
+      "loss": 1.0029,
+      "step": 15087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1040749262931497e-05,
+      "loss": 1.1768,
+      "step": 15088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1039718466862217e-05,
+      "loss": 0.9453,
+      "step": 15089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1038687659624659e-05,
+      "loss": 0.9722,
+      "step": 15090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103765684122989e-05,
+      "loss": 0.8994,
+      "step": 15091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1036626011688987e-05,
+      "loss": 0.9209,
+      "step": 15092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1035595171013017e-05,
+      "loss": 1.1201,
+      "step": 15093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1034564319213066e-05,
+      "loss": 1.0156,
+      "step": 15094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1033533456300192e-05,
+      "loss": 0.8477,
+      "step": 15095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1032502582285476e-05,
+      "loss": 1.001,
+      "step": 15096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103147169717999e-05,
+      "loss": 1.0942,
+      "step": 15097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1030440800994806e-05,
+      "loss": 1.1299,
+      "step": 15098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1029409893741001e-05,
+      "loss": 1.0605,
+      "step": 15099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1028378975429646e-05,
+      "loss": 0.9937,
+      "step": 15100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1027348046071813e-05,
+      "loss": 1.1514,
+      "step": 15101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102631710567858e-05,
+      "loss": 0.9336,
+      "step": 15102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1025286154261016e-05,
+      "loss": 1.1533,
+      "step": 15103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10242551918302e-05,
+      "loss": 0.7754,
+      "step": 15104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1023224218397205e-05,
+      "loss": 1.3145,
+      "step": 15105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1022193233973104e-05,
+      "loss": 1.1855,
+      "step": 15106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1021162238568974e-05,
+      "loss": 1.0586,
+      "step": 15107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1020131232195883e-05,
+      "loss": 0.8389,
+      "step": 15108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1019100214864916e-05,
+      "loss": 1.0513,
+      "step": 15109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1018069186587139e-05,
+      "loss": 0.9883,
+      "step": 15110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1017038147373631e-05,
+      "loss": 0.6455,
+      "step": 15111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1016007097235466e-05,
+      "loss": 0.7661,
+      "step": 15112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1014976036183718e-05,
+      "loss": 1.1836,
+      "step": 15113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1013944964229464e-05,
+      "loss": 0.8901,
+      "step": 15114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101291388138378e-05,
+      "loss": 1.3398,
+      "step": 15115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101188278765774e-05,
+      "loss": 0.875,
+      "step": 15116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101085168306242e-05,
+      "loss": 0.834,
+      "step": 15117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10098205676089e-05,
+      "loss": 1.3623,
+      "step": 15118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1008789441308245e-05,
+      "loss": 1.2007,
+      "step": 15119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1007758304171542e-05,
+      "loss": 1.0498,
+      "step": 15120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.100672715620986e-05,
+      "loss": 0.7578,
+      "step": 15121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.100569599743428e-05,
+      "loss": 0.8184,
+      "step": 15122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.100466482785588e-05,
+      "loss": 1.248,
+      "step": 15123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1003633647485725e-05,
+      "loss": 0.7993,
+      "step": 15124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1002602456334908e-05,
+      "loss": 1.0332,
+      "step": 15125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1001571254414491e-05,
+      "loss": 1.1211,
+      "step": 15126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1000540041735558e-05,
+      "loss": 1.1689,
+      "step": 15127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0999508818309188e-05,
+      "loss": 0.9668,
+      "step": 15128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0998477584146452e-05,
+      "loss": 0.8555,
+      "step": 15129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099744633925843e-05,
+      "loss": 1.2881,
+      "step": 15130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.09964150836562e-05,
+      "loss": 1.4658,
+      "step": 15131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099538381735084e-05,
+      "loss": 1.0723,
+      "step": 15132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0994352540353423e-05,
+      "loss": 0.7939,
+      "step": 15133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0993321252675033e-05,
+      "loss": 0.6299,
+      "step": 15134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0992289954326742e-05,
+      "loss": 1.0371,
+      "step": 15135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0991258645319636e-05,
+      "loss": 0.9463,
+      "step": 15136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0990227325664779e-05,
+      "loss": 0.8857,
+      "step": 15137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0989195995373266e-05,
+      "loss": 1.1152,
+      "step": 15138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0988164654456157e-05,
+      "loss": 0.7842,
+      "step": 15139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0987133302924546e-05,
+      "loss": 0.9346,
+      "step": 15140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0986101940789504e-05,
+      "loss": 0.8506,
+      "step": 15141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0985070568062111e-05,
+      "loss": 0.9072,
+      "step": 15142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0984039184753446e-05,
+      "loss": 1.1455,
+      "step": 15143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0983007790874585e-05,
+      "loss": 0.8345,
+      "step": 15144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.098197638643661e-05,
+      "loss": 0.77,
+      "step": 15145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.09809449714506e-05,
+      "loss": 1.376,
+      "step": 15146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0979913545927632e-05,
+      "loss": 0.8584,
+      "step": 15147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0978882109878786e-05,
+      "loss": 1.4668,
+      "step": 15148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0977850663315144e-05,
+      "loss": 0.9072,
+      "step": 15149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0976819206247778e-05,
+      "loss": 0.9873,
+      "step": 15150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0975787738687777e-05,
+      "loss": 1.0889,
+      "step": 15151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0974756260646214e-05,
+      "loss": 0.9648,
+      "step": 15152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0973724772134173e-05,
+      "loss": 1.4385,
+      "step": 15153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0972693273162728e-05,
+      "loss": 1.2812,
+      "step": 15154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0971661763742966e-05,
+      "loss": 1.124,
+      "step": 15155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0970630243885964e-05,
+      "loss": 0.9404,
+      "step": 15156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09695987136028e-05,
+      "loss": 0.7705,
+      "step": 15157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0968567172904558e-05,
+      "loss": 1.0654,
+      "step": 15158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0967535621802316e-05,
+      "loss": 1.1416,
+      "step": 15159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0966504060307157e-05,
+      "loss": 1.0371,
+      "step": 15160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0965472488430159e-05,
+      "loss": 1.0928,
+      "step": 15161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0964440906182406e-05,
+      "loss": 0.9097,
+      "step": 15162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0963409313574974e-05,
+      "loss": 1.2539,
+      "step": 15163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0962377710618947e-05,
+      "loss": 1.2012,
+      "step": 15164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0961346097325406e-05,
+      "loss": 0.8018,
+      "step": 15165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0960314473705433e-05,
+      "loss": 0.8281,
+      "step": 15166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0959282839770108e-05,
+      "loss": 1.0107,
+      "step": 15167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0958251195530511e-05,
+      "loss": 1.1904,
+      "step": 15168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.095721954099773e-05,
+      "loss": 1.5342,
+      "step": 15169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0956187876182837e-05,
+      "loss": 1.3516,
+      "step": 15170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0955156201096922e-05,
+      "loss": 1.2217,
+      "step": 15171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0954124515751063e-05,
+      "loss": 0.7539,
+      "step": 15172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0953092820156343e-05,
+      "loss": 0.7627,
+      "step": 15173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0952061114323842e-05,
+      "loss": 1.0889,
+      "step": 15174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0951029398264648e-05,
+      "loss": 1.4385,
+      "step": 15175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0949997671989833e-05,
+      "loss": 0.9209,
+      "step": 15176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0948965935510491e-05,
+      "loss": 1.1426,
+      "step": 15177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0947934188837697e-05,
+      "loss": 1.1914,
+      "step": 15178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0946902431982538e-05,
+      "loss": 0.9854,
+      "step": 15179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0945870664956096e-05,
+      "loss": 1.1348,
+      "step": 15180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0944838887769446e-05,
+      "loss": 0.4844,
+      "step": 15181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0943807100433683e-05,
+      "loss": 0.8604,
+      "step": 15182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0942775302959883e-05,
+      "loss": 0.9912,
+      "step": 15183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0941743495359131e-05,
+      "loss": 1.2344,
+      "step": 15184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0940711677642511e-05,
+      "loss": 1.1938,
+      "step": 15185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0939679849821106e-05,
+      "loss": 1.1318,
+      "step": 15186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0938648011905997e-05,
+      "loss": 1.1514,
+      "step": 15187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.093761616390827e-05,
+      "loss": 0.9023,
+      "step": 15188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0936584305839009e-05,
+      "loss": 1.3379,
+      "step": 15189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0935552437709298e-05,
+      "loss": 0.9365,
+      "step": 15190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0934520559530219e-05,
+      "loss": 0.918,
+      "step": 15191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0933488671312859e-05,
+      "loss": 1.2158,
+      "step": 15192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0932456773068301e-05,
+      "loss": 1.0283,
+      "step": 15193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0931424864807624e-05,
+      "loss": 1.2031,
+      "step": 15194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0930392946541923e-05,
+      "loss": 1.2109,
+      "step": 15195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0929361018282272e-05,
+      "loss": 1.0039,
+      "step": 15196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0928329080039766e-05,
+      "loss": 0.8135,
+      "step": 15197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092729713182548e-05,
+      "loss": 1.3652,
+      "step": 15198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0926265173650504e-05,
+      "loss": 1.1279,
+      "step": 15199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092523320552592e-05,
+      "loss": 0.8501,
+      "step": 15200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0924201227462816e-05,
+      "loss": 1.0859,
+      "step": 15201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0923169239472275e-05,
+      "loss": 1.0742,
+      "step": 15202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0922137241565383e-05,
+      "loss": 1.2529,
+      "step": 15203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0921105233753226e-05,
+      "loss": 1.0264,
+      "step": 15204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0920073216046888e-05,
+      "loss": 1.1045,
+      "step": 15205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.091904118845746e-05,
+      "loss": 1.2354,
+      "step": 15206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0918009150996016e-05,
+      "loss": 1.1816,
+      "step": 15207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0916977103673656e-05,
+      "loss": 0.9902,
+      "step": 15208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0915945046501453e-05,
+      "loss": 1.2188,
+      "step": 15209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0914912979490503e-05,
+      "loss": 1.0044,
+      "step": 15210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0913880902651886e-05,
+      "loss": 1.6191,
+      "step": 15211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.091284881599669e-05,
+      "loss": 0.9033,
+      "step": 15212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0911816719536001e-05,
+      "loss": 1.0972,
+      "step": 15213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0910784613280906e-05,
+      "loss": 0.8838,
+      "step": 15214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0909752497242491e-05,
+      "loss": 1.0732,
+      "step": 15215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0908720371431844e-05,
+      "loss": 0.9551,
+      "step": 15216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090768823586005e-05,
+      "loss": 1.0811,
+      "step": 15217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0906656090538196e-05,
+      "loss": 0.9424,
+      "step": 15218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0905623935477374e-05,
+      "loss": 0.9277,
+      "step": 15219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0904591770688659e-05,
+      "loss": 0.9487,
+      "step": 15220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0903559596183154e-05,
+      "loss": 1.019,
+      "step": 15221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0902527411971928e-05,
+      "loss": 0.9785,
+      "step": 15222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0901495218066085e-05,
+      "loss": 1.0645,
+      "step": 15223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0900463014476705e-05,
+      "loss": 0.9238,
+      "step": 15224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0899430801214872e-05,
+      "loss": 1.1562,
+      "step": 15225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0898398578291682e-05,
+      "loss": 0.6021,
+      "step": 15226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0897366345718217e-05,
+      "loss": 1.0215,
+      "step": 15227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0896334103505568e-05,
+      "loss": 1.1094,
+      "step": 15228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0895301851664821e-05,
+      "loss": 0.9014,
+      "step": 15229
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0894269590207062e-05,
+      "loss": 1.2314,
+      "step": 15230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0893237319143384e-05,
+      "loss": 0.917,
+      "step": 15231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0892205038484873e-05,
+      "loss": 1.0098,
+      "step": 15232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0891172748242617e-05,
+      "loss": 1.355,
+      "step": 15233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0890140448427705e-05,
+      "loss": 1.1426,
+      "step": 15234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0889108139051225e-05,
+      "loss": 0.8521,
+      "step": 15235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0888075820124266e-05,
+      "loss": 1.0586,
+      "step": 15236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088704349165792e-05,
+      "loss": 1.4414,
+      "step": 15237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088601115366327e-05,
+      "loss": 0.8274,
+      "step": 15238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088497880615141e-05,
+      "loss": 1.1699,
+      "step": 15239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0883946449133422e-05,
+      "loss": 0.7734,
+      "step": 15240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0882914082620406e-05,
+      "loss": 0.7285,
+      "step": 15241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0881881706623443e-05,
+      "loss": 1.1055,
+      "step": 15242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0880849321153624e-05,
+      "loss": 1.4355,
+      "step": 15243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087981692622204e-05,
+      "loss": 1.2686,
+      "step": 15244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087878452183978e-05,
+      "loss": 1.2188,
+      "step": 15245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0877752108017934e-05,
+      "loss": 1.3105,
+      "step": 15246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0876719684767592e-05,
+      "loss": 1.4365,
+      "step": 15247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0875687252099842e-05,
+      "loss": 1.2432,
+      "step": 15248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0874654810025775e-05,
+      "loss": 0.8877,
+      "step": 15249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0873622358556486e-05,
+      "loss": 0.9551,
+      "step": 15250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0872589897703056e-05,
+      "loss": 0.9854,
+      "step": 15251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0871557427476585e-05,
+      "loss": 1.021,
+      "step": 15252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0870524947888152e-05,
+      "loss": 0.8711,
+      "step": 15253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0869492458948861e-05,
+      "loss": 0.7861,
+      "step": 15254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0868459960669792e-05,
+      "loss": 0.9868,
+      "step": 15255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086742745306204e-05,
+      "loss": 1.0488,
+      "step": 15256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0866394936136695e-05,
+      "loss": 0.9707,
+      "step": 15257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0865362409904848e-05,
+      "loss": 1.1611,
+      "step": 15258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0864329874377591e-05,
+      "loss": 1.2363,
+      "step": 15259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0863297329566013e-05,
+      "loss": 1.1426,
+      "step": 15260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0862264775481209e-05,
+      "loss": 1.0449,
+      "step": 15261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0861232212134265e-05,
+      "loss": 1.1221,
+      "step": 15262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086019963953628e-05,
+      "loss": 0.8164,
+      "step": 15263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0859167057698335e-05,
+      "loss": 1.1992,
+      "step": 15264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0858134466631531e-05,
+      "loss": 1.1885,
+      "step": 15265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0857101866346953e-05,
+      "loss": 0.9746,
+      "step": 15266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0856069256855699e-05,
+      "loss": 1.0474,
+      "step": 15267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0855036638168858e-05,
+      "loss": 0.918,
+      "step": 15268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0854004010297521e-05,
+      "loss": 1.0156,
+      "step": 15269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0852971373252783e-05,
+      "loss": 1.1973,
+      "step": 15270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085193872704573e-05,
+      "loss": 1.1973,
+      "step": 15271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085090607168746e-05,
+      "loss": 1.1396,
+      "step": 15272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0849873407189067e-05,
+      "loss": 1.1387,
+      "step": 15273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0848840733561638e-05,
+      "loss": 1.04,
+      "step": 15274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.084780805081627e-05,
+      "loss": 1.3262,
+      "step": 15275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0846775358964053e-05,
+      "loss": 0.8857,
+      "step": 15276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0845742658016078e-05,
+      "loss": 0.8882,
+      "step": 15277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0844709947983443e-05,
+      "loss": 1.0518,
+      "step": 15278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0843677228877238e-05,
+      "loss": 1.0664,
+      "step": 15279
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0842644500708557e-05,
+      "loss": 1.0938,
+      "step": 15280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0841611763488493e-05,
+      "loss": 0.8125,
+      "step": 15281
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0840579017228137e-05,
+      "loss": 0.9575,
+      "step": 15282
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0839546261938589e-05,
+      "loss": 0.9482,
+      "step": 15283
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0838513497630932e-05,
+      "loss": 0.6787,
+      "step": 15284
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0837480724316272e-05,
+      "loss": 0.6968,
+      "step": 15285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0836447942005692e-05,
+      "loss": 0.9297,
+      "step": 15286
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0835415150710291e-05,
+      "loss": 1.1328,
+      "step": 15287
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0834382350441162e-05,
+      "loss": 1.2598,
+      "step": 15288
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0833349541209397e-05,
+      "loss": 0.6694,
+      "step": 15289
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0832316723026092e-05,
+      "loss": 1.2266,
+      "step": 15290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0831283895902345e-05,
+      "loss": 0.8867,
+      "step": 15291
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0830251059849244e-05,
+      "loss": 1.0215,
+      "step": 15292
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0829218214877883e-05,
+      "loss": 0.7051,
+      "step": 15293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0828185360999365e-05,
+      "loss": 0.834,
+      "step": 15294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0827152498224773e-05,
+      "loss": 1.0459,
+      "step": 15295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0826119626565213e-05,
+      "loss": 0.8545,
+      "step": 15296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0825086746031767e-05,
+      "loss": 1.1406,
+      "step": 15297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0824053856635543e-05,
+      "loss": 0.9116,
+      "step": 15298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0823020958387626e-05,
+      "loss": 1.1719,
+      "step": 15299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0821988051299117e-05,
+      "loss": 1.1523,
+      "step": 15300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0820955135381109e-05,
+      "loss": 0.9922,
+      "step": 15301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0819922210644695e-05,
+      "loss": 0.7529,
+      "step": 15302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0818889277100975e-05,
+      "loss": 0.9404,
+      "step": 15303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081785633476104e-05,
+      "loss": 1.126,
+      "step": 15304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0816823383635988e-05,
+      "loss": 1.0469,
+      "step": 15305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0815790423736914e-05,
+      "loss": 0.793,
+      "step": 15306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0814757455074911e-05,
+      "loss": 1.2402,
+      "step": 15307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0813724477661081e-05,
+      "loss": 0.6313,
+      "step": 15308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0812691491506517e-05,
+      "loss": 1.0625,
+      "step": 15309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081165849662231e-05,
+      "loss": 1.3408,
+      "step": 15310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0810625493019563e-05,
+      "loss": 0.96,
+      "step": 15311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080959248070937e-05,
+      "loss": 0.9778,
+      "step": 15312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0808559459702824e-05,
+      "loss": 1.041,
+      "step": 15313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0807526430011028e-05,
+      "loss": 0.666,
+      "step": 15314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080649339164507e-05,
+      "loss": 1.0693,
+      "step": 15315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0805460344616055e-05,
+      "loss": 1.1226,
+      "step": 15316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0804427288935073e-05,
+      "loss": 1.1572,
+      "step": 15317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0803394224613225e-05,
+      "loss": 1.2881,
+      "step": 15318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0802361151661605e-05,
+      "loss": 0.8848,
+      "step": 15319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080132807009131e-05,
+      "loss": 0.9209,
+      "step": 15320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0800294979913439e-05,
+      "loss": 1.0303,
+      "step": 15321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.079926188113909e-05,
+      "loss": 1.2285,
+      "step": 15322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0798228773779355e-05,
+      "loss": 0.9258,
+      "step": 15323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0797195657845337e-05,
+      "loss": 0.8809,
+      "step": 15324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0796162533348132e-05,
+      "loss": 1.001,
+      "step": 15325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0795129400298832e-05,
+      "loss": 1.127,
+      "step": 15326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0794096258708543e-05,
+      "loss": 1.291,
+      "step": 15327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0793063108588354e-05,
+      "loss": 0.562,
+      "step": 15328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0792029949949372e-05,
+      "loss": 1.1387,
+      "step": 15329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0790996782802687e-05,
+      "loss": 1.1113,
+      "step": 15330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0789963607159402e-05,
+      "loss": 1.2256,
+      "step": 15331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.078893042303061e-05,
+      "loss": 1.2627,
+      "step": 15332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0787897230427413e-05,
+      "loss": 1.3203,
+      "step": 15333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0786864029360911e-05,
+      "loss": 0.9717,
+      "step": 15334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0785830819842195e-05,
+      "loss": 1.4004,
+      "step": 15335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0784797601882373e-05,
+      "loss": 1.2041,
+      "step": 15336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0783764375492534e-05,
+      "loss": 0.7114,
+      "step": 15337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0782731140683786e-05,
+      "loss": 0.9268,
+      "step": 15338
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0781697897467216e-05,
+      "loss": 1.3145,
+      "step": 15339
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0780664645853935e-05,
+      "loss": 1.2852,
+      "step": 15340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0779631385855031e-05,
+      "loss": 1.0244,
+      "step": 15341
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0778598117481611e-05,
+      "loss": 1.1016,
+      "step": 15342
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0777564840744773e-05,
+      "loss": 1.041,
+      "step": 15343
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077653155565561e-05,
+      "loss": 1.1387,
+      "step": 15344
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0775498262225225e-05,
+      "loss": 0.991,
+      "step": 15345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0774464960464721e-05,
+      "loss": 0.6738,
+      "step": 15346
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0773431650385191e-05,
+      "loss": 1.2061,
+      "step": 15347
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0772398331997738e-05,
+      "loss": 0.7607,
+      "step": 15348
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077136500531346e-05,
+      "loss": 1.3066,
+      "step": 15349
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0770331670343459e-05,
+      "loss": 1.2061,
+      "step": 15350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0769298327098832e-05,
+      "loss": 1.2285,
+      "step": 15351
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0768264975590681e-05,
+      "loss": 0.6499,
+      "step": 15352
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0767231615830104e-05,
+      "loss": 0.9131,
+      "step": 15353
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0766198247828201e-05,
+      "loss": 1.1738,
+      "step": 15354
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0765164871596077e-05,
+      "loss": 1.1572,
+      "step": 15355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0764131487144824e-05,
+      "loss": 1.0957,
+      "step": 15356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0763098094485546e-05,
+      "loss": 1.0024,
+      "step": 15357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0762064693629346e-05,
+      "loss": 0.9038,
+      "step": 15358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076103128458732e-05,
+      "loss": 0.9092,
+      "step": 15359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0759997867370573e-05,
+      "loss": 1.375,
+      "step": 15360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0758964441990201e-05,
+      "loss": 1.1523,
+      "step": 15361
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0757931008457308e-05,
+      "loss": 1.207,
+      "step": 15362
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0756897566782991e-05,
+      "loss": 1.0098,
+      "step": 15363
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0755864116978357e-05,
+      "loss": 1.0029,
+      "step": 15364
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0754830659054501e-05,
+      "loss": 1.1279,
+      "step": 15365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0753797193022526e-05,
+      "loss": 1.1301,
+      "step": 15366
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0752763718893536e-05,
+      "loss": 1.1963,
+      "step": 15367
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0751730236678628e-05,
+      "loss": 0.6602,
+      "step": 15368
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0750696746388907e-05,
+      "loss": 1.1172,
+      "step": 15369
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0749663248035469e-05,
+      "loss": 0.8374,
+      "step": 15370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0748629741629421e-05,
+      "loss": 1.2734,
+      "step": 15371
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074759622718186e-05,
+      "loss": 0.8926,
+      "step": 15372
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0746562704703896e-05,
+      "loss": 1.0449,
+      "step": 15373
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0745529174206618e-05,
+      "loss": 0.9424,
+      "step": 15374
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074449563570114e-05,
+      "loss": 0.9619,
+      "step": 15375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0743462089198555e-05,
+      "loss": 1.061,
+      "step": 15376
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074242853470997e-05,
+      "loss": 0.8345,
+      "step": 15377
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0741394972246484e-05,
+      "loss": 1.0083,
+      "step": 15378
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0740361401819201e-05,
+      "loss": 0.9746,
+      "step": 15379
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0739327823439225e-05,
+      "loss": 1.0215,
+      "step": 15380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0738294237117655e-05,
+      "loss": 1.127,
+      "step": 15381
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0737260642865595e-05,
+      "loss": 0.9077,
+      "step": 15382
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0736227040694147e-05,
+      "loss": 1.2324,
+      "step": 15383
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0735193430614414e-05,
+      "loss": 1.0742,
+      "step": 15384
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0734159812637496e-05,
+      "loss": 1.002,
+      "step": 15385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0733126186774504e-05,
+      "loss": 1.6182,
+      "step": 15386
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073209255303653e-05,
+      "loss": 1.0945,
+      "step": 15387
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0731058911434685e-05,
+      "loss": 0.8467,
+      "step": 15388
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0730025261980066e-05,
+      "loss": 1.1445,
+      "step": 15389
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072899160468378e-05,
+      "loss": 1.1006,
+      "step": 15390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072795793955693e-05,
+      "loss": 1.0547,
+      "step": 15391
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0726924266610616e-05,
+      "loss": 1.2949,
+      "step": 15392
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0725890585855946e-05,
+      "loss": 0.8853,
+      "step": 15393
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0724856897304022e-05,
+      "loss": 1.1475,
+      "step": 15394
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0723823200965946e-05,
+      "loss": 1.0371,
+      "step": 15395
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0722789496852822e-05,
+      "loss": 1.5176,
+      "step": 15396
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0721755784975756e-05,
+      "loss": 1.0312,
+      "step": 15397
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0720722065345845e-05,
+      "loss": 1.0811,
+      "step": 15398
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0719688337974203e-05,
+      "loss": 0.9248,
+      "step": 15399
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0718654602871925e-05,
+      "loss": 1.1123,
+      "step": 15400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071762086005012e-05,
+      "loss": 1.0229,
+      "step": 15401
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0716587109519889e-05,
+      "loss": 1.0836,
+      "step": 15402
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0715553351292338e-05,
+      "loss": 1.248,
+      "step": 15403
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0714519585378574e-05,
+      "loss": 0.8633,
+      "step": 15404
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0713485811789696e-05,
+      "loss": 0.8921,
+      "step": 15405
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071245203053681e-05,
+      "loss": 0.9097,
+      "step": 15406
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0711418241631024e-05,
+      "loss": 0.8955,
+      "step": 15407
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071038444508344e-05,
+      "loss": 0.8589,
+      "step": 15408
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0709350640905161e-05,
+      "loss": 0.833,
+      "step": 15409
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0708316829107295e-05,
+      "loss": 1.0229,
+      "step": 15410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0707283009700941e-05,
+      "loss": 0.7324,
+      "step": 15411
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.070624918269721e-05,
+      "loss": 1.0864,
+      "step": 15412
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0705215348107205e-05,
+      "loss": 1.3848,
+      "step": 15413
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0704181505942034e-05,
+      "loss": 1.2461,
+      "step": 15414
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0703147656212799e-05,
+      "loss": 1.1055,
+      "step": 15415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0702113798930603e-05,
+      "loss": 0.8984,
+      "step": 15416
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0701079934106554e-05,
+      "loss": 1.2178,
+      "step": 15417
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0700046061751758e-05,
+      "loss": 0.9961,
+      "step": 15418
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0699012181877322e-05,
+      "loss": 1.2451,
+      "step": 15419
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0697978294494346e-05,
+      "loss": 1.418,
+      "step": 15420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0696944399613942e-05,
+      "loss": 1.2012,
+      "step": 15421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069591049724721e-05,
+      "loss": 1.1299,
+      "step": 15422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0694876587405261e-05,
+      "loss": 0.9761,
+      "step": 15423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0693842670099195e-05,
+      "loss": 1.0078,
+      "step": 15424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0692808745340124e-05,
+      "loss": 0.9697,
+      "step": 15425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0691774813139151e-05,
+      "loss": 0.7222,
+      "step": 15426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069074087350738e-05,
+      "loss": 0.876,
+      "step": 15427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0689706926455927e-05,
+      "loss": 0.9639,
+      "step": 15428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0688672971995884e-05,
+      "loss": 0.9902,
+      "step": 15429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0687639010138369e-05,
+      "loss": 1.1611,
+      "step": 15430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.068660504089448e-05,
+      "loss": 0.7075,
+      "step": 15431
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0685571064275329e-05,
+      "loss": 0.7959,
+      "step": 15432
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0684537080292022e-05,
+      "loss": 1.0771,
+      "step": 15433
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0683503088955663e-05,
+      "loss": 1.291,
+      "step": 15434
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0682469090277361e-05,
+      "loss": 1.2354,
+      "step": 15435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0681435084268221e-05,
+      "loss": 0.8843,
+      "step": 15436
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0680401070939353e-05,
+      "loss": 1.0088,
+      "step": 15437
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.067936705030186e-05,
+      "loss": 1.416,
+      "step": 15438
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0678333022366853e-05,
+      "loss": 0.6929,
+      "step": 15439
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0677298987145439e-05,
+      "loss": 1.1182,
+      "step": 15440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0676264944648724e-05,
+      "loss": 0.9385,
+      "step": 15441
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0675230894887811e-05,
+      "loss": 0.9463,
+      "step": 15442
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0674196837873817e-05,
+      "loss": 0.7061,
+      "step": 15443
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0673162773617836e-05,
+      "loss": 0.6787,
+      "step": 15444
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0672128702130988e-05,
+      "loss": 0.6494,
+      "step": 15445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0671094623424376e-05,
+      "loss": 1.1455,
+      "step": 15446
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0670060537509107e-05,
+      "loss": 0.9033,
+      "step": 15447
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0669026444396292e-05,
+      "loss": 0.958,
+      "step": 15448
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0667992344097031e-05,
+      "loss": 1.1328,
+      "step": 15449
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0666958236622442e-05,
+      "loss": 0.749,
+      "step": 15450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0665924121983627e-05,
+      "loss": 1.0503,
+      "step": 15451
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0664890000191695e-05,
+      "loss": 0.582,
+      "step": 15452
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0663855871257754e-05,
+      "loss": 0.9316,
+      "step": 15453
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0662821735192914e-05,
+      "loss": 0.623,
+      "step": 15454
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0661787592008279e-05,
+      "loss": 1.0742,
+      "step": 15455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0660753441714964e-05,
+      "loss": 0.876,
+      "step": 15456
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065971928432407e-05,
+      "loss": 0.8652,
+      "step": 15457
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0658685119846713e-05,
+      "loss": 1.3252,
+      "step": 15458
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0657650948293997e-05,
+      "loss": 0.8057,
+      "step": 15459
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0656616769677032e-05,
+      "loss": 0.9775,
+      "step": 15460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065558258400693e-05,
+      "loss": 0.8887,
+      "step": 15461
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0654548391294792e-05,
+      "loss": 1.2666,
+      "step": 15462
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0653514191551733e-05,
+      "loss": 1.2246,
+      "step": 15463
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0652479984788857e-05,
+      "loss": 0.8086,
+      "step": 15464
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0651445771017279e-05,
+      "loss": 0.7207,
+      "step": 15465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0650411550248107e-05,
+      "loss": 1.0107,
+      "step": 15466
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0649377322492447e-05,
+      "loss": 0.8823,
+      "step": 15467
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0648343087761411e-05,
+      "loss": 1.1699,
+      "step": 15468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0647308846066107e-05,
+      "loss": 0.9775,
+      "step": 15469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0646274597417646e-05,
+      "loss": 0.9365,
+      "step": 15470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0645240341827134e-05,
+      "loss": 1.0176,
+      "step": 15471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0644206079305686e-05,
+      "loss": 1.0605,
+      "step": 15472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0643171809864405e-05,
+      "loss": 1.1846,
+      "step": 15473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0642137533514409e-05,
+      "loss": 0.8257,
+      "step": 15474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0641103250266798e-05,
+      "loss": 1.0908,
+      "step": 15475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0640068960132692e-05,
+      "loss": 1.0176,
+      "step": 15476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0639034663123196e-05,
+      "loss": 1.46,
+      "step": 15477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0638000359249418e-05,
+      "loss": 0.9453,
+      "step": 15478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063696604852247e-05,
+      "loss": 0.9209,
+      "step": 15479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0635931730953464e-05,
+      "loss": 1.0938,
+      "step": 15480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0634897406553508e-05,
+      "loss": 1.0098,
+      "step": 15481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0633863075333713e-05,
+      "loss": 0.9603,
+      "step": 15482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063282873730519e-05,
+      "loss": 1.0518,
+      "step": 15483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063179439247905e-05,
+      "loss": 0.7544,
+      "step": 15484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0630760040866404e-05,
+      "loss": 1.2529,
+      "step": 15485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0629725682478355e-05,
+      "loss": 0.7275,
+      "step": 15486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0628691317326026e-05,
+      "loss": 1.2812,
+      "step": 15487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0627656945420516e-05,
+      "loss": 0.7793,
+      "step": 15488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0626622566772948e-05,
+      "loss": 1.1777,
+      "step": 15489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0625588181394424e-05,
+      "loss": 1.5195,
+      "step": 15490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0624553789296057e-05,
+      "loss": 1.0869,
+      "step": 15491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.062351939048896e-05,
+      "loss": 1.1396,
+      "step": 15492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0622484984984238e-05,
+      "loss": 1.1602,
+      "step": 15493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0621450572793011e-05,
+      "loss": 1.0391,
+      "step": 15494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0620416153926384e-05,
+      "loss": 1.4004,
+      "step": 15495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0619381728395473e-05,
+      "loss": 0.8877,
+      "step": 15496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0618347296211384e-05,
+      "loss": 1.3008,
+      "step": 15497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0617312857385235e-05,
+      "loss": 0.9702,
+      "step": 15498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0616278411928131e-05,
+      "loss": 1.0088,
+      "step": 15499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0615243959851189e-05,
+      "loss": 0.7949,
+      "step": 15500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0614209501165516e-05,
+      "loss": 1.2627,
+      "step": 15501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0613175035882228e-05,
+      "loss": 0.8506,
+      "step": 15502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0612140564012437e-05,
+      "loss": 0.9873,
+      "step": 15503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0611106085567247e-05,
+      "loss": 1.0669,
+      "step": 15504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.061007160055778e-05,
+      "loss": 0.9648,
+      "step": 15505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0609037108995143e-05,
+      "loss": 1.207,
+      "step": 15506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0608002610890451e-05,
+      "loss": 1.2104,
+      "step": 15507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060696810625481e-05,
+      "loss": 0.9854,
+      "step": 15508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060593359509934e-05,
+      "loss": 0.9062,
+      "step": 15509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0604899077435149e-05,
+      "loss": 1.04,
+      "step": 15510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0603864553273351e-05,
+      "loss": 0.8672,
+      "step": 15511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0602830022625056e-05,
+      "loss": 1.0781,
+      "step": 15512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0601795485501381e-05,
+      "loss": 0.915,
+      "step": 15513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0600760941913435e-05,
+      "loss": 0.8042,
+      "step": 15514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0599726391872331e-05,
+      "loss": 1.0029,
+      "step": 15515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0598691835389186e-05,
+      "loss": 0.9658,
+      "step": 15516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0597657272475104e-05,
+      "loss": 1.0156,
+      "step": 15517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0596622703141209e-05,
+      "loss": 1.2871,
+      "step": 15518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0595588127398601e-05,
+      "loss": 0.959,
+      "step": 15519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0594553545258407e-05,
+      "loss": 1.1416,
+      "step": 15520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.059351895673173e-05,
+      "loss": 0.8223,
+      "step": 15521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0592484361829687e-05,
+      "loss": 0.7163,
+      "step": 15522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.059144976056339e-05,
+      "loss": 1.2251,
+      "step": 15523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0590415152943954e-05,
+      "loss": 1.1094,
+      "step": 15524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058938053898249e-05,
+      "loss": 1.1299,
+      "step": 15525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0588345918690113e-05,
+      "loss": 0.9595,
+      "step": 15526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0587311292077937e-05,
+      "loss": 1.1875,
+      "step": 15527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0586276659157075e-05,
+      "loss": 1.248,
+      "step": 15528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0585242019938644e-05,
+      "loss": 0.8989,
+      "step": 15529
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058420737443375e-05,
+      "loss": 0.9834,
+      "step": 15530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0583172722653514e-05,
+      "loss": 0.9746,
+      "step": 15531
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0582138064609045e-05,
+      "loss": 0.6807,
+      "step": 15532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0581103400311461e-05,
+      "loss": 1.0811,
+      "step": 15533
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0580068729771872e-05,
+      "loss": 1.0449,
+      "step": 15534
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0579034053001394e-05,
+      "loss": 1.1582,
+      "step": 15535
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0577999370011142e-05,
+      "loss": 0.959,
+      "step": 15536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0576964680812229e-05,
+      "loss": 0.6767,
+      "step": 15537
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.057592998541577e-05,
+      "loss": 1.0557,
+      "step": 15538
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0574895283832877e-05,
+      "loss": 0.8213,
+      "step": 15539
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0573860576074667e-05,
+      "loss": 1.2324,
+      "step": 15540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0572825862152257e-05,
+      "loss": 0.9971,
+      "step": 15541
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0571791142076758e-05,
+      "loss": 1.2764,
+      "step": 15542
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.057075641585928e-05,
+      "loss": 0.7827,
+      "step": 15543
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0569721683510947e-05,
+      "loss": 0.9333,
+      "step": 15544
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0568686945042865e-05,
+      "loss": 1.1104,
+      "step": 15545
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0567652200466157e-05,
+      "loss": 1.0483,
+      "step": 15546
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0566617449791932e-05,
+      "loss": 0.917,
+      "step": 15547
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0565582693031304e-05,
+      "loss": 1.2021,
+      "step": 15548
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0564547930195396e-05,
+      "loss": 1.1436,
+      "step": 15549
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0563513161295314e-05,
+      "loss": 0.9912,
+      "step": 15550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.056247838634218e-05,
+      "loss": 1.1143,
+      "step": 15551
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0561443605347106e-05,
+      "loss": 1.0068,
+      "step": 15552
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0560408818321207e-05,
+      "loss": 1.4883,
+      "step": 15553
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0559374025275597e-05,
+      "loss": 1.1191,
+      "step": 15554
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0558339226221394e-05,
+      "loss": 0.8672,
+      "step": 15555
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0557304421169712e-05,
+      "loss": 0.8521,
+      "step": 15556
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0556269610131667e-05,
+      "loss": 1.1582,
+      "step": 15557
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0555234793118373e-05,
+      "loss": 0.9707,
+      "step": 15558
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.055419997014095e-05,
+      "loss": 0.9111,
+      "step": 15559
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0553165141210513e-05,
+      "loss": 1.0283,
+      "step": 15560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0552130306338168e-05,
+      "loss": 0.7954,
+      "step": 15561
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0551095465535047e-05,
+      "loss": 0.918,
+      "step": 15562
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.055006061881225e-05,
+      "loss": 1.1475,
+      "step": 15563
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0549025766180907e-05,
+      "loss": 1.2383,
+      "step": 15564
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0547990907652123e-05,
+      "loss": 1.1582,
+      "step": 15565
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054695604323702e-05,
+      "loss": 1.3398,
+      "step": 15566
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0545921172946712e-05,
+      "loss": 0.959,
+      "step": 15567
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0544886296792316e-05,
+      "loss": 1.0117,
+      "step": 15568
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0543851414784946e-05,
+      "loss": 0.9092,
+      "step": 15569
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0542816526935722e-05,
+      "loss": 0.8828,
+      "step": 15570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054178163325576e-05,
+      "loss": 1.0654,
+      "step": 15571
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0540746733756173e-05,
+      "loss": 1.1797,
+      "step": 15572
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0539711828448082e-05,
+      "loss": 0.8013,
+      "step": 15573
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0538676917342598e-05,
+      "loss": 0.9854,
+      "step": 15574
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0537642000450846e-05,
+      "loss": 1.2285,
+      "step": 15575
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0536607077783931e-05,
+      "loss": 1.0566,
+      "step": 15576
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0535572149352982e-05,
+      "loss": 1.0645,
+      "step": 15577
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0534537215169106e-05,
+      "loss": 1.123,
+      "step": 15578
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0533502275243428e-05,
+      "loss": 1.1143,
+      "step": 15579
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053246732958706e-05,
+      "loss": 1.2617,
+      "step": 15580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0531432378211117e-05,
+      "loss": 0.8779,
+      "step": 15581
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0530397421126721e-05,
+      "loss": 0.7607,
+      "step": 15582
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.052936245834499e-05,
+      "loss": 0.9336,
+      "step": 15583
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0528327489877035e-05,
+      "loss": 1.292,
+      "step": 15584
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0527292515733979e-05,
+      "loss": 1.5645,
+      "step": 15585
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0526257535926936e-05,
+      "loss": 1.2969,
+      "step": 15586
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0525222550467023e-05,
+      "loss": 1.209,
+      "step": 15587
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0524187559365362e-05,
+      "loss": 1.0742,
+      "step": 15588
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0523152562633063e-05,
+      "loss": 0.9648,
+      "step": 15589
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0522117560281251e-05,
+      "loss": 1.2402,
+      "step": 15590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0521082552321041e-05,
+      "loss": 1.1318,
+      "step": 15591
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0520047538763549e-05,
+      "loss": 0.7197,
+      "step": 15592
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0519012519619893e-05,
+      "loss": 1.1597,
+      "step": 15593
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0517977494901192e-05,
+      "loss": 0.8276,
+      "step": 15594
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0516942464618567e-05,
+      "loss": 0.9404,
+      "step": 15595
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.051590742878313e-05,
+      "loss": 0.8467,
+      "step": 15596
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0514872387406e-05,
+      "loss": 0.7153,
+      "step": 15597
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0513837340498299e-05,
+      "loss": 0.7539,
+      "step": 15598
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0512802288071144e-05,
+      "loss": 1.2266,
+      "step": 15599
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0511767230135648e-05,
+      "loss": 1.0068,
+      "step": 15600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0510732166702935e-05,
+      "loss": 0.6179,
+      "step": 15601
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0509697097784122e-05,
+      "loss": 1.0791,
+      "step": 15602
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0508662023390323e-05,
+      "loss": 1.1211,
+      "step": 15603
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0507626943532667e-05,
+      "loss": 0.9834,
+      "step": 15604
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.050659185822226e-05,
+      "loss": 1.0625,
+      "step": 15605
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0505556767470229e-05,
+      "loss": 1.0801,
+      "step": 15606
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0504521671287687e-05,
+      "loss": 1.1279,
+      "step": 15607
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0503486569685756e-05,
+      "loss": 0.5439,
+      "step": 15608
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0502451462675556e-05,
+      "loss": 1.3291,
+      "step": 15609
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0501416350268201e-05,
+      "loss": 0.8486,
+      "step": 15610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0500381232474814e-05,
+      "loss": 1.1738,
+      "step": 15611
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0499346109306512e-05,
+      "loss": 1.1963,
+      "step": 15612
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0498310980774413e-05,
+      "loss": 0.8374,
+      "step": 15613
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0497275846889638e-05,
+      "loss": 0.8945,
+      "step": 15614
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0496240707663306e-05,
+      "loss": 1.2422,
+      "step": 15615
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0495205563106534e-05,
+      "loss": 0.8491,
+      "step": 15616
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0494170413230446e-05,
+      "loss": 1.0601,
+      "step": 15617
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0493135258046155e-05,
+      "loss": 1.0674,
+      "step": 15618
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0492100097564784e-05,
+      "loss": 1.1162,
+      "step": 15619
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0491064931797447e-05,
+      "loss": 1.4541,
+      "step": 15620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0490029760755273e-05,
+      "loss": 1.1631,
+      "step": 15621
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0488994584449375e-05,
+      "loss": 0.9282,
+      "step": 15622
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0487959402890872e-05,
+      "loss": 0.9648,
+      "step": 15623
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0486924216090888e-05,
+      "loss": 0.9609,
+      "step": 15624
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0485889024060536e-05,
+      "loss": 1.0713,
+      "step": 15625
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0484853826810942e-05,
+      "loss": 1.207,
+      "step": 15626
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0483818624353222e-05,
+      "loss": 1.1191,
+      "step": 15627
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0482783416698496e-05,
+      "loss": 1.0713,
+      "step": 15628
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0481748203857887e-05,
+      "loss": 1.1885,
+      "step": 15629
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0480712985842513e-05,
+      "loss": 0.7832,
+      "step": 15630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0479677762663491e-05,
+      "loss": 1.0098,
+      "step": 15631
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0478642534331946e-05,
+      "loss": 1.042,
+      "step": 15632
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0477607300858991e-05,
+      "loss": 1.4121,
+      "step": 15633
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0476572062255758e-05,
+      "loss": 1.2656,
+      "step": 15634
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0475536818533357e-05,
+      "loss": 0.668,
+      "step": 15635
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0474501569702909e-05,
+      "loss": 0.9922,
+      "step": 15636
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0473466315775535e-05,
+      "loss": 0.6948,
+      "step": 15637
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0472431056762357e-05,
+      "loss": 1.2285,
+      "step": 15638
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0471395792674502e-05,
+      "loss": 1.0508,
+      "step": 15639
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0470360523523078e-05,
+      "loss": 1.0322,
+      "step": 15640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0469325249319211e-05,
+      "loss": 0.9541,
+      "step": 15641
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0468289970074021e-05,
+      "loss": 1.1348,
+      "step": 15642
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.046725468579863e-05,
+      "loss": 0.8691,
+      "step": 15643
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0466219396504158e-05,
+      "loss": 1.3672,
+      "step": 15644
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0465184102201727e-05,
+      "loss": 1.2734,
+      "step": 15645
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0464148802902451e-05,
+      "loss": 0.8716,
+      "step": 15646
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0463113498617461e-05,
+      "loss": 0.7969,
+      "step": 15647
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0462078189357874e-05,
+      "loss": 0.981,
+      "step": 15648
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0461042875134804e-05,
+      "loss": 0.7739,
+      "step": 15649
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0460007555959384e-05,
+      "loss": 1.082,
+      "step": 15650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0458972231842724e-05,
+      "loss": 0.9688,
+      "step": 15651
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0457936902795954e-05,
+      "loss": 1.2803,
+      "step": 15652
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0456901568830188e-05,
+      "loss": 1.0635,
+      "step": 15653
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0455866229956551e-05,
+      "loss": 1.1543,
+      "step": 15654
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0454830886186163e-05,
+      "loss": 1.2812,
+      "step": 15655
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0453795537530145e-05,
+      "loss": 0.8799,
+      "step": 15656
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045276018399962e-05,
+      "loss": 1.2017,
+      "step": 15657
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045172482560571e-05,
+      "loss": 0.8037,
+      "step": 15658
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045068946235953e-05,
+      "loss": 1.4863,
+      "step": 15659
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0449654094272212e-05,
+      "loss": 1.0352,
+      "step": 15660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0448618721354874e-05,
+      "loss": 1.2637,
+      "step": 15661
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0447583343618629e-05,
+      "loss": 1.1992,
+      "step": 15662
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044654796107461e-05,
+      "loss": 0.7783,
+      "step": 15663
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0445512573733928e-05,
+      "loss": 0.8389,
+      "step": 15664
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0444477181607715e-05,
+      "loss": 0.9272,
+      "step": 15665
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0443441784707088e-05,
+      "loss": 1.0137,
+      "step": 15666
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044240638304317e-05,
+      "loss": 1.3369,
+      "step": 15667
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0441370976627082e-05,
+      "loss": 1.0869,
+      "step": 15668
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0440335565469944e-05,
+      "loss": 1.2051,
+      "step": 15669
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0439300149582881e-05,
+      "loss": 1.0337,
+      "step": 15670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0438264728977017e-05,
+      "loss": 1.4956,
+      "step": 15671
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0437229303663468e-05,
+      "loss": 1.1084,
+      "step": 15672
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0436193873653362e-05,
+      "loss": 1.2539,
+      "step": 15673
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043515843895782e-05,
+      "loss": 1.0098,
+      "step": 15674
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0434122999587956e-05,
+      "loss": 1.0959,
+      "step": 15675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0433087555554908e-05,
+      "loss": 0.9971,
+      "step": 15676
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0432052106869783e-05,
+      "loss": 1.0488,
+      "step": 15677
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0431016653543715e-05,
+      "loss": 0.9443,
+      "step": 15678
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0429981195587822e-05,
+      "loss": 1.2871,
+      "step": 15679
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0428945733013223e-05,
+      "loss": 0.959,
+      "step": 15680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0427910265831044e-05,
+      "loss": 0.9023,
+      "step": 15681
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.042687479405241e-05,
+      "loss": 1.3076,
+      "step": 15682
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0425839317688437e-05,
+      "loss": 0.9678,
+      "step": 15683
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0424803836750254e-05,
+      "loss": 1.3281,
+      "step": 15684
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.042376835124898e-05,
+      "loss": 0.9336,
+      "step": 15685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0422732861195742e-05,
+      "loss": 1.0562,
+      "step": 15686
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0421697366601657e-05,
+      "loss": 0.9351,
+      "step": 15687
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0420661867477853e-05,
+      "loss": 0.7422,
+      "step": 15688
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0419626363835454e-05,
+      "loss": 0.8906,
+      "step": 15689
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0418590855685573e-05,
+      "loss": 0.6987,
+      "step": 15690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0417555343039345e-05,
+      "loss": 0.9238,
+      "step": 15691
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0416519825907888e-05,
+      "loss": 0.7334,
+      "step": 15692
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041548430430232e-05,
+      "loss": 1.1162,
+      "step": 15693
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0414448778233776e-05,
+      "loss": 1.0332,
+      "step": 15694
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0413413247713369e-05,
+      "loss": 0.9639,
+      "step": 15695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0412377712752229e-05,
+      "loss": 1.0557,
+      "step": 15696
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0411342173361474e-05,
+      "loss": 0.7349,
+      "step": 15697
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0410306629552231e-05,
+      "loss": 1.125,
+      "step": 15698
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0409271081335622e-05,
+      "loss": 0.9126,
+      "step": 15699
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040823552872277e-05,
+      "loss": 0.915,
+      "step": 15700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04071999717248e-05,
+      "loss": 1.0469,
+      "step": 15701
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0406164410352835e-05,
+      "loss": 0.7979,
+      "step": 15702
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0405128844617998e-05,
+      "loss": 1.1699,
+      "step": 15703
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0404093274531413e-05,
+      "loss": 0.8955,
+      "step": 15704
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0403057700104206e-05,
+      "loss": 0.834,
+      "step": 15705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0402022121347495e-05,
+      "loss": 0.7969,
+      "step": 15706
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040098653827241e-05,
+      "loss": 1.084,
+      "step": 15707
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039995095089007e-05,
+      "loss": 1.0508,
+      "step": 15708
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0398915359211605e-05,
+      "loss": 1.0615,
+      "step": 15709
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0397879763248131e-05,
+      "loss": 1.2236,
+      "step": 15710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039684416301078e-05,
+      "loss": 1.1416,
+      "step": 15711
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0395808558510669e-05,
+      "loss": 1.4766,
+      "step": 15712
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0394772949758927e-05,
+      "loss": 1.0703,
+      "step": 15713
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0393737336766674e-05,
+      "loss": 0.6592,
+      "step": 15714
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039270171954504e-05,
+      "loss": 0.9512,
+      "step": 15715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0391666098105143e-05,
+      "loss": 1.2061,
+      "step": 15716
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0390630472458112e-05,
+      "loss": 1.1436,
+      "step": 15717
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0389594842615067e-05,
+      "loss": 1.0342,
+      "step": 15718
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0388559208587137e-05,
+      "loss": 1.6797,
+      "step": 15719
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0387523570385446e-05,
+      "loss": 1.2402,
+      "step": 15720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0386487928021112e-05,
+      "loss": 0.8018,
+      "step": 15721
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0385452281505269e-05,
+      "loss": 0.8574,
+      "step": 15722
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0384416630849033e-05,
+      "loss": 0.8848,
+      "step": 15723
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0383380976063533e-05,
+      "loss": 0.8438,
+      "step": 15724
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0382345317159892e-05,
+      "loss": 1.0225,
+      "step": 15725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0381309654149237e-05,
+      "loss": 0.8521,
+      "step": 15726
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038027398704269e-05,
+      "loss": 1.0088,
+      "step": 15727
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0379238315851379e-05,
+      "loss": 0.7489,
+      "step": 15728
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0378202640586426e-05,
+      "loss": 1.0293,
+      "step": 15729
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0377166961258955e-05,
+      "loss": 0.9541,
+      "step": 15730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0376131277880096e-05,
+      "loss": 1.0986,
+      "step": 15731
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0375095590460968e-05,
+      "loss": 1.4326,
+      "step": 15732
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.03740598990127e-05,
+      "loss": 1.4785,
+      "step": 15733
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0373024203546413e-05,
+      "loss": 1.2876,
+      "step": 15734
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0371988504073238e-05,
+      "loss": 0.6621,
+      "step": 15735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0370952800604295e-05,
+      "loss": 0.7822,
+      "step": 15736
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.036991709315071e-05,
+      "loss": 1.0215,
+      "step": 15737
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0368881381723612e-05,
+      "loss": 1.3887,
+      "step": 15738
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0367845666334118e-05,
+      "loss": 1.0605,
+      "step": 15739
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0366809946993364e-05,
+      "loss": 0.8018,
+      "step": 15740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0365774223712467e-05,
+      "loss": 0.915,
+      "step": 15741
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0364738496502555e-05,
+      "loss": 0.7656,
+      "step": 15742
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0363702765374757e-05,
+      "loss": 0.7217,
+      "step": 15743
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0362667030340193e-05,
+      "loss": 0.7212,
+      "step": 15744
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0361631291409989e-05,
+      "loss": 0.9316,
+      "step": 15745
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0360595548595274e-05,
+      "loss": 1.0127,
+      "step": 15746
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0359559801907172e-05,
+      "loss": 1.2227,
+      "step": 15747
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0358524051356807e-05,
+      "loss": 0.9302,
+      "step": 15748
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0357488296955305e-05,
+      "loss": 0.8374,
+      "step": 15749
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0356452538713795e-05,
+      "loss": 1.46,
+      "step": 15750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0355416776643401e-05,
+      "loss": 0.9229,
+      "step": 15751
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0354381010755245e-05,
+      "loss": 1.123,
+      "step": 15752
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.035334524106046e-05,
+      "loss": 0.4099,
+      "step": 15753
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0352309467570165e-05,
+      "loss": 0.896,
+      "step": 15754
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0351273690295489e-05,
+      "loss": 1.2334,
+      "step": 15755
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.035023790924756e-05,
+      "loss": 0.7783,
+      "step": 15756
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0349202124437498e-05,
+      "loss": 1.1309,
+      "step": 15757
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0348166335876437e-05,
+      "loss": 0.8882,
+      "step": 15758
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0347130543575496e-05,
+      "loss": 0.8892,
+      "step": 15759
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0346094747545805e-05,
+      "loss": 1.2715,
+      "step": 15760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0345058947798487e-05,
+      "loss": 0.9932,
+      "step": 15761
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0344023144344673e-05,
+      "loss": 0.5942,
+      "step": 15762
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0342987337195487e-05,
+      "loss": 0.7456,
+      "step": 15763
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0341951526362055e-05,
+      "loss": 0.812,
+      "step": 15764
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0340915711855499e-05,
+      "loss": 1.3398,
+      "step": 15765
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0339879893686954e-05,
+      "loss": 0.9609,
+      "step": 15766
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0338844071867539e-05,
+      "loss": 1.2461,
+      "step": 15767
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0337808246408386e-05,
+      "loss": 1.1904,
+      "step": 15768
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0336772417320615e-05,
+      "loss": 0.832,
+      "step": 15769
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0335736584615357e-05,
+      "loss": 0.8955,
+      "step": 15770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0334700748303738e-05,
+      "loss": 1.3486,
+      "step": 15771
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0333664908396885e-05,
+      "loss": 1.0723,
+      "step": 15772
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0332629064905925e-05,
+      "loss": 1.0957,
+      "step": 15773
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0331593217841982e-05,
+      "loss": 1.1348,
+      "step": 15774
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0330557367216186e-05,
+      "loss": 1.0146,
+      "step": 15775
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.032952151303966e-05,
+      "loss": 1.1475,
+      "step": 15776
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0328485655323535e-05,
+      "loss": 1.125,
+      "step": 15777
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0327449794078932e-05,
+      "loss": 1.3457,
+      "step": 15778
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0326413929316986e-05,
+      "loss": 0.9341,
+      "step": 15779
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0325378061048815e-05,
+      "loss": 1.0498,
+      "step": 15780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0324342189285555e-05,
+      "loss": 1.4277,
+      "step": 15781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0323306314038324e-05,
+      "loss": 0.9229,
+      "step": 15782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0322270435318254e-05,
+      "loss": 1.1514,
+      "step": 15783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0321234553136472e-05,
+      "loss": 1.333,
+      "step": 15784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0320198667504107e-05,
+      "loss": 1.0928,
+      "step": 15785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031916277843228e-05,
+      "loss": 1.1865,
+      "step": 15786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0318126885932122e-05,
+      "loss": 1.0605,
+      "step": 15787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0317090990014759e-05,
+      "loss": 1.041,
+      "step": 15788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0316055090691321e-05,
+      "loss": 0.7373,
+      "step": 15789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0315019187972934e-05,
+      "loss": 1.1494,
+      "step": 15790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0313983281870722e-05,
+      "loss": 0.9707,
+      "step": 15791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0312947372395814e-05,
+      "loss": 0.8682,
+      "step": 15792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031191145955934e-05,
+      "loss": 0.8628,
+      "step": 15793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0310875543372425e-05,
+      "loss": 0.9072,
+      "step": 15794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.03098396238462e-05,
+      "loss": 0.8921,
+      "step": 15795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0308803700991786e-05,
+      "loss": 1.0571,
+      "step": 15796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0307767774820317e-05,
+      "loss": 0.8379,
+      "step": 15797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0306731845342917e-05,
+      "loss": 0.6421,
+      "step": 15798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0305695912570712e-05,
+      "loss": 1.1191,
+      "step": 15799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0304659976514833e-05,
+      "loss": 0.7891,
+      "step": 15800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0303624037186406e-05,
+      "loss": 1.1826,
+      "step": 15801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0302588094596559e-05,
+      "loss": 1.1445,
+      "step": 15802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0301552148756423e-05,
+      "loss": 1.3369,
+      "step": 15803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0300516199677119e-05,
+      "loss": 0.9199,
+      "step": 15804
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.029948024736978e-05,
+      "loss": 0.9658,
+      "step": 15805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.029844429184553e-05,
+      "loss": 0.9453,
+      "step": 15806
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0297408333115501e-05,
+      "loss": 1.0723,
+      "step": 15807
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0296372371190822e-05,
+      "loss": 1.2529,
+      "step": 15808
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0295336406082612e-05,
+      "loss": 1.1982,
+      "step": 15809
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.029430043780201e-05,
+      "loss": 0.8877,
+      "step": 15810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0293264466360134e-05,
+      "loss": 1.0947,
+      "step": 15811
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0292228491768121e-05,
+      "loss": 1.21,
+      "step": 15812
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0291192514037095e-05,
+      "loss": 0.9268,
+      "step": 15813
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0290156533178182e-05,
+      "loss": 0.5188,
+      "step": 15814
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0289120549202511e-05,
+      "loss": 1.2236,
+      "step": 15815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0288084562121215e-05,
+      "loss": 1.1191,
+      "step": 15816
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0287048571945414e-05,
+      "loss": 1.2666,
+      "step": 15817
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0286012578686244e-05,
+      "loss": 0.7241,
+      "step": 15818
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0284976582354831e-05,
+      "loss": 1.0557,
+      "step": 15819
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0283940582962299e-05,
+      "loss": 0.7471,
+      "step": 15820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0282904580519784e-05,
+      "loss": 1.209,
+      "step": 15821
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0281868575038404e-05,
+      "loss": 0.9087,
+      "step": 15822
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0280832566529302e-05,
+      "loss": 0.8301,
+      "step": 15823
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0279796555003589e-05,
+      "loss": 0.9453,
+      "step": 15824
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027876054047241e-05,
+      "loss": 1.1152,
+      "step": 15825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0277724522946882e-05,
+      "loss": 1.4629,
+      "step": 15826
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0276688502438136e-05,
+      "loss": 1.0049,
+      "step": 15827
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0275652478957305e-05,
+      "loss": 0.834,
+      "step": 15828
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0274616452515514e-05,
+      "loss": 1.0664,
+      "step": 15829
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027358042312389e-05,
+      "loss": 1.1396,
+      "step": 15830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0272544390793564e-05,
+      "loss": 1.3125,
+      "step": 15831
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0271508355535666e-05,
+      "loss": 1.3496,
+      "step": 15832
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0270472317361322e-05,
+      "loss": 0.9541,
+      "step": 15833
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0269436276281665e-05,
+      "loss": 0.9067,
+      "step": 15834
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0268400232307815e-05,
+      "loss": 0.9258,
+      "step": 15835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0267364185450914e-05,
+      "loss": 1.0186,
+      "step": 15836
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0266328135722076e-05,
+      "loss": 1.0547,
+      "step": 15837
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0265292083132441e-05,
+      "loss": 1.2334,
+      "step": 15838
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0264256027693136e-05,
+      "loss": 1.2178,
+      "step": 15839
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0263219969415283e-05,
+      "loss": 0.8613,
+      "step": 15840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0262183908310021e-05,
+      "loss": 1.3486,
+      "step": 15841
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0261147844388472e-05,
+      "loss": 1.2324,
+      "step": 15842
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0260111777661767e-05,
+      "loss": 0.6533,
+      "step": 15843
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0259075708141035e-05,
+      "loss": 1.0596,
+      "step": 15844
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0258039635837407e-05,
+      "loss": 0.8408,
+      "step": 15845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0257003560762008e-05,
+      "loss": 1.4385,
+      "step": 15846
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025596748292597e-05,
+      "loss": 0.9336,
+      "step": 15847
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0254931402340422e-05,
+      "loss": 1.0332,
+      "step": 15848
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0253895319016494e-05,
+      "loss": 0.7256,
+      "step": 15849
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025285923296531e-05,
+      "loss": 1.0381,
+      "step": 15850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0251823144198008e-05,
+      "loss": 1.0566,
+      "step": 15851
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0250787052725712e-05,
+      "loss": 0.8594,
+      "step": 15852
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024975095855955e-05,
+      "loss": 0.9443,
+      "step": 15853
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0248714861710655e-05,
+      "loss": 1.1631,
+      "step": 15854
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0247678762190151e-05,
+      "loss": 0.7354,
+      "step": 15855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0246642660009178e-05,
+      "loss": 0.7026,
+      "step": 15856
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0245606555178854e-05,
+      "loss": 1.4766,
+      "step": 15857
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0244570447710314e-05,
+      "loss": 0.9609,
+      "step": 15858
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0243534337614684e-05,
+      "loss": 0.998,
+      "step": 15859
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0242498224903099e-05,
+      "loss": 1.0195,
+      "step": 15860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0241462109586684e-05,
+      "loss": 0.7817,
+      "step": 15861
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024042599167657e-05,
+      "loss": 0.9062,
+      "step": 15862
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0239389871183888e-05,
+      "loss": 1.0,
+      "step": 15863
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0238353748119767e-05,
+      "loss": 1.1729,
+      "step": 15864
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0237317622495336e-05,
+      "loss": 1.0381,
+      "step": 15865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.023628149432172e-05,
+      "loss": 1.2432,
+      "step": 15866
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.023524536361006e-05,
+      "loss": 0.8027,
+      "step": 15867
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0234209230371475e-05,
+      "loss": 0.876,
+      "step": 15868
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0233173094617102e-05,
+      "loss": 1.3379,
+      "step": 15869
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0232136956358065e-05,
+      "loss": 0.8818,
+      "step": 15870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0231100815605497e-05,
+      "loss": 1.1582,
+      "step": 15871
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0230064672370527e-05,
+      "loss": 0.9141,
+      "step": 15872
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0229028526664288e-05,
+      "loss": 1.0859,
+      "step": 15873
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0227992378497905e-05,
+      "loss": 1.252,
+      "step": 15874
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.022695622788251e-05,
+      "loss": 1.1655,
+      "step": 15875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0225920074829235e-05,
+      "loss": 1.0391,
+      "step": 15876
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0224883919349206e-05,
+      "loss": 1.041,
+      "step": 15877
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0223847761453558e-05,
+      "loss": 0.8203,
+      "step": 15878
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0222811601153413e-05,
+      "loss": 1.083,
+      "step": 15879
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0221775438459911e-05,
+      "loss": 1.4473,
+      "step": 15880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0220739273384173e-05,
+      "loss": 0.8501,
+      "step": 15881
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0219703105937335e-05,
+      "loss": 1.1689,
+      "step": 15882
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021866693613053e-05,
+      "loss": 0.8696,
+      "step": 15883
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0217630763974877e-05,
+      "loss": 0.9712,
+      "step": 15884
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0216594589481517e-05,
+      "loss": 1.2637,
+      "step": 15885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0215558412661573e-05,
+      "loss": 1.0146,
+      "step": 15886
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0214522233526183e-05,
+      "loss": 0.9512,
+      "step": 15887
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021348605208647e-05,
+      "loss": 0.6069,
+      "step": 15888
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0212449868353564e-05,
+      "loss": 1.125,
+      "step": 15889
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0211413682338603e-05,
+      "loss": 1.1177,
+      "step": 15890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0210377494052707e-05,
+      "loss": 1.1211,
+      "step": 15891
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0209341303507017e-05,
+      "loss": 1.0034,
+      "step": 15892
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0208305110712655e-05,
+      "loss": 1.2793,
+      "step": 15893
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0207268915680757e-05,
+      "loss": 0.9243,
+      "step": 15894
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0206232718422451e-05,
+      "loss": 0.9355,
+      "step": 15895
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0205196518948868e-05,
+      "loss": 0.876,
+      "step": 15896
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0204160317271134e-05,
+      "loss": 1.2559,
+      "step": 15897
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020312411340039e-05,
+      "loss": 1.1313,
+      "step": 15898
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0202087907347752e-05,
+      "loss": 0.6257,
+      "step": 15899
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0201051699124365e-05,
+      "loss": 0.9863,
+      "step": 15900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0200015488741351e-05,
+      "loss": 1.1572,
+      "step": 15901
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0198979276209842e-05,
+      "loss": 0.8594,
+      "step": 15902
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.019794306154097e-05,
+      "loss": 0.999,
+      "step": 15903
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0196906844745867e-05,
+      "loss": 1.1885,
+      "step": 15904
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.019587062583566e-05,
+      "loss": 0.9258,
+      "step": 15905
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0194834404821481e-05,
+      "loss": 1.1309,
+      "step": 15906
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0193798181714461e-05,
+      "loss": 1.0466,
+      "step": 15907
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0192761956525732e-05,
+      "loss": 0.957,
+      "step": 15908
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0191725729266424e-05,
+      "loss": 1.0879,
+      "step": 15909
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0190689499947664e-05,
+      "loss": 1.582,
+      "step": 15910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018965326858059e-05,
+      "loss": 1.3613,
+      "step": 15911
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0188617035176327e-05,
+      "loss": 1.4414,
+      "step": 15912
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018758079974601e-05,
+      "loss": 1.1787,
+      "step": 15913
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0186544562300766e-05,
+      "loss": 1.0469,
+      "step": 15914
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0185508322851727e-05,
+      "loss": 1.0625,
+      "step": 15915
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0184472081410026e-05,
+      "loss": 1.0947,
+      "step": 15916
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0183435837986792e-05,
+      "loss": 1.002,
+      "step": 15917
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0182399592593155e-05,
+      "loss": 0.9805,
+      "step": 15918
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0181363345240249e-05,
+      "loss": 0.8164,
+      "step": 15919
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0180327095939202e-05,
+      "loss": 1.1309,
+      "step": 15920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0179290844701149e-05,
+      "loss": 0.9932,
+      "step": 15921
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0178254591537219e-05,
+      "loss": 1.2432,
+      "step": 15922
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0177218336458536e-05,
+      "loss": 1.2803,
+      "step": 15923
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0176182079476244e-05,
+      "loss": 0.8984,
+      "step": 15924
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0175145820601464e-05,
+      "loss": 0.8044,
+      "step": 15925
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0174109559845335e-05,
+      "loss": 1.1621,
+      "step": 15926
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0173073297218981e-05,
+      "loss": 0.9033,
+      "step": 15927
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0172037032733535e-05,
+      "loss": 1.1045,
+      "step": 15928
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0171000766400132e-05,
+      "loss": 1.0247,
+      "step": 15929
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0169964498229896e-05,
+      "loss": 0.7432,
+      "step": 15930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016892822823397e-05,
+      "loss": 1.2285,
+      "step": 15931
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0167891956423473e-05,
+      "loss": 1.0571,
+      "step": 15932
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0166855682809542e-05,
+      "loss": 0.4683,
+      "step": 15933
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0165819407403305e-05,
+      "loss": 1.2695,
+      "step": 15934
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0164783130215901e-05,
+      "loss": 1.0635,
+      "step": 15935
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0163746851258452e-05,
+      "loss": 1.0674,
+      "step": 15936
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0162710570542094e-05,
+      "loss": 1.2461,
+      "step": 15937
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0161674288077959e-05,
+      "loss": 0.8804,
+      "step": 15938
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0160638003877174e-05,
+      "loss": 0.8916,
+      "step": 15939
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.015960171795088e-05,
+      "loss": 1.2822,
+      "step": 15940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0158565430310195e-05,
+      "loss": 0.6704,
+      "step": 15941
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0157529140966262e-05,
+      "loss": 0.7725,
+      "step": 15942
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0156492849930205e-05,
+      "loss": 0.9766,
+      "step": 15943
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0155456557213161e-05,
+      "loss": 0.7388,
+      "step": 15944
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0154420262826257e-05,
+      "loss": 1.0684,
+      "step": 15945
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0153383966780626e-05,
+      "loss": 0.9248,
+      "step": 15946
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0152347669087401e-05,
+      "loss": 0.7944,
+      "step": 15947
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.015131136975771e-05,
+      "loss": 0.8843,
+      "step": 15948
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0150275068802689e-05,
+      "loss": 0.8276,
+      "step": 15949
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0149238766233466e-05,
+      "loss": 1.291,
+      "step": 15950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0148202462061175e-05,
+      "loss": 0.9258,
+      "step": 15951
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0147166156296946e-05,
+      "loss": 1.001,
+      "step": 15952
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0146129848951913e-05,
+      "loss": 1.1064,
+      "step": 15953
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0145093540037204e-05,
+      "loss": 0.6973,
+      "step": 15954
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0144057229563954e-05,
+      "loss": 0.9717,
+      "step": 15955
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0143020917543289e-05,
+      "loss": 1.1992,
+      "step": 15956
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0141984603986352e-05,
+      "loss": 0.8848,
+      "step": 15957
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0140948288904264e-05,
+      "loss": 0.7832,
+      "step": 15958
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0139911972308159e-05,
+      "loss": 1.1367,
+      "step": 15959
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.013887565420917e-05,
+      "loss": 0.9844,
+      "step": 15960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0137839334618431e-05,
+      "loss": 1.1279,
+      "step": 15961
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0136803013547071e-05,
+      "loss": 1.0781,
+      "step": 15962
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0135766691006221e-05,
+      "loss": 1.1055,
+      "step": 15963
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0134730367007015e-05,
+      "loss": 1.0176,
+      "step": 15964
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0133694041560584e-05,
+      "loss": 0.9229,
+      "step": 15965
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0132657714678062e-05,
+      "loss": 1.1045,
+      "step": 15966
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0131621386370576e-05,
+      "loss": 1.4082,
+      "step": 15967
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0130585056649262e-05,
+      "loss": 0.915,
+      "step": 15968
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0129548725525249e-05,
+      "loss": 1.0811,
+      "step": 15969
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0128512393009673e-05,
+      "loss": 0.8701,
+      "step": 15970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.012747605911366e-05,
+      "loss": 0.7715,
+      "step": 15971
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0126439723848346e-05,
+      "loss": 1.0596,
+      "step": 15972
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0125403387224864e-05,
+      "loss": 0.9727,
+      "step": 15973
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.012436704925434e-05,
+      "loss": 1.0918,
+      "step": 15974
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0123330709947914e-05,
+      "loss": 1.0153,
+      "step": 15975
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0122294369316714e-05,
+      "loss": 1.2227,
+      "step": 15976
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0121258027371869e-05,
+      "loss": 1.3096,
+      "step": 15977
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0120221684124516e-05,
+      "loss": 1.2148,
+      "step": 15978
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0119185339585787e-05,
+      "loss": 0.7319,
+      "step": 15979
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0118148993766808e-05,
+      "loss": 1.3906,
+      "step": 15980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0117112646678718e-05,
+      "loss": 0.9253,
+      "step": 15981
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0116076298332645e-05,
+      "loss": 0.9287,
+      "step": 15982
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0115039948739723e-05,
+      "loss": 1.041,
+      "step": 15983
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0114003597911084e-05,
+      "loss": 1.1484,
+      "step": 15984
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0112967245857856e-05,
+      "loss": 1.0146,
+      "step": 15985
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011193089259118e-05,
+      "loss": 0.9834,
+      "step": 15986
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0110894538122176e-05,
+      "loss": 1.3047,
+      "step": 15987
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010985818246199e-05,
+      "loss": 0.8711,
+      "step": 15988
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0108821825621741e-05,
+      "loss": 1.3418,
+      "step": 15989
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0107785467612572e-05,
+      "loss": 1.2236,
+      "step": 15990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0106749108445608e-05,
+      "loss": 1.0625,
+      "step": 15991
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0105712748131984e-05,
+      "loss": 0.8342,
+      "step": 15992
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010467638668283e-05,
+      "loss": 1.1826,
+      "step": 15993
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0103640024109282e-05,
+      "loss": 0.9531,
+      "step": 15994
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010260366042247e-05,
+      "loss": 1.1274,
+      "step": 15995
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0101567295633526e-05,
+      "loss": 0.7314,
+      "step": 15996
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0100530929753585e-05,
+      "loss": 1.1172,
+      "step": 15997
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0099494562793773e-05,
+      "loss": 0.9146,
+      "step": 15998
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0098458194765231e-05,
+      "loss": 0.9541,
+      "step": 15999
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.009742182567908e-05,
+      "loss": 0.9438,
+      "step": 16000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0096385455546467e-05,
+      "loss": 1.0342,
+      "step": 16001
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0095349084378512e-05,
+      "loss": 1.1484,
+      "step": 16002
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0094312712186351e-05,
+      "loss": 0.9221,
+      "step": 16003
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0093276338981116e-05,
+      "loss": 1.0176,
+      "step": 16004
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0092239964773944e-05,
+      "loss": 0.9961,
+      "step": 16005
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0091203589575962e-05,
+      "loss": 0.9033,
+      "step": 16006
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0090167213398303e-05,
+      "loss": 0.9585,
+      "step": 16007
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0089130836252098e-05,
+      "loss": 1.2383,
+      "step": 16008
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0088094458148487e-05,
+      "loss": 1.2949,
+      "step": 16009
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0087058079098595e-05,
+      "loss": 0.7695,
+      "step": 16010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0086021699113555e-05,
+      "loss": 1.0254,
+      "step": 16011
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0084985318204503e-05,
+      "loss": 1.0264,
+      "step": 16012
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0083948936382566e-05,
+      "loss": 0.71,
+      "step": 16013
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0082912553658886e-05,
+      "loss": 1.3066,
+      "step": 16014
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081876170044585e-05,
+      "loss": 0.9775,
+      "step": 16015
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00808397855508e-05,
+      "loss": 1.3662,
+      "step": 16016
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0079803400188664e-05,
+      "loss": 0.9124,
+      "step": 16017
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0078767013969303e-05,
+      "loss": 1.2363,
+      "step": 16018
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0077730626903864e-05,
+      "loss": 1.3125,
+      "step": 16019
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0076694239003466e-05,
+      "loss": 1.2275,
+      "step": 16020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0075657850279248e-05,
+      "loss": 0.6426,
+      "step": 16021
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.007462146074234e-05,
+      "loss": 1.335,
+      "step": 16022
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0073585070403873e-05,
+      "loss": 1.083,
+      "step": 16023
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0072548679274986e-05,
+      "loss": 0.9666,
+      "step": 16024
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0071512287366804e-05,
+      "loss": 1.1719,
+      "step": 16025
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0070475894690463e-05,
+      "loss": 1.0117,
+      "step": 16026
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0069439501257094e-05,
+      "loss": 1.0869,
+      "step": 16027
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0068403107077838e-05,
+      "loss": 0.8545,
+      "step": 16028
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0067366712163811e-05,
+      "loss": 1.1919,
+      "step": 16029
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0066330316526162e-05,
+      "loss": 1.0928,
+      "step": 16030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0065293920176012e-05,
+      "loss": 0.6045,
+      "step": 16031
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0064257523124503e-05,
+      "loss": 0.7769,
+      "step": 16032
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006322112538276e-05,
+      "loss": 1.0488,
+      "step": 16033
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0062184726961918e-05,
+      "loss": 1.0488,
+      "step": 16034
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0061148327873112e-05,
+      "loss": 1.0684,
+      "step": 16035
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0060111928127472e-05,
+      "loss": 1.167,
+      "step": 16036
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.005907552773613e-05,
+      "loss": 1.0439,
+      "step": 16037
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0058039126710221e-05,
+      "loss": 1.0234,
+      "step": 16038
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0057002725060876e-05,
+      "loss": 1.043,
+      "step": 16039
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0055966322799228e-05,
+      "loss": 0.854,
+      "step": 16040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0054929919936414e-05,
+      "loss": 0.9385,
+      "step": 16041
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0053893516483557e-05,
+      "loss": 1.085,
+      "step": 16042
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0052857112451802e-05,
+      "loss": 1.0547,
+      "step": 16043
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0051820707852267e-05,
+      "loss": 1.0244,
+      "step": 16044
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0050784302696099e-05,
+      "loss": 1.0078,
+      "step": 16045
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0049747896994422e-05,
+      "loss": 0.9912,
+      "step": 16046
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004871149075837e-05,
+      "loss": 1.3447,
+      "step": 16047
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0047675083999077e-05,
+      "loss": 1.0811,
+      "step": 16048
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0046638676727676e-05,
+      "loss": 1.04,
+      "step": 16049
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0045602268955298e-05,
+      "loss": 0.8318,
+      "step": 16050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0044565860693078e-05,
+      "loss": 1.1035,
+      "step": 16051
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004352945195215e-05,
+      "loss": 0.9717,
+      "step": 16052
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0042493042743642e-05,
+      "loss": 1.3232,
+      "step": 16053
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0041456633078688e-05,
+      "loss": 1.0293,
+      "step": 16054
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0040420222968424e-05,
+      "loss": 0.6055,
+      "step": 16055
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003938381242398e-05,
+      "loss": 0.9951,
+      "step": 16056
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0038347401456487e-05,
+      "loss": 0.7583,
+      "step": 16057
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0037310990077083e-05,
+      "loss": 0.8564,
+      "step": 16058
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0036274578296896e-05,
+      "loss": 0.8184,
+      "step": 16059
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0035238166127062e-05,
+      "loss": 1.2617,
+      "step": 16060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0034201753578712e-05,
+      "loss": 1.0098,
+      "step": 16061
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0033165340662979e-05,
+      "loss": 0.9395,
+      "step": 16062
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0032128927390993e-05,
+      "loss": 1.1934,
+      "step": 16063
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0031092513773893e-05,
+      "loss": 0.9775,
+      "step": 16064
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0030056099822806e-05,
+      "loss": 1.1992,
+      "step": 16065
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002901968554887e-05,
+      "loss": 0.9893,
+      "step": 16066
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0027983270963214e-05,
+      "loss": 1.2646,
+      "step": 16067
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0026946856076972e-05,
+      "loss": 1.0264,
+      "step": 16068
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0025910440901278e-05,
+      "loss": 1.0527,
+      "step": 16069
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002487402544726e-05,
+      "loss": 0.9885,
+      "step": 16070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0023837609726057e-05,
+      "loss": 0.8218,
+      "step": 16071
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0022801193748796e-05,
+      "loss": 0.9971,
+      "step": 16072
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0021764777526614e-05,
+      "loss": 1.1768,
+      "step": 16073
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0020728361070644e-05,
+      "loss": 1.2324,
+      "step": 16074
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0019691944392016e-05,
+      "loss": 1.041,
+      "step": 16075
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0018655527501867e-05,
+      "loss": 0.9346,
+      "step": 16076
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0017619110411323e-05,
+      "loss": 1.1934,
+      "step": 16077
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0016582693131525e-05,
+      "loss": 1.0576,
+      "step": 16078
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0015546275673598e-05,
+      "loss": 0.7461,
+      "step": 16079
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001450985804868e-05,
+      "loss": 1.0479,
+      "step": 16080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0013473440267903e-05,
+      "loss": 0.9053,
+      "step": 16081
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0012437022342399e-05,
+      "loss": 1.3135,
+      "step": 16082
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0011400604283298e-05,
+      "loss": 0.9214,
+      "step": 16083
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0010364186101739e-05,
+      "loss": 1.1689,
+      "step": 16084
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0009327767808853e-05,
+      "loss": 1.0459,
+      "step": 16085
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0008291349415766e-05,
+      "loss": 0.7178,
+      "step": 16086
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0007254930933623e-05,
+      "loss": 0.9307,
+      "step": 16087
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0006218512373545e-05,
+      "loss": 0.9385,
+      "step": 16088
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0005182093746672e-05,
+      "loss": 0.6616,
+      "step": 16089
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0004145675064137e-05,
+      "loss": 1.4072,
+      "step": 16090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0003109256337066e-05,
+      "loss": 0.9932,
+      "step": 16091
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0002072837576599e-05,
+      "loss": 1.166,
+      "step": 16092
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0001036418793868e-05,
+      "loss": 0.981,
+      "step": 16093
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1e-05,
+      "loss": 0.9893,
+      "step": 16094
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.998963581206135e-06,
+      "loss": 1.0078,
+      "step": 16095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.997927162423406e-06,
+      "loss": 1.0293,
+      "step": 16096
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.996890743662935e-06,
+      "loss": 1.0391,
+      "step": 16097
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.995854324935867e-06,
+      "loss": 1.2607,
+      "step": 16098
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99481790625333e-06,
+      "loss": 1.0889,
+      "step": 16099
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.993781487626458e-06,
+      "loss": 1.1797,
+      "step": 16100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.992745069066382e-06,
+      "loss": 1.2578,
+      "step": 16101
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.991708650584233e-06,
+      "loss": 0.9248,
+      "step": 16102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.990672232191149e-06,
+      "loss": 0.8994,
+      "step": 16103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.989635813898264e-06,
+      "loss": 0.7554,
+      "step": 16104
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988599395716703e-06,
+      "loss": 1.0083,
+      "step": 16105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.987562977657605e-06,
+      "loss": 1.0039,
+      "step": 16106
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.986526559732099e-06,
+      "loss": 1.0078,
+      "step": 16107
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.985490141951321e-06,
+      "loss": 0.833,
+      "step": 16108
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.984453724326405e-06,
+      "loss": 1.2773,
+      "step": 16109
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.98341730686848e-06,
+      "loss": 1.3525,
+      "step": 16110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.98238088958868e-06,
+      "loss": 1.062,
+      "step": 16111
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.981344472498138e-06,
+      "loss": 0.6997,
+      "step": 16112
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.980308055607986e-06,
+      "loss": 0.9346,
+      "step": 16113
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.979271638929358e-06,
+      "loss": 0.9961,
+      "step": 16114
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.97823522247339e-06,
+      "loss": 0.9795,
+      "step": 16115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.977198806251207e-06,
+      "loss": 1.2725,
+      "step": 16116
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.976162390273948e-06,
+      "loss": 1.0845,
+      "step": 16117
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.975125974552742e-06,
+      "loss": 0.6846,
+      "step": 16118
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.974089559098725e-06,
+      "loss": 0.7188,
+      "step": 16119
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.973053143923032e-06,
+      "loss": 0.8125,
+      "step": 16120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.972016729036788e-06,
+      "loss": 0.6963,
+      "step": 16121
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.970980314451136e-06,
+      "loss": 1.3701,
+      "step": 16122
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.969943900177192e-06,
+      "loss": 1.2373,
+      "step": 16123
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.968907486226108e-06,
+      "loss": 1.125,
+      "step": 16124
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.967871072609008e-06,
+      "loss": 1.0508,
+      "step": 16125
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.966834659337025e-06,
+      "loss": 1.2769,
+      "step": 16126
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.965798246421293e-06,
+      "loss": 0.9463,
+      "step": 16127
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.964761833872942e-06,
+      "loss": 1.1807,
+      "step": 16128
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.963725421703106e-06,
+      "loss": 0.9775,
+      "step": 16129
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.962689009922918e-06,
+      "loss": 0.9219,
+      "step": 16130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.961652598543517e-06,
+      "loss": 1.1396,
+      "step": 16131
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.960616187576025e-06,
+      "loss": 1.0752,
+      "step": 16132
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.959579777031578e-06,
+      "loss": 1.4395,
+      "step": 16133
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.958543366921312e-06,
+      "loss": 0.918,
+      "step": 16134
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.957506957256363e-06,
+      "loss": 1.0225,
+      "step": 16135
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.956470548047854e-06,
+      "loss": 0.8369,
+      "step": 16136
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.955434139306924e-06,
+      "loss": 0.8389,
+      "step": 16137
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.954397731044702e-06,
+      "loss": 1.0425,
+      "step": 16138
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.953361323272325e-06,
+      "loss": 0.9111,
+      "step": 16139
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.952324916000925e-06,
+      "loss": 0.7646,
+      "step": 16140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.951288509241632e-06,
+      "loss": 0.8945,
+      "step": 16141
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.950252103005583e-06,
+      "loss": 0.7266,
+      "step": 16142
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.949215697303906e-06,
+      "loss": 1.0166,
+      "step": 16143
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.948179292147733e-06,
+      "loss": 0.8555,
+      "step": 16144
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.947142887548202e-06,
+      "loss": 0.9561,
+      "step": 16145
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.946106483516446e-06,
+      "loss": 0.709,
+      "step": 16146
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94507008006359e-06,
+      "loss": 1.0773,
+      "step": 16147
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.944033677200773e-06,
+      "loss": 0.6934,
+      "step": 16148
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.942997274939124e-06,
+      "loss": 0.7896,
+      "step": 16149
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.941960873289782e-06,
+      "loss": 1.2988,
+      "step": 16150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.940924472263873e-06,
+      "loss": 0.8374,
+      "step": 16151
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.939888071872531e-06,
+      "loss": 1.2012,
+      "step": 16152
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.938851672126893e-06,
+      "loss": 1.1045,
+      "step": 16153
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.937815273038082e-06,
+      "loss": 0.8691,
+      "step": 16154
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.936778874617242e-06,
+      "loss": 1.0029,
+      "step": 16155
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.935742476875499e-06,
+      "loss": 0.7852,
+      "step": 16156
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93470607982399e-06,
+      "loss": 1.0215,
+      "step": 16157
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.933669683473843e-06,
+      "loss": 0.6626,
+      "step": 16158
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.932633287836189e-06,
+      "loss": 1.084,
+      "step": 16159
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.931596892922166e-06,
+      "loss": 1.0176,
+      "step": 16160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.930560498742908e-06,
+      "loss": 0.8164,
+      "step": 16161
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.92952410530954e-06,
+      "loss": 0.8071,
+      "step": 16162
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9284877126332e-06,
+      "loss": 1.1641,
+      "step": 16163
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.927451320725016e-06,
+      "loss": 0.668,
+      "step": 16164
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.926414929596127e-06,
+      "loss": 1.1279,
+      "step": 16165
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.925378539257664e-06,
+      "loss": 0.9688,
+      "step": 16166
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.924342149720755e-06,
+      "loss": 0.8447,
+      "step": 16167
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.923305760996539e-06,
+      "loss": 0.832,
+      "step": 16168
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.922269373096138e-06,
+      "loss": 1.4512,
+      "step": 16169
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.921232986030695e-06,
+      "loss": 1.0479,
+      "step": 16170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.920196599811341e-06,
+      "loss": 0.7568,
+      "step": 16171
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.919160214449203e-06,
+      "loss": 0.8022,
+      "step": 16172
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91812382995542e-06,
+      "loss": 1.0273,
+      "step": 16173
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.917087446341121e-06,
+      "loss": 0.8325,
+      "step": 16174
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.916051063617434e-06,
+      "loss": 0.7671,
+      "step": 16175
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.915014681795499e-06,
+      "loss": 0.9907,
+      "step": 16176
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.913978300886447e-06,
+      "loss": 1.1201,
+      "step": 16177
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.912941920901408e-06,
+      "loss": 1.2549,
+      "step": 16178
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91190554185152e-06,
+      "loss": 1.5605,
+      "step": 16179
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9108691637479e-06,
+      "loss": 1.0469,
+      "step": 16180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9098327866017e-06,
+      "loss": 1.3672,
+      "step": 16181
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.908796410424043e-06,
+      "loss": 1.1973,
+      "step": 16182
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.90776003522606e-06,
+      "loss": 0.5874,
+      "step": 16183
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.906723661018887e-06,
+      "loss": 0.9072,
+      "step": 16184
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.90568728781365e-06,
+      "loss": 1.127,
+      "step": 16185
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.904650915621491e-06,
+      "loss": 0.8955,
+      "step": 16186
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.903614544453535e-06,
+      "loss": 1.3984,
+      "step": 16187
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.902578174320921e-06,
+      "loss": 1.0322,
+      "step": 16188
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.901541805234774e-06,
+      "loss": 1.2354,
+      "step": 16189
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.900505437206227e-06,
+      "loss": 1.1074,
+      "step": 16190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.899469070246417e-06,
+      "loss": 1.1523,
+      "step": 16191
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.898432704366476e-06,
+      "loss": 1.1055,
+      "step": 16192
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.897396339577533e-06,
+      "loss": 1.0742,
+      "step": 16193
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89635997589072e-06,
+      "loss": 0.897,
+      "step": 16194
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89532361331717e-06,
+      "loss": 1.1821,
+      "step": 16195
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89428725186802e-06,
+      "loss": 0.8789,
+      "step": 16196
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.893250891554395e-06,
+      "loss": 1.0576,
+      "step": 16197
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.892214532387433e-06,
+      "loss": 1.2158,
+      "step": 16198
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.891178174378264e-06,
+      "loss": 1.1855,
+      "step": 16199
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89014181753801e-06,
+      "loss": 0.957,
+      "step": 16200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.889105461877825e-06,
+      "loss": 1.2324,
+      "step": 16201
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.888069107408824e-06,
+      "loss": 1.1924,
+      "step": 16202
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.887032754142147e-06,
+      "loss": 1.2314,
+      "step": 16203
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.88599640208892e-06,
+      "loss": 1.0283,
+      "step": 16204
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.884960051260282e-06,
+      "loss": 1.0532,
+      "step": 16205
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.883923701667357e-06,
+      "loss": 1.2871,
+      "step": 16206
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.882887353321286e-06,
+      "loss": 1.2422,
+      "step": 16207
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.881851006233194e-06,
+      "loss": 0.9697,
+      "step": 16208
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.880814660414217e-06,
+      "loss": 0.9453,
+      "step": 16209
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.879778315875487e-06,
+      "loss": 1.3516,
+      "step": 16210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.878741972628131e-06,
+      "loss": 1.3281,
+      "step": 16211
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87770563068329e-06,
+      "loss": 0.7974,
+      "step": 16212
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.876669290052087e-06,
+      "loss": 1.0527,
+      "step": 16213
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.875632950745662e-06,
+      "loss": 0.9067,
+      "step": 16214
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.874596612775141e-06,
+      "loss": 1.3711,
+      "step": 16215
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.873560276151655e-06,
+      "loss": 1.001,
+      "step": 16216
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.872523940886341e-06,
+      "loss": 0.8101,
+      "step": 16217
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87148760699033e-06,
+      "loss": 1.2715,
+      "step": 16218
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.870451274474754e-06,
+      "loss": 0.9238,
+      "step": 16219
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.869414943350743e-06,
+      "loss": 1.2012,
+      "step": 16220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.868378613629424e-06,
+      "loss": 0.957,
+      "step": 16221
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.86734228532194e-06,
+      "loss": 0.7754,
+      "step": 16222
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866305958439418e-06,
+      "loss": 0.7651,
+      "step": 16223
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.865269632992987e-06,
+      "loss": 0.7295,
+      "step": 16224
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.864233308993782e-06,
+      "loss": 1.0264,
+      "step": 16225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.86319698645293e-06,
+      "loss": 1.1895,
+      "step": 16226
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.862160665381572e-06,
+      "loss": 1.0381,
+      "step": 16227
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.861124345790831e-06,
+      "loss": 0.9697,
+      "step": 16228
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.860088027691843e-06,
+      "loss": 1.1299,
+      "step": 16229
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.859051711095741e-06,
+      "loss": 0.8818,
+      "step": 16230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.858015396013651e-06,
+      "loss": 0.7842,
+      "step": 16231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85697908245671e-06,
+      "loss": 0.8281,
+      "step": 16232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.855942770436049e-06,
+      "loss": 1.1899,
+      "step": 16233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8549064599628e-06,
+      "loss": 0.7168,
+      "step": 16234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.853870151048092e-06,
+      "loss": 0.8521,
+      "step": 16235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.852833843703058e-06,
+      "loss": 1.126,
+      "step": 16236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.851797537938825e-06,
+      "loss": 1.043,
+      "step": 16237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.850761233766537e-06,
+      "loss": 1.2852,
+      "step": 16238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.849724931197314e-06,
+      "loss": 1.2754,
+      "step": 16239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.848688630242293e-06,
+      "loss": 1.1094,
+      "step": 16240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.847652330912605e-06,
+      "loss": 1.2695,
+      "step": 16241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.846616033219377e-06,
+      "loss": 1.1768,
+      "step": 16242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.845579737173747e-06,
+      "loss": 1.2344,
+      "step": 16243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.844543442786842e-06,
+      "loss": 1.1797,
+      "step": 16244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.843507150069799e-06,
+      "loss": 1.041,
+      "step": 16245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.842470859033743e-06,
+      "loss": 1.0996,
+      "step": 16246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.841434569689805e-06,
+      "loss": 1.1748,
+      "step": 16247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.840398282049123e-06,
+      "loss": 1.3047,
+      "step": 16248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.839361996122827e-06,
+      "loss": 1.082,
+      "step": 16249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838325711922045e-06,
+      "loss": 0.9336,
+      "step": 16250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.83728942945791e-06,
+      "loss": 1.2285,
+      "step": 16251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.83625314874155e-06,
+      "loss": 0.7178,
+      "step": 16252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8352168697841e-06,
+      "loss": 0.8945,
+      "step": 16253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.834180592596696e-06,
+      "loss": 1.1689,
+      "step": 16254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.833144317190462e-06,
+      "loss": 1.3359,
+      "step": 16255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.832108043576532e-06,
+      "loss": 1.1631,
+      "step": 16256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.831071771766032e-06,
+      "loss": 1.1582,
+      "step": 16257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.830035501770104e-06,
+      "loss": 1.1533,
+      "step": 16258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82899923359987e-06,
+      "loss": 0.6265,
+      "step": 16259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82796296726647e-06,
+      "loss": 1.1279,
+      "step": 16260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.826926702781024e-06,
+      "loss": 0.7666,
+      "step": 16261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.825890440154668e-06,
+      "loss": 1.6816,
+      "step": 16262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.824854179398536e-06,
+      "loss": 1.1816,
+      "step": 16263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.823817920523757e-06,
+      "loss": 1.5479,
+      "step": 16264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.822781663541465e-06,
+      "loss": 0.8027,
+      "step": 16265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.821745408462787e-06,
+      "loss": 1.167,
+      "step": 16266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.820709155298856e-06,
+      "loss": 1.0039,
+      "step": 16267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.819672904060798e-06,
+      "loss": 1.0781,
+      "step": 16268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.818636654759753e-06,
+      "loss": 0.9819,
+      "step": 16269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.817600407406848e-06,
+      "loss": 0.9912,
+      "step": 16270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.816564162013212e-06,
+      "loss": 1.3105,
+      "step": 16271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.815527918589979e-06,
+      "loss": 0.8555,
+      "step": 16272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.814491677148277e-06,
+      "loss": 0.833,
+      "step": 16273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.813455437699238e-06,
+      "loss": 0.8252,
+      "step": 16274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812419200253993e-06,
+      "loss": 0.9941,
+      "step": 16275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.811382964823676e-06,
+      "loss": 0.7681,
+      "step": 16276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.810346731419415e-06,
+      "loss": 1.0107,
+      "step": 16277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.809310500052336e-06,
+      "loss": 0.7832,
+      "step": 16278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80827427073358e-06,
+      "loss": 0.8838,
+      "step": 16279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.807238043474271e-06,
+      "loss": 1.1562,
+      "step": 16280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.806201818285542e-06,
+      "loss": 1.3105,
+      "step": 16281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.805165595178524e-06,
+      "loss": 0.8242,
+      "step": 16282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.804129374164342e-06,
+      "loss": 0.937,
+      "step": 16283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.803093155254136e-06,
+      "loss": 1.1865,
+      "step": 16284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.802056938459033e-06,
+      "loss": 1.0029,
+      "step": 16285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80102072379016e-06,
+      "loss": 0.832,
+      "step": 16286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.799984511258654e-06,
+      "loss": 0.7842,
+      "step": 16287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.798948300875639e-06,
+      "loss": 1.166,
+      "step": 16288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79791209265225e-06,
+      "loss": 1.0303,
+      "step": 16289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.796875886599616e-06,
+      "loss": 0.8872,
+      "step": 16290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79583968272887e-06,
+      "loss": 1.0107,
+      "step": 16291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.794803481051137e-06,
+      "loss": 0.8809,
+      "step": 16292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.793767281577552e-06,
+      "loss": 1.2822,
+      "step": 16293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.792731084319243e-06,
+      "loss": 0.6973,
+      "step": 16294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.791694889287347e-06,
+      "loss": 0.9248,
+      "step": 16295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.790658696492986e-06,
+      "loss": 0.7417,
+      "step": 16296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.789622505947295e-06,
+      "loss": 0.8167,
+      "step": 16297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.788586317661404e-06,
+      "loss": 1.4785,
+      "step": 16298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.787550131646436e-06,
+      "loss": 1.1172,
+      "step": 16299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.786513947913534e-06,
+      "loss": 1.0684,
+      "step": 16300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78547776647382e-06,
+      "loss": 1.0869,
+      "step": 16301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78444158733843e-06,
+      "loss": 1.0938,
+      "step": 16302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.783405410518488e-06,
+      "loss": 1.4844,
+      "step": 16303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.782369236025124e-06,
+      "loss": 1.2305,
+      "step": 16304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.781333063869473e-06,
+      "loss": 1.168,
+      "step": 16305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.780296894062667e-06,
+      "loss": 0.9106,
+      "step": 16306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77926072661583e-06,
+      "loss": 0.7288,
+      "step": 16307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.778224561540094e-06,
+      "loss": 1.1758,
+      "step": 16308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.777188398846587e-06,
+      "loss": 1.0957,
+      "step": 16309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.776152238546447e-06,
+      "loss": 1.2119,
+      "step": 16310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.775116080650797e-06,
+      "loss": 0.9917,
+      "step": 16311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.774079925170769e-06,
+      "loss": 0.7495,
+      "step": 16312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.773043772117495e-06,
+      "loss": 1.2354,
+      "step": 16313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.772007621502095e-06,
+      "loss": 0.6694,
+      "step": 16314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.770971473335715e-06,
+      "loss": 0.9575,
+      "step": 16315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.769935327629474e-06,
+      "loss": 1.0688,
+      "step": 16316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.768899184394505e-06,
+      "loss": 1.0771,
+      "step": 16317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.76786304364194e-06,
+      "loss": 1.4092,
+      "step": 16318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.766826905382901e-06,
+      "loss": 0.9834,
+      "step": 16319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.765790769628525e-06,
+      "loss": 0.832,
+      "step": 16320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.764754636389942e-06,
+      "loss": 1.373,
+      "step": 16321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.763718505678282e-06,
+      "loss": 0.6548,
+      "step": 16322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.762682377504668e-06,
+      "loss": 1.251,
+      "step": 16323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.761646251880237e-06,
+      "loss": 1.1289,
+      "step": 16324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.760610128816113e-06,
+      "loss": 0.9302,
+      "step": 16325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.759574008323431e-06,
+      "loss": 0.6475,
+      "step": 16326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.758537890413319e-06,
+      "loss": 1.2686,
+      "step": 16327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.757501775096904e-06,
+      "loss": 0.8389,
+      "step": 16328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.756465662385321e-06,
+      "loss": 1.1729,
+      "step": 16329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75542955228969e-06,
+      "loss": 0.7695,
+      "step": 16330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.754393444821148e-06,
+      "loss": 0.8027,
+      "step": 16331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.753357339990825e-06,
+      "loss": 0.9385,
+      "step": 16332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75232123780985e-06,
+      "loss": 0.5972,
+      "step": 16333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.751285138289348e-06,
+      "loss": 0.9941,
+      "step": 16334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.750249041440451e-06,
+      "loss": 1.0107,
+      "step": 16335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.74921294727429e-06,
+      "loss": 1.0586,
+      "step": 16336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.748176855801995e-06,
+      "loss": 0.8398,
+      "step": 16337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.747140767034692e-06,
+      "loss": 1.0215,
+      "step": 16338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.746104680983511e-06,
+      "loss": 0.9092,
+      "step": 16339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.745068597659578e-06,
+      "loss": 0.5454,
+      "step": 16340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.744032517074032e-06,
+      "loss": 0.9004,
+      "step": 16341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.742996439237995e-06,
+      "loss": 0.8477,
+      "step": 16342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.741960364162596e-06,
+      "loss": 1.1826,
+      "step": 16343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.740924291858969e-06,
+      "loss": 0.8525,
+      "step": 16344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.739888222338233e-06,
+      "loss": 0.7021,
+      "step": 16345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.738852155611531e-06,
+      "loss": 0.9302,
+      "step": 16346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73781609168998e-06,
+      "loss": 0.7734,
+      "step": 16347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.736780030584718e-06,
+      "loss": 0.9595,
+      "step": 16348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.735743972306869e-06,
+      "loss": 1.0459,
+      "step": 16349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.734707916867559e-06,
+      "loss": 1.2344,
+      "step": 16350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.733671864277923e-06,
+      "loss": 1.0645,
+      "step": 16351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.732635814549088e-06,
+      "loss": 0.9092,
+      "step": 16352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.731599767692187e-06,
+      "loss": 0.7822,
+      "step": 16353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.730563723718338e-06,
+      "loss": 1.3506,
+      "step": 16354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.729527682638681e-06,
+      "loss": 0.7146,
+      "step": 16355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.728491644464334e-06,
+      "loss": 0.9121,
+      "step": 16356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.727455609206438e-06,
+      "loss": 0.8857,
+      "step": 16357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.726419576876113e-06,
+      "loss": 0.9761,
+      "step": 16358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.725383547484492e-06,
+      "loss": 1.0332,
+      "step": 16359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7243475210427e-06,
+      "loss": 1.0088,
+      "step": 16360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.723311497561865e-06,
+      "loss": 0.7156,
+      "step": 16361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.722275477053122e-06,
+      "loss": 1.2324,
+      "step": 16362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.721239459527594e-06,
+      "loss": 0.8726,
+      "step": 16363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.720203444996413e-06,
+      "loss": 0.998,
+      "step": 16364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.719167433470705e-06,
+      "loss": 1.0088,
+      "step": 16365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.718131424961594e-06,
+      "loss": 0.7979,
+      "step": 16366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.717095419480218e-06,
+      "loss": 1.1504,
+      "step": 16367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.716059417037703e-06,
+      "loss": 0.9507,
+      "step": 16368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.715023417645172e-06,
+      "loss": 1.1973,
+      "step": 16369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71398742131376e-06,
+      "loss": 1.2568,
+      "step": 16370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.712951428054586e-06,
+      "loss": 0.9219,
+      "step": 16371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71191543787879e-06,
+      "loss": 1.126,
+      "step": 16372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.710879450797492e-06,
+      "loss": 1.2754,
+      "step": 16373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.709843466821822e-06,
+      "loss": 1.1504,
+      "step": 16374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70880748596291e-06,
+      "loss": 1.041,
+      "step": 16375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.707771508231882e-06,
+      "loss": 0.9219,
+      "step": 16376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.706735533639868e-06,
+      "loss": 0.9805,
+      "step": 16377
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.705699562197993e-06,
+      "loss": 0.7358,
+      "step": 16378
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.704663593917391e-06,
+      "loss": 1.1865,
+      "step": 16379
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.703627628809183e-06,
+      "loss": 1.082,
+      "step": 16380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.702591666884499e-06,
+      "loss": 0.9111,
+      "step": 16381
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70155570815447e-06,
+      "loss": 1.2715,
+      "step": 16382
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.700519752630224e-06,
+      "loss": 0.8403,
+      "step": 16383
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.699483800322884e-06,
+      "loss": 0.9146,
+      "step": 16384
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69844785124358e-06,
+      "loss": 0.8813,
+      "step": 16385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.697411905403444e-06,
+      "loss": 1.0996,
+      "step": 16386
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.696375962813595e-06,
+      "loss": 1.2998,
+      "step": 16387
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.695340023485169e-06,
+      "loss": 0.8662,
+      "step": 16388
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69430408742929e-06,
+      "loss": 1.3633,
+      "step": 16389
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.693268154657088e-06,
+      "loss": 0.8887,
+      "step": 16390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.692232225179687e-06,
+      "loss": 1.2695,
+      "step": 16391
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.691196299008214e-06,
+      "loss": 1.1113,
+      "step": 16392
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.690160376153803e-06,
+      "loss": 1.3594,
+      "step": 16393
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.689124456627577e-06,
+      "loss": 1.3086,
+      "step": 16394
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.688088540440663e-06,
+      "loss": 0.8296,
+      "step": 16395
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.687052627604188e-06,
+      "loss": 0.7677,
+      "step": 16396
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68601671812928e-06,
+      "loss": 1.1025,
+      "step": 16397
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.684980812027068e-06,
+      "loss": 1.5547,
+      "step": 16398
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.683944909308682e-06,
+      "loss": 1.1182,
+      "step": 16399
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.682909009985243e-06,
+      "loss": 0.8789,
+      "step": 16400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.681873114067883e-06,
+      "loss": 0.8672,
+      "step": 16401
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.680837221567722e-06,
+      "loss": 0.9951,
+      "step": 16402
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.679801332495897e-06,
+      "loss": 0.8755,
+      "step": 16403
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.67876544686353e-06,
+      "loss": 0.958,
+      "step": 16404
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.67772956468175e-06,
+      "loss": 1.0928,
+      "step": 16405
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.67669368596168e-06,
+      "loss": 1.502,
+      "step": 16406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.675657810714448e-06,
+      "loss": 0.8828,
+      "step": 16407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.674621938951185e-06,
+      "loss": 1.4531,
+      "step": 16408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.673586070683016e-06,
+      "loss": 1.0586,
+      "step": 16409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.672550205921071e-06,
+      "loss": 1.2803,
+      "step": 16410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.671514344676468e-06,
+      "loss": 0.9468,
+      "step": 16411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.67047848696034e-06,
+      "loss": 0.8066,
+      "step": 16412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.669442632783816e-06,
+      "loss": 0.9414,
+      "step": 16413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.668406782158021e-06,
+      "loss": 0.8975,
+      "step": 16414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.667370935094078e-06,
+      "loss": 1.2188,
+      "step": 16415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.666335091603117e-06,
+      "loss": 1.3125,
+      "step": 16416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.665299251696267e-06,
+      "loss": 0.8789,
+      "step": 16417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.664263415384644e-06,
+      "loss": 1.3643,
+      "step": 16418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.663227582679387e-06,
+      "loss": 1.0361,
+      "step": 16419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.662191753591619e-06,
+      "loss": 1.1289,
+      "step": 16420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.661155928132466e-06,
+      "loss": 0.7539,
+      "step": 16421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.66012010631305e-06,
+      "loss": 1.1924,
+      "step": 16422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.659084288144503e-06,
+      "loss": 0.9248,
+      "step": 16423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.658048473637948e-06,
+      "loss": 1.0107,
+      "step": 16424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.657012662804517e-06,
+      "loss": 0.8154,
+      "step": 16425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.655976855655329e-06,
+      "loss": 1.1602,
+      "step": 16426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.654941052201514e-06,
+      "loss": 0.8984,
+      "step": 16427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.653905252454195e-06,
+      "loss": 0.8174,
+      "step": 16428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.652869456424505e-06,
+      "loss": 1.0498,
+      "step": 16429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.651833664123567e-06,
+      "loss": 1.168,
+      "step": 16430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.650797875562504e-06,
+      "loss": 0.98,
+      "step": 16431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.649762090752446e-06,
+      "loss": 0.8101,
+      "step": 16432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.648726309704513e-06,
+      "loss": 0.8057,
+      "step": 16433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.647690532429837e-06,
+      "loss": 0.8545,
+      "step": 16434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.646654758939543e-06,
+      "loss": 0.7734,
+      "step": 16435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.645618989244758e-06,
+      "loss": 1.0332,
+      "step": 16436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.644583223356604e-06,
+      "loss": 0.8516,
+      "step": 16437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.643547461286207e-06,
+      "loss": 0.6621,
+      "step": 16438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.642511703044696e-06,
+      "loss": 1.0181,
+      "step": 16439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.641475948643198e-06,
+      "loss": 0.8779,
+      "step": 16440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.640440198092834e-06,
+      "loss": 0.9678,
+      "step": 16441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.639404451404729e-06,
+      "loss": 1.208,
+      "step": 16442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.638368708590011e-06,
+      "loss": 1.1133,
+      "step": 16443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.637332969659809e-06,
+      "loss": 1.0303,
+      "step": 16444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.636297234625248e-06,
+      "loss": 1.0146,
+      "step": 16445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.635261503497447e-06,
+      "loss": 0.9854,
+      "step": 16446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634225776287538e-06,
+      "loss": 1.1211,
+      "step": 16447
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.633190053006641e-06,
+      "loss": 1.1562,
+      "step": 16448
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.632154333665882e-06,
+      "loss": 1.0723,
+      "step": 16449
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.631118618276391e-06,
+      "loss": 0.6006,
+      "step": 16450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.630082906849294e-06,
+      "loss": 0.9424,
+      "step": 16451
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.629047199395709e-06,
+      "loss": 0.4243,
+      "step": 16452
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.628011495926766e-06,
+      "loss": 0.9893,
+      "step": 16453
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.626975796453587e-06,
+      "loss": 0.9521,
+      "step": 16454
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.625940100987301e-06,
+      "loss": 1.2617,
+      "step": 16455
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.624904409539035e-06,
+      "loss": 1.1621,
+      "step": 16456
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.623868722119908e-06,
+      "loss": 1.1816,
+      "step": 16457
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.622833038741048e-06,
+      "loss": 0.8525,
+      "step": 16458
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.621797359413574e-06,
+      "loss": 0.8018,
+      "step": 16459
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.620761684148624e-06,
+      "loss": 1.0664,
+      "step": 16460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.619726012957312e-06,
+      "loss": 1.0762,
+      "step": 16461
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.618690345850765e-06,
+      "loss": 0.8467,
+      "step": 16462
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.617654682840112e-06,
+      "loss": 1.209,
+      "step": 16463
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.616619023936469e-06,
+      "loss": 0.9492,
+      "step": 16464
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61558336915097e-06,
+      "loss": 0.6763,
+      "step": 16465
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.614547718494736e-06,
+      "loss": 1.0449,
+      "step": 16466
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.613512071978891e-06,
+      "loss": 1.0176,
+      "step": 16467
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61247642961456e-06,
+      "loss": 1.1582,
+      "step": 16468
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.611440791412864e-06,
+      "loss": 1.373,
+      "step": 16469
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.610405157384933e-06,
+      "loss": 0.8008,
+      "step": 16470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.609369527541893e-06,
+      "loss": 0.793,
+      "step": 16471
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.608333901894859e-06,
+      "loss": 1.4775,
+      "step": 16472
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.607298280454963e-06,
+      "loss": 1.0029,
+      "step": 16473
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.606262663233326e-06,
+      "loss": 1.1855,
+      "step": 16474
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.605227050241078e-06,
+      "loss": 1.0645,
+      "step": 16475
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.604191441489334e-06,
+      "loss": 1.1274,
+      "step": 16476
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.603155836989224e-06,
+      "loss": 0.9395,
+      "step": 16477
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.602120236751872e-06,
+      "loss": 1.1221,
+      "step": 16478
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.6010846407884e-06,
+      "loss": 0.8135,
+      "step": 16479
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.60004904910993e-06,
+      "loss": 1.0957,
+      "step": 16480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.599013461727591e-06,
+      "loss": 1.0457,
+      "step": 16481
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.597977878652508e-06,
+      "loss": 0.8462,
+      "step": 16482
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.596942299895799e-06,
+      "loss": 0.8994,
+      "step": 16483
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59590672546859e-06,
+      "loss": 1.5205,
+      "step": 16484
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.594871155382003e-06,
+      "loss": 1.0215,
+      "step": 16485
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.593835589647169e-06,
+      "loss": 1.0137,
+      "step": 16486
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.592800028275202e-06,
+      "loss": 1.042,
+      "step": 16487
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.591764471277231e-06,
+      "loss": 1.4707,
+      "step": 16488
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.590728918664383e-06,
+      "loss": 1.2324,
+      "step": 16489
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.589693370447769e-06,
+      "loss": 0.9639,
+      "step": 16490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.588657826638528e-06,
+      "loss": 1.3594,
+      "step": 16491
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.587622287247773e-06,
+      "loss": 0.9407,
+      "step": 16492
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.586586752286634e-06,
+      "loss": 0.8076,
+      "step": 16493
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.585551221766229e-06,
+      "loss": 0.9531,
+      "step": 16494
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.584515695697681e-06,
+      "loss": 0.9297,
+      "step": 16495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.583480174092116e-06,
+      "loss": 1.1162,
+      "step": 16496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.582444656960658e-06,
+      "loss": 0.8521,
+      "step": 16497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58140914431443e-06,
+      "loss": 1.0342,
+      "step": 16498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.580373636164553e-06,
+      "loss": 1.1797,
+      "step": 16499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.579338132522147e-06,
+      "loss": 1.2695,
+      "step": 16500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.578302633398343e-06,
+      "loss": 1.1357,
+      "step": 16501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.577267138804262e-06,
+      "loss": 0.6321,
+      "step": 16502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.576231648751021e-06,
+      "loss": 1.1152,
+      "step": 16503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.575196163249751e-06,
+      "loss": 1.2148,
+      "step": 16504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.574160682311564e-06,
+      "loss": 0.9443,
+      "step": 16505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.573125205947594e-06,
+      "loss": 1.3457,
+      "step": 16506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.572089734168959e-06,
+      "loss": 1.084,
+      "step": 16507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57105426698678e-06,
+      "loss": 0.9038,
+      "step": 16508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.570018804412185e-06,
+      "loss": 0.9648,
+      "step": 16509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.568983346456289e-06,
+      "loss": 1.0586,
+      "step": 16510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.567947893130217e-06,
+      "loss": 0.8389,
+      "step": 16511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.566912444445095e-06,
+      "loss": 1.2012,
+      "step": 16512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.565877000412046e-06,
+      "loss": 0.9053,
+      "step": 16513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.564841561042187e-06,
+      "loss": 0.522,
+      "step": 16514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.563806126346643e-06,
+      "loss": 0.9917,
+      "step": 16515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.562770696336534e-06,
+      "loss": 1.1304,
+      "step": 16516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.561735271022987e-06,
+      "loss": 0.9238,
+      "step": 16517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56069985041712e-06,
+      "loss": 0.7988,
+      "step": 16518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.55966443453006e-06,
+      "loss": 0.6821,
+      "step": 16519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.558629023372925e-06,
+      "loss": 1.0322,
+      "step": 16520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.557593616956834e-06,
+      "loss": 0.873,
+      "step": 16521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556558215292915e-06,
+      "loss": 0.9707,
+      "step": 16522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.555522818392287e-06,
+      "loss": 0.9766,
+      "step": 16523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.554487426266076e-06,
+      "loss": 0.8711,
+      "step": 16524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.553452038925397e-06,
+      "loss": 1.0078,
+      "step": 16525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.552416656381373e-06,
+      "loss": 0.7378,
+      "step": 16526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.551381278645131e-06,
+      "loss": 1.0071,
+      "step": 16527
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.550345905727791e-06,
+      "loss": 1.0469,
+      "step": 16528
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.549310537640471e-06,
+      "loss": 0.7471,
+      "step": 16529
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.548275174394295e-06,
+      "loss": 1.1562,
+      "step": 16530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.547239816000381e-06,
+      "loss": 0.9634,
+      "step": 16531
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.546204462469858e-06,
+      "loss": 1.1162,
+      "step": 16532
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54516911381384e-06,
+      "loss": 0.6416,
+      "step": 16533
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.544133770043454e-06,
+      "loss": 1.5371,
+      "step": 16534
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.543098431169817e-06,
+      "loss": 0.916,
+      "step": 16535
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.542063097204048e-06,
+      "loss": 0.9314,
+      "step": 16536
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.541027768157278e-06,
+      "loss": 1.2598,
+      "step": 16537
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.53999244404062e-06,
+      "loss": 0.915,
+      "step": 16538
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5389571248652e-06,
+      "loss": 0.8813,
+      "step": 16539
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.537921810642133e-06,
+      "loss": 0.999,
+      "step": 16540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.536886501382542e-06,
+      "loss": 1.3662,
+      "step": 16541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.535851197097549e-06,
+      "loss": 0.9209,
+      "step": 16542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.534815897798276e-06,
+      "loss": 1.166,
+      "step": 16543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.533780603495846e-06,
+      "loss": 0.7549,
+      "step": 16544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532745314201372e-06,
+      "loss": 1.001,
+      "step": 16545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.531710029925985e-06,
+      "loss": 0.877,
+      "step": 16546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.53067475068079e-06,
+      "loss": 0.7388,
+      "step": 16547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.529639476476927e-06,
+      "loss": 0.8408,
+      "step": 16548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.528604207325503e-06,
+      "loss": 1.0708,
+      "step": 16549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.527568943237646e-06,
+      "loss": 0.9746,
+      "step": 16550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.526533684224468e-06,
+      "loss": 0.9932,
+      "step": 16551
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.525498430297094e-06,
+      "loss": 0.7666,
+      "step": 16552
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.524463181466648e-06,
+      "loss": 1.0283,
+      "step": 16553
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.523427937744245e-06,
+      "loss": 1.1143,
+      "step": 16554
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52239269914101e-06,
+      "loss": 0.9121,
+      "step": 16555
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52135746566806e-06,
+      "loss": 0.6499,
+      "step": 16556
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52032223733651e-06,
+      "loss": 1.4033,
+      "step": 16557
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519287014157488e-06,
+      "loss": 1.0703,
+      "step": 16558
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.518251796142114e-06,
+      "loss": 0.9775,
+      "step": 16559
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.517216583301505e-06,
+      "loss": 0.9189,
+      "step": 16560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51618137564678e-06,
+      "loss": 1.166,
+      "step": 16561
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.515146173189058e-06,
+      "loss": 1.0283,
+      "step": 16562
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.514110975939466e-06,
+      "loss": 1.2158,
+      "step": 16563
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.513075783909117e-06,
+      "loss": 0.7861,
+      "step": 16564
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51204059710913e-06,
+      "loss": 0.9951,
+      "step": 16565
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51100541555063e-06,
+      "loss": 1.2412,
+      "step": 16566
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.509970239244729e-06,
+      "loss": 1.2891,
+      "step": 16567
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.508935068202553e-06,
+      "loss": 1.2188,
+      "step": 16568
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.507899902435219e-06,
+      "loss": 0.958,
+      "step": 16569
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50686474195385e-06,
+      "loss": 0.998,
+      "step": 16570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.505829586769557e-06,
+      "loss": 1.1006,
+      "step": 16571
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.504794436893468e-06,
+      "loss": 1.0752,
+      "step": 16572
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.503759292336694e-06,
+      "loss": 1.0864,
+      "step": 16573
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.502724153110364e-06,
+      "loss": 0.8018,
+      "step": 16574
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.501689019225589e-06,
+      "loss": 1.5879,
+      "step": 16575
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.500653890693493e-06,
+      "loss": 1.0605,
+      "step": 16576
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.499618767525192e-06,
+      "loss": 1.0151,
+      "step": 16577
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.498583649731802e-06,
+      "loss": 1.0869,
+      "step": 16578
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.497548537324449e-06,
+      "loss": 1.4521,
+      "step": 16579
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.496513430314245e-06,
+      "loss": 1.0801,
+      "step": 16580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.495478328712317e-06,
+      "loss": 1.2607,
+      "step": 16581
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.494443232529776e-06,
+      "loss": 0.9248,
+      "step": 16582
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.493408141777742e-06,
+      "loss": 0.7524,
+      "step": 16583
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.492373056467336e-06,
+      "loss": 0.8076,
+      "step": 16584
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.491337976609679e-06,
+      "loss": 1.1523,
+      "step": 16585
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.490302902215882e-06,
+      "loss": 1.1396,
+      "step": 16586
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.489267833297068e-06,
+      "loss": 1.1553,
+      "step": 16587
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.488232769864352e-06,
+      "loss": 1.25,
+      "step": 16588
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.487197711928859e-06,
+      "loss": 1.373,
+      "step": 16589
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.486162659501703e-06,
+      "loss": 0.9583,
+      "step": 16590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485127612594001e-06,
+      "loss": 1.2891,
+      "step": 16591
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.484092571216876e-06,
+      "loss": 1.1357,
+      "step": 16592
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.483057535381435e-06,
+      "loss": 1.0498,
+      "step": 16593
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.48202250509881e-06,
+      "loss": 1.0557,
+      "step": 16594
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.480987480380108e-06,
+      "loss": 1.0713,
+      "step": 16595
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.479952461236453e-06,
+      "loss": 1.2881,
+      "step": 16596
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.478917447678964e-06,
+      "loss": 0.7549,
+      "step": 16597
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47788243971875e-06,
+      "loss": 1.2266,
+      "step": 16598
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.476847437366939e-06,
+      "loss": 1.3193,
+      "step": 16599
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.475812440634641e-06,
+      "loss": 1.0752,
+      "step": 16600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474777449532982e-06,
+      "loss": 0.959,
+      "step": 16601
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.473742464073069e-06,
+      "loss": 0.9854,
+      "step": 16602
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.472707484266028e-06,
+      "loss": 0.9141,
+      "step": 16603
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.471672510122966e-06,
+      "loss": 0.782,
+      "step": 16604
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.470637541655015e-06,
+      "loss": 0.9355,
+      "step": 16605
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46960257887328e-06,
+      "loss": 1.1309,
+      "step": 16606
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.468567621788886e-06,
+      "loss": 0.9844,
+      "step": 16607
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.467532670412947e-06,
+      "loss": 0.96,
+      "step": 16608
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.466497724756575e-06,
+      "loss": 0.8975,
+      "step": 16609
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.465462784830895e-06,
+      "loss": 1.4844,
+      "step": 16610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.464427850647021e-06,
+      "loss": 0.9434,
+      "step": 16611
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46339292221607e-06,
+      "loss": 0.7983,
+      "step": 16612
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46235799954916e-06,
+      "loss": 0.9463,
+      "step": 16613
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.461323082657402e-06,
+      "loss": 1.1367,
+      "step": 16614
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46028817155192e-06,
+      "loss": 1.125,
+      "step": 16615
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45925326624383e-06,
+      "loss": 0.8477,
+      "step": 16616
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458218366744244e-06,
+      "loss": 0.9893,
+      "step": 16617
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457183473064279e-06,
+      "loss": 0.9072,
+      "step": 16618
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.456148585215054e-06,
+      "loss": 0.9678,
+      "step": 16619
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.455113703207687e-06,
+      "loss": 1.0752,
+      "step": 16620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454078827053291e-06,
+      "loss": 1.2734,
+      "step": 16621
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.453043956762982e-06,
+      "loss": 0.5859,
+      "step": 16622
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45200909234788e-06,
+      "loss": 1.0371,
+      "step": 16623
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450974233819093e-06,
+      "loss": 1.0039,
+      "step": 16624
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44993938118775e-06,
+      "loss": 1.2637,
+      "step": 16625
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448904534464957e-06,
+      "loss": 1.6328,
+      "step": 16626
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.447869693661833e-06,
+      "loss": 0.8462,
+      "step": 16627
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446834858789494e-06,
+      "loss": 0.9854,
+      "step": 16628
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.445800029859052e-06,
+      "loss": 1.1582,
+      "step": 16629
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.444765206881627e-06,
+      "loss": 1.3008,
+      "step": 16630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443730389868337e-06,
+      "loss": 1.3203,
+      "step": 16631
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442695578830291e-06,
+      "loss": 1.2871,
+      "step": 16632
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44166077377861e-06,
+      "loss": 1.3047,
+      "step": 16633
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440625974724408e-06,
+      "loss": 1.2061,
+      "step": 16634
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439591181678795e-06,
+      "loss": 1.0684,
+      "step": 16635
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438556394652897e-06,
+      "loss": 0.7197,
+      "step": 16636
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437521613657821e-06,
+      "loss": 1.0498,
+      "step": 16637
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436486838704687e-06,
+      "loss": 1.1338,
+      "step": 16638
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435452069804609e-06,
+      "loss": 1.0254,
+      "step": 16639
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434417306968695e-06,
+      "loss": 1.0605,
+      "step": 16640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43338255020807e-06,
+      "loss": 0.6509,
+      "step": 16641
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.432347799533845e-06,
+      "loss": 1.2031,
+      "step": 16642
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.431313054957139e-06,
+      "loss": 0.8398,
+      "step": 16643
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.430278316489058e-06,
+      "loss": 1.0615,
+      "step": 16644
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.429243584140721e-06,
+      "loss": 0.9663,
+      "step": 16645
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.428208857923246e-06,
+      "loss": 1.0215,
+      "step": 16646
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.427174137847747e-06,
+      "loss": 1.0625,
+      "step": 16647
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.426139423925334e-06,
+      "loss": 0.8052,
+      "step": 16648
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.425104716167128e-06,
+      "loss": 0.7432,
+      "step": 16649
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.424070014584232e-06,
+      "loss": 1.165,
+      "step": 16650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.423035319187773e-06,
+      "loss": 0.8574,
+      "step": 16651
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.422000629988861e-06,
+      "loss": 1.4873,
+      "step": 16652
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.420965946998608e-06,
+      "loss": 0.999,
+      "step": 16653
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.419931270228133e-06,
+      "loss": 0.9165,
+      "step": 16654
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.418896599688542e-06,
+      "loss": 1.1738,
+      "step": 16655
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.417861935390957e-06,
+      "loss": 1.0449,
+      "step": 16656
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.416827277346488e-06,
+      "loss": 0.7632,
+      "step": 16657
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.415792625566253e-06,
+      "loss": 1.0664,
+      "step": 16658
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.414757980061361e-06,
+      "loss": 0.7183,
+      "step": 16659
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.413723340842926e-06,
+      "loss": 0.9326,
+      "step": 16660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.412688707922063e-06,
+      "loss": 0.8228,
+      "step": 16661
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.411654081309888e-06,
+      "loss": 0.7188,
+      "step": 16662
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.410619461017514e-06,
+      "loss": 1.2754,
+      "step": 16663
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.409584847056051e-06,
+      "loss": 1.1572,
+      "step": 16664
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.408550239436615e-06,
+      "loss": 1.145,
+      "step": 16665
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.407515638170317e-06,
+      "loss": 0.8057,
+      "step": 16666
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.406481043268274e-06,
+      "loss": 0.8447,
+      "step": 16667
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.405446454741597e-06,
+      "loss": 1.5977,
+      "step": 16668
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.404411872601402e-06,
+      "loss": 0.876,
+      "step": 16669
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.403377296858798e-06,
+      "loss": 1.1934,
+      "step": 16670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402342727524897e-06,
+      "loss": 1.1113,
+      "step": 16671
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.401308164610817e-06,
+      "loss": 0.7432,
+      "step": 16672
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400273608127672e-06,
+      "loss": 1.0879,
+      "step": 16673
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.399239058086568e-06,
+      "loss": 0.9917,
+      "step": 16674
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39820451449862e-06,
+      "loss": 1.3457,
+      "step": 16675
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.397169977374942e-06,
+      "loss": 1.0117,
+      "step": 16676
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39613544672665e-06,
+      "loss": 1.4248,
+      "step": 16677
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.395100922564853e-06,
+      "loss": 0.9355,
+      "step": 16678
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.394066404900662e-06,
+      "loss": 1.2148,
+      "step": 16679
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.393031893745194e-06,
+      "loss": 1.207,
+      "step": 16680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39199738910955e-06,
+      "loss": 0.9971,
+      "step": 16681
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.390962891004858e-06,
+      "loss": 1.3164,
+      "step": 16682
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.389928399442221e-06,
+      "loss": 0.9492,
+      "step": 16683
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.388893914432757e-06,
+      "loss": 0.8066,
+      "step": 16684
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38785943598757e-06,
+      "loss": 0.9902,
+      "step": 16685
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.386824964117773e-06,
+      "loss": 0.6875,
+      "step": 16686
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.385790498834484e-06,
+      "loss": 1.0762,
+      "step": 16687
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.384756040148813e-06,
+      "loss": 1.0713,
+      "step": 16688
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.383721588071872e-06,
+      "loss": 0.9824,
+      "step": 16689
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.382687142614768e-06,
+      "loss": 0.8389,
+      "step": 16690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38165270378862e-06,
+      "loss": 1.3076,
+      "step": 16691
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.380618271604529e-06,
+      "loss": 0.8613,
+      "step": 16692
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.379583846073617e-06,
+      "loss": 1.1064,
+      "step": 16693
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.378549427206992e-06,
+      "loss": 1.4121,
+      "step": 16694
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.377515015015765e-06,
+      "loss": 1.0107,
+      "step": 16695
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.376480609511047e-06,
+      "loss": 0.9473,
+      "step": 16696
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.375446210703946e-06,
+      "loss": 1.2002,
+      "step": 16697
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37441181860558e-06,
+      "loss": 1.3369,
+      "step": 16698
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.373377433227054e-06,
+      "loss": 0.9543,
+      "step": 16699
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.372343054579485e-06,
+      "loss": 1.124,
+      "step": 16700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37130868267398e-06,
+      "loss": 0.9785,
+      "step": 16701
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.370274317521645e-06,
+      "loss": 0.6782,
+      "step": 16702
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.3692399591336e-06,
+      "loss": 1.0811,
+      "step": 16703
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.368205607520954e-06,
+      "loss": 0.9346,
+      "step": 16704
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.367171262694814e-06,
+      "loss": 0.9561,
+      "step": 16705
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.366136924666289e-06,
+      "loss": 0.9023,
+      "step": 16706
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.365102593446492e-06,
+      "loss": 1.167,
+      "step": 16707
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36406826904654e-06,
+      "loss": 0.9336,
+      "step": 16708
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.363033951477533e-06,
+      "loss": 1.1504,
+      "step": 16709
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.361999640750585e-06,
+      "loss": 0.9902,
+      "step": 16710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.360965336876809e-06,
+      "loss": 1.1475,
+      "step": 16711
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.35993103986731e-06,
+      "loss": 1.1055,
+      "step": 16712
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.358896749733202e-06,
+      "loss": 1.1211,
+      "step": 16713
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.357862466485594e-06,
+      "loss": 0.9209,
+      "step": 16714
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.356828190135598e-06,
+      "loss": 1.0801,
+      "step": 16715
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.355793920694319e-06,
+      "loss": 1.1514,
+      "step": 16716
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.354759658172867e-06,
+      "loss": 1.126,
+      "step": 16717
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.353725402582356e-06,
+      "loss": 1.0957,
+      "step": 16718
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.352691153933896e-06,
+      "loss": 1.2148,
+      "step": 16719
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.351656912238592e-06,
+      "loss": 1.2402,
+      "step": 16720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.350622677507556e-06,
+      "loss": 0.917,
+      "step": 16721
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.349588449751898e-06,
+      "loss": 0.4868,
+      "step": 16722
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.348554228982721e-06,
+      "loss": 1.0742,
+      "step": 16723
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.347520015211144e-06,
+      "loss": 0.7666,
+      "step": 16724
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34648580844827e-06,
+      "loss": 0.9458,
+      "step": 16725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.345451608705215e-06,
+      "loss": 0.7588,
+      "step": 16726
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.344417415993078e-06,
+      "loss": 1.0908,
+      "step": 16727
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.343383230322969e-06,
+      "loss": 1.0352,
+      "step": 16728
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.342349051706004e-06,
+      "loss": 0.5688,
+      "step": 16729
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34131488015329e-06,
+      "loss": 0.8037,
+      "step": 16730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.340280715675933e-06,
+      "loss": 0.7783,
+      "step": 16731
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33924655828504e-06,
+      "loss": 0.8755,
+      "step": 16732
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.338212407991721e-06,
+      "loss": 1.3105,
+      "step": 16733
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.337178264807088e-06,
+      "loss": 0.6826,
+      "step": 16734
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.336144128742251e-06,
+      "loss": 1.1367,
+      "step": 16735
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33510999980831e-06,
+      "loss": 1.1533,
+      "step": 16736
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33407587801638e-06,
+      "loss": 1.3213,
+      "step": 16737
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33304176337756e-06,
+      "loss": 0.9551,
+      "step": 16738
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33200765590297e-06,
+      "loss": 0.8652,
+      "step": 16739
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.330973555603713e-06,
+      "loss": 1.1553,
+      "step": 16740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.329939462490896e-06,
+      "loss": 1.127,
+      "step": 16741
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.328905376575629e-06,
+      "loss": 1.084,
+      "step": 16742
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.327871297869014e-06,
+      "loss": 1.042,
+      "step": 16743
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.326837226382166e-06,
+      "loss": 0.8237,
+      "step": 16744
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.325803162126188e-06,
+      "loss": 0.9067,
+      "step": 16745
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.324769105112192e-06,
+      "loss": 0.8999,
+      "step": 16746
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.323735055351281e-06,
+      "loss": 1.1299,
+      "step": 16747
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.322701012854563e-06,
+      "loss": 1.1953,
+      "step": 16748
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.321666977633145e-06,
+      "loss": 1.0996,
+      "step": 16749
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32063294969814e-06,
+      "loss": 0.7949,
+      "step": 16750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.319598929060649e-06,
+      "loss": 1.2285,
+      "step": 16751
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.318564915731782e-06,
+      "loss": 1.02,
+      "step": 16752
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.317530909722644e-06,
+      "loss": 0.8994,
+      "step": 16753
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31649691104434e-06,
+      "loss": 1.248,
+      "step": 16754
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.315462919707981e-06,
+      "loss": 1.2725,
+      "step": 16755
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.314428935724673e-06,
+      "loss": 1.2637,
+      "step": 16756
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.313394959105525e-06,
+      "loss": 1.1709,
+      "step": 16757
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.312360989861638e-06,
+      "loss": 1.3125,
+      "step": 16758
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.311327028004116e-06,
+      "loss": 0.793,
+      "step": 16759
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.310293073544077e-06,
+      "loss": 1.082,
+      "step": 16760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.309259126492621e-06,
+      "loss": 1.002,
+      "step": 16761
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.308225186860852e-06,
+      "loss": 1.3379,
+      "step": 16762
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.307191254659878e-06,
+      "loss": 0.915,
+      "step": 16763
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.306157329900805e-06,
+      "loss": 1.1953,
+      "step": 16764
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.305123412594742e-06,
+      "loss": 0.8916,
+      "step": 16765
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.304089502752792e-06,
+      "loss": 0.8628,
+      "step": 16766
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.303055600386061e-06,
+      "loss": 1.1289,
+      "step": 16767
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.302021705505659e-06,
+      "loss": 1.3311,
+      "step": 16768
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.300987818122678e-06,
+      "loss": 0.8711,
+      "step": 16769
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.299953938248244e-06,
+      "loss": 0.9482,
+      "step": 16770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.298920065893447e-06,
+      "loss": 1.041,
+      "step": 16771
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.297886201069402e-06,
+      "loss": 1.3301,
+      "step": 16772
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.296852343787206e-06,
+      "loss": 1.2129,
+      "step": 16773
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.295818494057968e-06,
+      "loss": 1.6104,
+      "step": 16774
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.294784651892795e-06,
+      "loss": 0.9268,
+      "step": 16775
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.293750817302793e-06,
+      "loss": 0.8906,
+      "step": 16776
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.292716990299062e-06,
+      "loss": 0.9287,
+      "step": 16777
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.291683170892712e-06,
+      "loss": 1.1406,
+      "step": 16778
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.29064935909484e-06,
+      "loss": 0.9824,
+      "step": 16779
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.289615554916562e-06,
+      "loss": 1.1016,
+      "step": 16780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.288581758368978e-06,
+      "loss": 0.6523,
+      "step": 16781
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.287547969463192e-06,
+      "loss": 1.1445,
+      "step": 16782
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.286514188210309e-06,
+      "loss": 1.0977,
+      "step": 16783
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.285480414621431e-06,
+      "loss": 0.7988,
+      "step": 16784
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.284446648707663e-06,
+      "loss": 0.7793,
+      "step": 16785
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.283412890480114e-06,
+      "loss": 0.7656,
+      "step": 16786
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.282379139949882e-06,
+      "loss": 1.2656,
+      "step": 16787
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.28134539712808e-06,
+      "loss": 1.1504,
+      "step": 16788
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.280311662025803e-06,
+      "loss": 0.6118,
+      "step": 16789
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.279277934654157e-06,
+      "loss": 0.9814,
+      "step": 16790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.278244215024248e-06,
+      "loss": 0.7322,
+      "step": 16791
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.277210503147182e-06,
+      "loss": 0.9917,
+      "step": 16792
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.276176799034057e-06,
+      "loss": 1.2441,
+      "step": 16793
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.275143102695983e-06,
+      "loss": 1.0625,
+      "step": 16794
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.274109414144054e-06,
+      "loss": 0.7114,
+      "step": 16795
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.273075733389385e-06,
+      "loss": 1.2168,
+      "step": 16796
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.272042060443074e-06,
+      "loss": 0.8325,
+      "step": 16797
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.271008395316223e-06,
+      "loss": 1.1572,
+      "step": 16798
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.269974738019939e-06,
+      "loss": 1.0938,
+      "step": 16799
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.268941088565319e-06,
+      "loss": 1.1123,
+      "step": 16800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.267907446963473e-06,
+      "loss": 1.1396,
+      "step": 16801
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.266873813225498e-06,
+      "loss": 1.4297,
+      "step": 16802
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.265840187362505e-06,
+      "loss": 1.1025,
+      "step": 16803
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.264806569385589e-06,
+      "loss": 1.1436,
+      "step": 16804
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.263772959305854e-06,
+      "loss": 0.9346,
+      "step": 16805
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.262739357134405e-06,
+      "loss": 1.2168,
+      "step": 16806
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.261705762882348e-06,
+      "loss": 1.0049,
+      "step": 16807
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.260672176560778e-06,
+      "loss": 0.7139,
+      "step": 16808
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2596385981808e-06,
+      "loss": 1.2061,
+      "step": 16809
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.258605027753516e-06,
+      "loss": 1.0518,
+      "step": 16810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.257571465290033e-06,
+      "loss": 0.6958,
+      "step": 16811
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.256537910801448e-06,
+      "loss": 1.0781,
+      "step": 16812
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.255504364298864e-06,
+      "loss": 0.8354,
+      "step": 16813
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.254470825793385e-06,
+      "loss": 1.0879,
+      "step": 16814
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25343729529611e-06,
+      "loss": 1.2988,
+      "step": 16815
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25240377281814e-06,
+      "loss": 0.9688,
+      "step": 16816
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25137025837058e-06,
+      "loss": 1.1523,
+      "step": 16817
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.250336751964534e-06,
+      "loss": 1.1836,
+      "step": 16818
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.249303253611098e-06,
+      "loss": 1.1895,
+      "step": 16819
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.248269763321375e-06,
+      "loss": 1.5479,
+      "step": 16820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.247236281106466e-06,
+      "loss": 1.1992,
+      "step": 16821
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.246202806977476e-06,
+      "loss": 0.9668,
+      "step": 16822
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.245169340945502e-06,
+      "loss": 1.1113,
+      "step": 16823
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.244135883021647e-06,
+      "loss": 1.0088,
+      "step": 16824
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.243102433217014e-06,
+      "loss": 0.9658,
+      "step": 16825
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.242068991542694e-06,
+      "loss": 1.0869,
+      "step": 16826
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.241035558009802e-06,
+      "loss": 0.6665,
+      "step": 16827
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24000213262943e-06,
+      "loss": 0.8379,
+      "step": 16828
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.238968715412684e-06,
+      "loss": 1.0439,
+      "step": 16829
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.237935306370659e-06,
+      "loss": 0.7344,
+      "step": 16830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.236901905514456e-06,
+      "loss": 1.2041,
+      "step": 16831
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.235868512855178e-06,
+      "loss": 0.9326,
+      "step": 16832
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.234835128403926e-06,
+      "loss": 0.9575,
+      "step": 16833
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.233801752171802e-06,
+      "loss": 1.1064,
+      "step": 16834
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2327683841699e-06,
+      "loss": 1.0195,
+      "step": 16835
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.23173502440932e-06,
+      "loss": 1.1377,
+      "step": 16836
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.230701672901168e-06,
+      "loss": 1.1714,
+      "step": 16837
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.229668329656544e-06,
+      "loss": 1.1235,
+      "step": 16838
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.228634994686543e-06,
+      "loss": 1.3271,
+      "step": 16839
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.227601668002265e-06,
+      "loss": 1.2666,
+      "step": 16840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.226568349614809e-06,
+      "loss": 1.0928,
+      "step": 16841
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.225535039535282e-06,
+      "loss": 0.8408,
+      "step": 16842
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.224501737774776e-06,
+      "loss": 0.7861,
+      "step": 16843
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.223468444344394e-06,
+      "loss": 1.167,
+      "step": 16844
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.222435159255234e-06,
+      "loss": 0.7168,
+      "step": 16845
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.221401882518392e-06,
+      "loss": 1.1221,
+      "step": 16846
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22036861414497e-06,
+      "loss": 0.6406,
+      "step": 16847
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.219335354146068e-06,
+      "loss": 1.1953,
+      "step": 16848
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.218302102532786e-06,
+      "loss": 1.2637,
+      "step": 16849
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.217268859316219e-06,
+      "loss": 1.0322,
+      "step": 16850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.216235624507467e-06,
+      "loss": 1.2891,
+      "step": 16851
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.215202398117628e-06,
+      "loss": 0.8535,
+      "step": 16852
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.214169180157807e-06,
+      "loss": 0.7876,
+      "step": 16853
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.213135970639092e-06,
+      "loss": 1.4492,
+      "step": 16854
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.212102769572589e-06,
+      "loss": 1.0234,
+      "step": 16855
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.211069576969395e-06,
+      "loss": 0.7617,
+      "step": 16856
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.210036392840603e-06,
+      "loss": 1.334,
+      "step": 16857
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.209003217197316e-06,
+      "loss": 1.2275,
+      "step": 16858
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.207970050050632e-06,
+      "loss": 0.8818,
+      "step": 16859
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20693689141165e-06,
+      "loss": 1.3945,
+      "step": 16860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.205903741291462e-06,
+      "loss": 1.123,
+      "step": 16861
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20487059970117e-06,
+      "loss": 1.0703,
+      "step": 16862
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.203837466651872e-06,
+      "loss": 1.0869,
+      "step": 16863
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.202804342154667e-06,
+      "loss": 0.9502,
+      "step": 16864
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.201771226220646e-06,
+      "loss": 0.9717,
+      "step": 16865
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.200738118860914e-06,
+      "loss": 1.0742,
+      "step": 16866
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.199705020086561e-06,
+      "loss": 0.8252,
+      "step": 16867
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19867192990869e-06,
+      "loss": 0.9702,
+      "step": 16868
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.197638848338398e-06,
+      "loss": 1.0205,
+      "step": 16869
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.196605775386777e-06,
+      "loss": 1.5,
+      "step": 16870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.195572711064932e-06,
+      "loss": 1.0068,
+      "step": 16871
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.194539655383945e-06,
+      "loss": 1.3262,
+      "step": 16872
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19350660835493e-06,
+      "loss": 0.7637,
+      "step": 16873
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.192473569988975e-06,
+      "loss": 0.8936,
+      "step": 16874
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.191440540297178e-06,
+      "loss": 1.2568,
+      "step": 16875
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190407519290633e-06,
+      "loss": 1.1191,
+      "step": 16876
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18937450698044e-06,
+      "loss": 1.3359,
+      "step": 16877
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18834150337769e-06,
+      "loss": 0.813,
+      "step": 16878
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.187308508493485e-06,
+      "loss": 0.8892,
+      "step": 16879
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.186275522338922e-06,
+      "loss": 0.7529,
+      "step": 16880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18524254492509e-06,
+      "loss": 0.8643,
+      "step": 16881
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.184209576263093e-06,
+      "loss": 1.2637,
+      "step": 16882
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.183176616364014e-06,
+      "loss": 1.2285,
+      "step": 16883
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.182143665238962e-06,
+      "loss": 0.9238,
+      "step": 16884
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.181110722899029e-06,
+      "loss": 1.5449,
+      "step": 16885
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.180077789355307e-06,
+      "loss": 1.0518,
+      "step": 16886
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.179044864618896e-06,
+      "loss": 0.8672,
+      "step": 16887
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.178011948700885e-06,
+      "loss": 1.0996,
+      "step": 16888
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.176979041612375e-06,
+      "loss": 0.79,
+      "step": 16889
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17594614336446e-06,
+      "loss": 0.9082,
+      "step": 16890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.174913253968235e-06,
+      "loss": 1.4297,
+      "step": 16891
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.173880373434792e-06,
+      "loss": 1.2334,
+      "step": 16892
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.172847501775226e-06,
+      "loss": 0.7119,
+      "step": 16893
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.171814639000638e-06,
+      "loss": 1.0234,
+      "step": 16894
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.170781785122118e-06,
+      "loss": 1.3945,
+      "step": 16895
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16974894015076e-06,
+      "loss": 1.1182,
+      "step": 16896
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.168716104097659e-06,
+      "loss": 1.123,
+      "step": 16897
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.167683276973906e-06,
+      "loss": 0.8706,
+      "step": 16898
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166650458790605e-06,
+      "loss": 1.1777,
+      "step": 16899
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.165617649558843e-06,
+      "loss": 0.8223,
+      "step": 16900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.164584849289712e-06,
+      "loss": 0.9375,
+      "step": 16901
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.163552057994313e-06,
+      "loss": 0.7363,
+      "step": 16902
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.162519275683731e-06,
+      "loss": 1.2236,
+      "step": 16903
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.161486502369068e-06,
+      "loss": 1.2598,
+      "step": 16904
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.160453738061414e-06,
+      "loss": 1.0117,
+      "step": 16905
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.159420982771865e-06,
+      "loss": 1.0205,
+      "step": 16906
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15838823651151e-06,
+      "loss": 1.2949,
+      "step": 16907
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.157355499291446e-06,
+      "loss": 1.1855,
+      "step": 16908
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.156322771122763e-06,
+      "loss": 0.918,
+      "step": 16909
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15529005201656e-06,
+      "loss": 1.0771,
+      "step": 16910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.154257341983924e-06,
+      "loss": 1.1323,
+      "step": 16911
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.153224641035952e-06,
+      "loss": 0.8784,
+      "step": 16912
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.152191949183736e-06,
+      "loss": 0.584,
+      "step": 16913
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.151159266438362e-06,
+      "loss": 1.1797,
+      "step": 16914
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.150126592810935e-06,
+      "loss": 0.793,
+      "step": 16915
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.149093928312541e-06,
+      "loss": 0.9224,
+      "step": 16916
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.148061272954275e-06,
+      "loss": 0.9561,
+      "step": 16917
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.147028626747224e-06,
+      "loss": 1.0605,
+      "step": 16918
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.14599598970248e-06,
+      "loss": 1.0332,
+      "step": 16919
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.144963361831144e-06,
+      "loss": 0.9463,
+      "step": 16920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.143930743144303e-06,
+      "loss": 1.3623,
+      "step": 16921
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.142898133653049e-06,
+      "loss": 0.8477,
+      "step": 16922
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.141865533368474e-06,
+      "loss": 1.0498,
+      "step": 16923
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.140832942301667e-06,
+      "loss": 0.7754,
+      "step": 16924
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.139800360463722e-06,
+      "loss": 1.2666,
+      "step": 16925
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138767787865737e-06,
+      "loss": 0.8984,
+      "step": 16926
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.137735224518795e-06,
+      "loss": 0.9424,
+      "step": 16927
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.136702670433992e-06,
+      "loss": 1.249,
+      "step": 16928
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.13567012562241e-06,
+      "loss": 0.8975,
+      "step": 16929
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.134637590095154e-06,
+      "loss": 0.9893,
+      "step": 16930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.133605063863308e-06,
+      "loss": 0.8232,
+      "step": 16931
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132572546937963e-06,
+      "loss": 1.0015,
+      "step": 16932
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.131540039330213e-06,
+      "loss": 1.0078,
+      "step": 16933
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.130507541051142e-06,
+      "loss": 1.083,
+      "step": 16934
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.129475052111848e-06,
+      "loss": 0.5044,
+      "step": 16935
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.128442572523418e-06,
+      "loss": 1.0928,
+      "step": 16936
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.127410102296946e-06,
+      "loss": 1.0039,
+      "step": 16937
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.126377641443518e-06,
+      "loss": 1.0879,
+      "step": 16938
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.125345189974227e-06,
+      "loss": 1.0732,
+      "step": 16939
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12431274790016e-06,
+      "loss": 0.7676,
+      "step": 16940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.123280315232412e-06,
+      "loss": 0.9688,
+      "step": 16941
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12224789198207e-06,
+      "loss": 0.8838,
+      "step": 16942
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.121215478160224e-06,
+      "loss": 1.2793,
+      "step": 16943
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.120183073777965e-06,
+      "loss": 0.8906,
+      "step": 16944
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.119150678846379e-06,
+      "loss": 1.1855,
+      "step": 16945
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11811829337656e-06,
+      "loss": 0.5488,
+      "step": 16946
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.117085917379598e-06,
+      "loss": 0.9429,
+      "step": 16947
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.116053550866582e-06,
+      "loss": 1.0674,
+      "step": 16948
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.115021193848597e-06,
+      "loss": 1.4463,
+      "step": 16949
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.113988846336732e-06,
+      "loss": 1.0254,
+      "step": 16950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.112956508342082e-06,
+      "loss": 0.7642,
+      "step": 16951
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.111924179875736e-06,
+      "loss": 0.7764,
+      "step": 16952
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.110891860948777e-06,
+      "loss": 1.4492,
+      "step": 16953
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.109859551572297e-06,
+      "loss": 1.1787,
+      "step": 16954
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.108827251757384e-06,
+      "loss": 0.917,
+      "step": 16955
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10779496151513e-06,
+      "loss": 1.002,
+      "step": 16956
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.106762680856618e-06,
+      "loss": 0.812,
+      "step": 16957
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10573040979294e-06,
+      "loss": 1.1978,
+      "step": 16958
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.104698148335185e-06,
+      "loss": 0.8066,
+      "step": 16959
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.103665896494433e-06,
+      "loss": 0.9238,
+      "step": 16960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.102633654281784e-06,
+      "loss": 0.8936,
+      "step": 16961
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10160142170832e-06,
+      "loss": 1.1133,
+      "step": 16962
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10056919878513e-06,
+      "loss": 0.915,
+      "step": 16963
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0995369855233e-06,
+      "loss": 0.7754,
+      "step": 16964
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.098504781933919e-06,
+      "loss": 0.8262,
+      "step": 16965
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.097472588028072e-06,
+      "loss": 0.8936,
+      "step": 16966
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.096440403816851e-06,
+      "loss": 1.0332,
+      "step": 16967
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.095408229311343e-06,
+      "loss": 1.084,
+      "step": 16968
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.094376064522631e-06,
+      "loss": 0.9785,
+      "step": 16969
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.093343909461807e-06,
+      "loss": 0.7305,
+      "step": 16970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09231176413995e-06,
+      "loss": 1.2598,
+      "step": 16971
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.091279628568157e-06,
+      "loss": 1.2354,
+      "step": 16972
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09024750275751e-06,
+      "loss": 1.1084,
+      "step": 16973
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.089215386719098e-06,
+      "loss": 1.0088,
+      "step": 16974
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.088183280464004e-06,
+      "loss": 1.0986,
+      "step": 16975
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.087151184003313e-06,
+      "loss": 1.127,
+      "step": 16976
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086119097348116e-06,
+      "loss": 1.1719,
+      "step": 16977
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0850870205095e-06,
+      "loss": 0.8008,
+      "step": 16978
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08405495349855e-06,
+      "loss": 1.2646,
+      "step": 16979
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08302289632635e-06,
+      "loss": 0.5166,
+      "step": 16980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.081990849003983e-06,
+      "loss": 1.5068,
+      "step": 16981
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.080958811542543e-06,
+      "loss": 1.4473,
+      "step": 16982
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.079926783953113e-06,
+      "loss": 0.8506,
+      "step": 16983
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.078894766246776e-06,
+      "loss": 1.2305,
+      "step": 16984
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.077862758434618e-06,
+      "loss": 0.8408,
+      "step": 16985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.076830760527727e-06,
+      "loss": 1.1973,
+      "step": 16986
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.075798772537187e-06,
+      "loss": 1.1504,
+      "step": 16987
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.074766794474084e-06,
+      "loss": 1.209,
+      "step": 16988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0737348263495e-06,
+      "loss": 1.3223,
+      "step": 16989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.072702868174526e-06,
+      "loss": 1.2441,
+      "step": 16990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.071670919960237e-06,
+      "loss": 0.7295,
+      "step": 16991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.070638981717728e-06,
+      "loss": 0.9404,
+      "step": 16992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.069607053458078e-06,
+      "loss": 1.0391,
+      "step": 16993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.068575135192377e-06,
+      "loss": 0.5415,
+      "step": 16994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.067543226931704e-06,
+      "loss": 0.9082,
+      "step": 16995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.066511328687144e-06,
+      "loss": 0.9941,
+      "step": 16996
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065479440469781e-06,
+      "loss": 0.748,
+      "step": 16997
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.064447562290705e-06,
+      "loss": 1.2539,
+      "step": 16998
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.063415694160993e-06,
+      "loss": 0.8081,
+      "step": 16999
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.062383836091733e-06,
+      "loss": 0.9629,
+      "step": 17000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.061351988094008e-06,
+      "loss": 1.1836,
+      "step": 17001
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.060320150178898e-06,
+      "loss": 1.2158,
+      "step": 17002
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05928832235749e-06,
+      "loss": 1.1387,
+      "step": 17003
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.058256504640872e-06,
+      "loss": 1.2236,
+      "step": 17004
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.057224697040122e-06,
+      "loss": 1.1172,
+      "step": 17005
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05619289956632e-06,
+      "loss": 0.9375,
+      "step": 17006
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.055161112230555e-06,
+      "loss": 0.9238,
+      "step": 17007
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.054129335043909e-06,
+      "loss": 1.0342,
+      "step": 17008
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.053097568017465e-06,
+      "loss": 0.7573,
+      "step": 17009
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.052065811162305e-06,
+      "loss": 1.0479,
+      "step": 17010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.051034064489512e-06,
+      "loss": 0.8467,
+      "step": 17011
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.050002328010165e-06,
+      "loss": 0.9463,
+      "step": 17012
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.048970601735356e-06,
+      "loss": 1.2988,
+      "step": 17013
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04793888567616e-06,
+      "loss": 0.7998,
+      "step": 17014
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04690717984366e-06,
+      "loss": 1.4199,
+      "step": 17015
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.045875484248942e-06,
+      "loss": 0.8418,
+      "step": 17016
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.044843798903078e-06,
+      "loss": 1.0029,
+      "step": 17017
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.043812123817165e-06,
+      "loss": 1.3027,
+      "step": 17018
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.042780459002273e-06,
+      "loss": 1.0781,
+      "step": 17019
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.041748804469492e-06,
+      "loss": 1.0723,
+      "step": 17020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.040717160229897e-06,
+      "loss": 1.3164,
+      "step": 17021
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.039685526294569e-06,
+      "loss": 1.0581,
+      "step": 17022
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.038653902674596e-06,
+      "loss": 1.2539,
+      "step": 17023
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037622289381054e-06,
+      "loss": 1.2402,
+      "step": 17024
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.036590686425031e-06,
+      "loss": 1.0537,
+      "step": 17025
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.035559093817598e-06,
+      "loss": 0.9072,
+      "step": 17026
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.034527511569847e-06,
+      "loss": 0.8613,
+      "step": 17027
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.033495939692843e-06,
+      "loss": 1.0938,
+      "step": 17028
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.032464378197687e-06,
+      "loss": 1.2441,
+      "step": 17029
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.031432827095444e-06,
+      "loss": 1.041,
+      "step": 17030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.030401286397203e-06,
+      "loss": 1.127,
+      "step": 17031
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.029369756114042e-06,
+      "loss": 0.8857,
+      "step": 17032
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028338236257035e-06,
+      "loss": 1.0176,
+      "step": 17033
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.027306726837273e-06,
+      "loss": 1.0068,
+      "step": 17034
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02627522786583e-06,
+      "loss": 1.1865,
+      "step": 17035
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.025243739353791e-06,
+      "loss": 0.8877,
+      "step": 17036
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.024212261312228e-06,
+      "loss": 0.9165,
+      "step": 17037
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.023180793752222e-06,
+      "loss": 0.8506,
+      "step": 17038
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02214933668486e-06,
+      "loss": 0.9272,
+      "step": 17039
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.021117890121217e-06,
+      "loss": 1.1504,
+      "step": 17040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.020086454072373e-06,
+      "loss": 0.8936,
+      "step": 17041
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.019055028549403e-06,
+      "loss": 1.2666,
+      "step": 17042
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.01802361356339e-06,
+      "loss": 1.1045,
+      "step": 17043
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.016992209125418e-06,
+      "loss": 1.0615,
+      "step": 17044
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.015960815246557e-06,
+      "loss": 1.5869,
+      "step": 17045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.01492943193789e-06,
+      "loss": 0.7505,
+      "step": 17046
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.0138980592105e-06,
+      "loss": 1.2266,
+      "step": 17047
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.012866697075456e-06,
+      "loss": 0.7412,
+      "step": 17048
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.011835345543843e-06,
+      "loss": 0.8477,
+      "step": 17049
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.010804004626739e-06,
+      "loss": 0.9756,
+      "step": 17050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.009772674335223e-06,
+      "loss": 1.0762,
+      "step": 17051
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00874135468037e-06,
+      "loss": 1.5117,
+      "step": 17052
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.007710045673257e-06,
+      "loss": 0.9346,
+      "step": 17053
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.006678747324966e-06,
+      "loss": 1.167,
+      "step": 17054
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.005647459646578e-06,
+      "loss": 0.9395,
+      "step": 17055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.004616182649162e-06,
+      "loss": 0.8274,
+      "step": 17056
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.003584916343801e-06,
+      "loss": 1.1191,
+      "step": 17057
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.002553660741575e-06,
+      "loss": 0.957,
+      "step": 17058
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.001522415853548e-06,
+      "loss": 0.8867,
+      "step": 17059
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.000491181690815e-06,
+      "loss": 1.5322,
+      "step": 17060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.999459958264443e-06,
+      "loss": 0.959,
+      "step": 17061
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.998428745585514e-06,
+      "loss": 1.1074,
+      "step": 17062
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.997397543665098e-06,
+      "loss": 1.459,
+      "step": 17063
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.996366352514273e-06,
+      "loss": 0.9795,
+      "step": 17064
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.995335172144125e-06,
+      "loss": 0.8984,
+      "step": 17065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.994304002565723e-06,
+      "loss": 1.1133,
+      "step": 17066
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.993272843790142e-06,
+      "loss": 1.2114,
+      "step": 17067
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.992241695828461e-06,
+      "loss": 1.1562,
+      "step": 17068
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.991210558691757e-06,
+      "loss": 1.1211,
+      "step": 17069
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.990179432391104e-06,
+      "loss": 1.0488,
+      "step": 17070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.989148316937582e-06,
+      "loss": 0.9238,
+      "step": 17071
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.988117212342262e-06,
+      "loss": 0.8379,
+      "step": 17072
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.987086118616225e-06,
+      "loss": 0.9351,
+      "step": 17073
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.986055035770536e-06,
+      "loss": 1.041,
+      "step": 17074
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.985023963816286e-06,
+      "loss": 0.9609,
+      "step": 17075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.983992902764538e-06,
+      "loss": 0.8096,
+      "step": 17076
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.982961852626374e-06,
+      "loss": 0.8828,
+      "step": 17077
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.981930813412866e-06,
+      "loss": 1.4277,
+      "step": 17078
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.980899785135086e-06,
+      "loss": 0.8525,
+      "step": 17079
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.979868767804117e-06,
+      "loss": 0.8247,
+      "step": 17080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97883776143103e-06,
+      "loss": 1.0088,
+      "step": 17081
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.9778067660269e-06,
+      "loss": 0.7656,
+      "step": 17082
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.976775781602798e-06,
+      "loss": 0.7444,
+      "step": 17083
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.9757448081698e-06,
+      "loss": 0.937,
+      "step": 17084
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.974713845738984e-06,
+      "loss": 0.8633,
+      "step": 17085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.973682894321425e-06,
+      "loss": 1.1455,
+      "step": 17086
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97265195392819e-06,
+      "loss": 1.2695,
+      "step": 17087
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.971621024570359e-06,
+      "loss": 1.2383,
+      "step": 17088
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.970590106259005e-06,
+      "loss": 0.8711,
+      "step": 17089
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.969559199005196e-06,
+      "loss": 1.1035,
+      "step": 17090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.968528302820014e-06,
+      "loss": 1.0674,
+      "step": 17091
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.967497417714527e-06,
+      "loss": 0.7622,
+      "step": 17092
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.966466543699812e-06,
+      "loss": 0.8511,
+      "step": 17093
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.96543568078694e-06,
+      "loss": 0.9746,
+      "step": 17094
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.96440482898698e-06,
+      "loss": 0.9932,
+      "step": 17095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.963373988311015e-06,
+      "loss": 0.8198,
+      "step": 17096
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.962343158770114e-06,
+      "loss": 0.9365,
+      "step": 17097
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.961312340375346e-06,
+      "loss": 0.4739,
+      "step": 17098
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.960281533137786e-06,
+      "loss": 0.9199,
+      "step": 17099
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.959250737068503e-06,
+      "loss": 0.7886,
+      "step": 17100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95821995217858e-06,
+      "loss": 0.9434,
+      "step": 17101
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.957189178479077e-06,
+      "loss": 0.9922,
+      "step": 17102
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.956158415981075e-06,
+      "loss": 1.0449,
+      "step": 17103
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.955127664695643e-06,
+      "loss": 1.123,
+      "step": 17104
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.954096924633846e-06,
+      "loss": 0.873,
+      "step": 17105
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95306619580677e-06,
+      "loss": 0.8037,
+      "step": 17106
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.952035478225475e-06,
+      "loss": 1.2373,
+      "step": 17107
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95100477190104e-06,
+      "loss": 0.9727,
+      "step": 17108
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.949974076844533e-06,
+      "loss": 1.1543,
+      "step": 17109
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94894339306702e-06,
+      "loss": 0.8213,
+      "step": 17110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.947912720579584e-06,
+      "loss": 0.9268,
+      "step": 17111
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.946882059393288e-06,
+      "loss": 1.0996,
+      "step": 17112
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.945851409519206e-06,
+      "loss": 1.0518,
+      "step": 17113
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.944820770968407e-06,
+      "loss": 0.8047,
+      "step": 17114
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.943790143751959e-06,
+      "loss": 1.2227,
+      "step": 17115
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94275952788094e-06,
+      "loss": 0.8447,
+      "step": 17116
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.941728923366419e-06,
+      "loss": 1.0264,
+      "step": 17117
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.940698330219459e-06,
+      "loss": 1.1299,
+      "step": 17118
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.939667748451141e-06,
+      "loss": 1.4775,
+      "step": 17119
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.938637178072525e-06,
+      "loss": 0.9307,
+      "step": 17120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.937606619094684e-06,
+      "loss": 0.8242,
+      "step": 17121
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.936576071528693e-06,
+      "loss": 1.0879,
+      "step": 17122
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.935545535385617e-06,
+      "loss": 1.2754,
+      "step": 17123
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.934515010676528e-06,
+      "loss": 0.6602,
+      "step": 17124
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.933484497412493e-06,
+      "loss": 1.2734,
+      "step": 17125
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.93245399560458e-06,
+      "loss": 0.9932,
+      "step": 17126
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.931423505263864e-06,
+      "loss": 1.144,
+      "step": 17127
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.930393026401413e-06,
+      "loss": 1.4414,
+      "step": 17128
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.929362559028293e-06,
+      "loss": 0.9829,
+      "step": 17129
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.928332103155572e-06,
+      "loss": 1.3203,
+      "step": 17130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.927301658794321e-06,
+      "loss": 0.8643,
+      "step": 17131
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.926271225955612e-06,
+      "loss": 1.2119,
+      "step": 17132
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.925240804650507e-06,
+      "loss": 0.9551,
+      "step": 17133
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.924210394890078e-06,
+      "loss": 1.3301,
+      "step": 17134
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.923179996685396e-06,
+      "loss": 1.1094,
+      "step": 17135
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.922149610047521e-06,
+      "loss": 1.0186,
+      "step": 17136
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.921119234987526e-06,
+      "loss": 1.3096,
+      "step": 17137
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.920088871516482e-06,
+      "loss": 0.9067,
+      "step": 17138
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.919058519645454e-06,
+      "loss": 1.1406,
+      "step": 17139
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.918028179385508e-06,
+      "loss": 1.0957,
+      "step": 17140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.91699785074771e-06,
+      "loss": 0.8135,
+      "step": 17141
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.915967533743135e-06,
+      "loss": 0.9199,
+      "step": 17142
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914937228382845e-06,
+      "loss": 1.1572,
+      "step": 17143
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.913906934677905e-06,
+      "loss": 1.0312,
+      "step": 17144
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.912876652639387e-06,
+      "loss": 1.1914,
+      "step": 17145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.911846382278354e-06,
+      "loss": 1.1094,
+      "step": 17146
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.910816123605878e-06,
+      "loss": 1.3438,
+      "step": 17147
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90978587663302e-06,
+      "loss": 1.0732,
+      "step": 17148
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.908755641370848e-06,
+      "loss": 0.7451,
+      "step": 17149
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.907725417830432e-06,
+      "loss": 1.4521,
+      "step": 17150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.906695206022833e-06,
+      "loss": 0.5454,
+      "step": 17151
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.905665005959118e-06,
+      "loss": 1.1006,
+      "step": 17152
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.904634817650356e-06,
+      "loss": 1.4766,
+      "step": 17153
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.903604641107615e-06,
+      "loss": 1.1846,
+      "step": 17154
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.902574476341953e-06,
+      "loss": 0.9521,
+      "step": 17155
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.901544323364443e-06,
+      "loss": 1.2842,
+      "step": 17156
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.900514182186144e-06,
+      "loss": 1.0996,
+      "step": 17157
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.899484052818127e-06,
+      "loss": 1.0176,
+      "step": 17158
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.898453935271457e-06,
+      "loss": 0.8096,
+      "step": 17159
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.897423829557196e-06,
+      "loss": 1.1689,
+      "step": 17160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.896393735686412e-06,
+      "loss": 0.731,
+      "step": 17161
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.895363653670162e-06,
+      "loss": 1.5703,
+      "step": 17162
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.894333583519523e-06,
+      "loss": 0.8555,
+      "step": 17163
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.893303525245552e-06,
+      "loss": 0.9297,
+      "step": 17164
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.892273478859318e-06,
+      "loss": 0.9326,
+      "step": 17165
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.89124344437188e-06,
+      "loss": 1.3818,
+      "step": 17166
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890213421794302e-06,
+      "loss": 0.7275,
+      "step": 17167
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.889183411137654e-06,
+      "loss": 0.8926,
+      "step": 17168
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.888153412412997e-06,
+      "loss": 1.3877,
+      "step": 17169
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.887123425631396e-06,
+      "loss": 0.9951,
+      "step": 17170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.886093450803913e-06,
+      "loss": 0.8259,
+      "step": 17171
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.88506348794161e-06,
+      "loss": 1.0928,
+      "step": 17172
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.884033537055553e-06,
+      "loss": 1.373,
+      "step": 17173
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.88300359815681e-06,
+      "loss": 1.0771,
+      "step": 17174
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.881973671256435e-06,
+      "loss": 1.1777,
+      "step": 17175
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.880943756365496e-06,
+      "loss": 1.084,
+      "step": 17176
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.879913853495053e-06,
+      "loss": 0.8721,
+      "step": 17177
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.878883962656176e-06,
+      "loss": 0.8843,
+      "step": 17178
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.877854083859922e-06,
+      "loss": 0.7798,
+      "step": 17179
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.876824217117352e-06,
+      "loss": 0.8369,
+      "step": 17180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.875794362439534e-06,
+      "loss": 0.8066,
+      "step": 17181
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.874764519837527e-06,
+      "loss": 0.9766,
+      "step": 17182
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.87373468932239e-06,
+      "loss": 1.2676,
+      "step": 17183
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.872704870905192e-06,
+      "loss": 1.0215,
+      "step": 17184
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.87167506459699e-06,
+      "loss": 0.9712,
+      "step": 17185
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.870645270408849e-06,
+      "loss": 1.3477,
+      "step": 17186
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869615488351828e-06,
+      "loss": 1.0459,
+      "step": 17187
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.868585718436985e-06,
+      "loss": 1.126,
+      "step": 17188
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.867555960675394e-06,
+      "loss": 1.4102,
+      "step": 17189
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.866526215078105e-06,
+      "loss": 0.6831,
+      "step": 17190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.865496481656182e-06,
+      "loss": 1.0195,
+      "step": 17191
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.864466760420688e-06,
+      "loss": 0.9844,
+      "step": 17192
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.863437051382679e-06,
+      "loss": 0.98,
+      "step": 17193
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86240735455322e-06,
+      "loss": 1.4697,
+      "step": 17194
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.861377669943373e-06,
+      "loss": 0.8735,
+      "step": 17195
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.860347997564196e-06,
+      "loss": 0.6011,
+      "step": 17196
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85931833742675e-06,
+      "loss": 0.8174,
+      "step": 17197
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.858288689542089e-06,
+      "loss": 0.917,
+      "step": 17198
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.857259053921284e-06,
+      "loss": 0.9814,
+      "step": 17199
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.856229430575391e-06,
+      "loss": 1.1025,
+      "step": 17200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.855199819515468e-06,
+      "loss": 1.1152,
+      "step": 17201
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.854170220752573e-06,
+      "loss": 1.3506,
+      "step": 17202
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853140634297767e-06,
+      "loss": 1.0176,
+      "step": 17203
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.852111060162114e-06,
+      "loss": 0.9395,
+      "step": 17204
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.851081498356671e-06,
+      "loss": 1.0029,
+      "step": 17205
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.850051948892494e-06,
+      "loss": 0.6797,
+      "step": 17206
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.849022411780647e-06,
+      "loss": 0.938,
+      "step": 17207
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.847992887032178e-06,
+      "loss": 0.6328,
+      "step": 17208
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84696337465816e-06,
+      "loss": 0.8916,
+      "step": 17209
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.845933874669645e-06,
+      "loss": 1.0703,
+      "step": 17210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84490438707769e-06,
+      "loss": 0.9092,
+      "step": 17211
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84387491189336e-06,
+      "loss": 0.9626,
+      "step": 17212
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.842845449127703e-06,
+      "loss": 1.3301,
+      "step": 17213
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.841815998791785e-06,
+      "loss": 1.0879,
+      "step": 17214
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.840786560896661e-06,
+      "loss": 1.2402,
+      "step": 17215
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.839757135453391e-06,
+      "loss": 1.0996,
+      "step": 17216
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83872772247303e-06,
+      "loss": 1.3159,
+      "step": 17217
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.837698321966638e-06,
+      "loss": 0.9854,
+      "step": 17218
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.836668933945265e-06,
+      "loss": 0.9692,
+      "step": 17219
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.835639558419983e-06,
+      "loss": 1.0713,
+      "step": 17220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.834610195401837e-06,
+      "loss": 0.8389,
+      "step": 17221
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.833580844901888e-06,
+      "loss": 0.8328,
+      "step": 17222
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.832551506931196e-06,
+      "loss": 1.0176,
+      "step": 17223
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.831522181500809e-06,
+      "loss": 1.2402,
+      "step": 17224
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83049286862179e-06,
+      "loss": 1.0059,
+      "step": 17225
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.829463568305197e-06,
+      "loss": 1.5918,
+      "step": 17226
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.828434280562084e-06,
+      "loss": 1.208,
+      "step": 17227
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.827405005403508e-06,
+      "loss": 0.8647,
+      "step": 17228
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.826375742840521e-06,
+      "loss": 0.854,
+      "step": 17229
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.825346492884183e-06,
+      "loss": 1.041,
+      "step": 17230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.824317255545552e-06,
+      "loss": 1.3984,
+      "step": 17231
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82328803083568e-06,
+      "loss": 1.1318,
+      "step": 17232
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.822258818765622e-06,
+      "loss": 1.1562,
+      "step": 17233
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.821229619346433e-06,
+      "loss": 1.1016,
+      "step": 17234
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.820200432589175e-06,
+      "loss": 1.1377,
+      "step": 17235
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819171258504895e-06,
+      "loss": 0.9238,
+      "step": 17236
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.818142097104651e-06,
+      "loss": 0.8369,
+      "step": 17237
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.817112948399503e-06,
+      "loss": 0.5811,
+      "step": 17238
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.816083812400493e-06,
+      "loss": 0.938,
+      "step": 17239
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.815054689118688e-06,
+      "loss": 1.0684,
+      "step": 17240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.814025578565139e-06,
+      "loss": 1.0439,
+      "step": 17241
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.812996480750899e-06,
+      "loss": 0.9229,
+      "step": 17242
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.811967395687021e-06,
+      "loss": 0.9131,
+      "step": 17243
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.810938323384561e-06,
+      "loss": 0.999,
+      "step": 17244
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.809909263854571e-06,
+      "loss": 1.1475,
+      "step": 17245
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80888021710811e-06,
+      "loss": 0.9922,
+      "step": 17246
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.807851183156225e-06,
+      "loss": 0.8447,
+      "step": 17247
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.806822162009974e-06,
+      "loss": 0.6592,
+      "step": 17248
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80579315368041e-06,
+      "loss": 1.2188,
+      "step": 17249
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.804764158178579e-06,
+      "loss": 0.9746,
+      "step": 17250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.803735175515546e-06,
+      "loss": 0.8936,
+      "step": 17251
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.802706205702355e-06,
+      "loss": 0.8687,
+      "step": 17252
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.801677248750068e-06,
+      "loss": 0.8936,
+      "step": 17253
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.800648304669726e-06,
+      "loss": 0.9902,
+      "step": 17254
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.799619373472386e-06,
+      "loss": 1.1064,
+      "step": 17255
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.798590455169102e-06,
+      "loss": 1.3018,
+      "step": 17256
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.797561549770927e-06,
+      "loss": 1.1289,
+      "step": 17257
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.796532657288916e-06,
+      "loss": 0.9326,
+      "step": 17258
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.795503777734113e-06,
+      "loss": 1.0107,
+      "step": 17259
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.794474911117572e-06,
+      "loss": 0.9785,
+      "step": 17260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.793446057450349e-06,
+      "loss": 1.0911,
+      "step": 17261
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.792417216743496e-06,
+      "loss": 1.0078,
+      "step": 17262
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.791388389008057e-06,
+      "loss": 0.8149,
+      "step": 17263
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.790359574255091e-06,
+      "loss": 0.8164,
+      "step": 17264
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.789330772495641e-06,
+      "loss": 0.9404,
+      "step": 17265
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.78830198374077e-06,
+      "loss": 0.9746,
+      "step": 17266
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.787273208001516e-06,
+      "loss": 0.5044,
+      "step": 17267
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.786244445288938e-06,
+      "loss": 1.043,
+      "step": 17268
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785215695614088e-06,
+      "loss": 1.3691,
+      "step": 17269
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.784186958988006e-06,
+      "loss": 0.8516,
+      "step": 17270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.783158235421752e-06,
+      "loss": 0.9272,
+      "step": 17271
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.782129524926373e-06,
+      "loss": 0.8936,
+      "step": 17272
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.78110082751292e-06,
+      "loss": 0.8955,
+      "step": 17273
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.780072143192444e-06,
+      "loss": 0.9697,
+      "step": 17274
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.779043471975989e-06,
+      "loss": 1.1992,
+      "step": 17275
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.778014813874607e-06,
+      "loss": 0.8857,
+      "step": 17276
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.776986168899355e-06,
+      "loss": 0.8047,
+      "step": 17277
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.775957537061273e-06,
+      "loss": 0.8389,
+      "step": 17278
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.774928918371412e-06,
+      "loss": 0.918,
+      "step": 17279
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.773900312840827e-06,
+      "loss": 0.6523,
+      "step": 17280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.772871720480558e-06,
+      "loss": 1.0361,
+      "step": 17281
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.771843141301659e-06,
+      "loss": 1.3467,
+      "step": 17282
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.770814575315178e-06,
+      "loss": 1.1035,
+      "step": 17283
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.769786022532166e-06,
+      "loss": 1.0527,
+      "step": 17284
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.768757482963667e-06,
+      "loss": 1.1543,
+      "step": 17285
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.767728956620727e-06,
+      "loss": 0.7004,
+      "step": 17286
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.766700443514403e-06,
+      "loss": 0.9141,
+      "step": 17287
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.765671943655738e-06,
+      "loss": 1.1602,
+      "step": 17288
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.764643457055778e-06,
+      "loss": 0.9629,
+      "step": 17289
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763614983725573e-06,
+      "loss": 1.2129,
+      "step": 17290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.762586523676166e-06,
+      "loss": 1.2256,
+      "step": 17291
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.761558076918612e-06,
+      "loss": 1.0684,
+      "step": 17292
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.760529643463955e-06,
+      "loss": 1.2065,
+      "step": 17293
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75950122332324e-06,
+      "loss": 1.2598,
+      "step": 17294
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.758472816507516e-06,
+      "loss": 0.9482,
+      "step": 17295
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.757444423027823e-06,
+      "loss": 1.1514,
+      "step": 17296
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.756416042895222e-06,
+      "loss": 1.0225,
+      "step": 17297
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.755387676120747e-06,
+      "loss": 0.6611,
+      "step": 17298
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.754359322715451e-06,
+      "loss": 1.4365,
+      "step": 17299
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.753330982690376e-06,
+      "loss": 1.3271,
+      "step": 17300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75230265605657e-06,
+      "loss": 1.002,
+      "step": 17301
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.751274342825077e-06,
+      "loss": 1.1582,
+      "step": 17302
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.750246043006946e-06,
+      "loss": 0.7842,
+      "step": 17303
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.749217756613223e-06,
+      "loss": 1.0254,
+      "step": 17304
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74818948365495e-06,
+      "loss": 0.9609,
+      "step": 17305
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.747161224143175e-06,
+      "loss": 1.0088,
+      "step": 17306
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.746132978088938e-06,
+      "loss": 1.0303,
+      "step": 17307
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.745104745503292e-06,
+      "loss": 0.9829,
+      "step": 17308
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.744076526397276e-06,
+      "loss": 1.0146,
+      "step": 17309
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.743048320781937e-06,
+      "loss": 0.9199,
+      "step": 17310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.742020128668323e-06,
+      "loss": 1.2148,
+      "step": 17311
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74099195006747e-06,
+      "loss": 1.1221,
+      "step": 17312
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.739963784990428e-06,
+      "loss": 1.2578,
+      "step": 17313
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.738935633448242e-06,
+      "loss": 0.6348,
+      "step": 17314
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.737907495451957e-06,
+      "loss": 0.9453,
+      "step": 17315
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.736879371012611e-06,
+      "loss": 1.0669,
+      "step": 17316
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.735851260141248e-06,
+      "loss": 0.834,
+      "step": 17317
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.734823162848919e-06,
+      "loss": 1.2324,
+      "step": 17318
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.733795079146664e-06,
+      "loss": 1.0723,
+      "step": 17319
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.732767009045523e-06,
+      "loss": 1.2725,
+      "step": 17320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.731738952556543e-06,
+      "loss": 0.7495,
+      "step": 17321
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.730710909690761e-06,
+      "loss": 0.8501,
+      "step": 17322
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.729682880459232e-06,
+      "loss": 1.4785,
+      "step": 17323
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.728654864872987e-06,
+      "loss": 1.1504,
+      "step": 17324
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.727626862943072e-06,
+      "loss": 1.3457,
+      "step": 17325
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.726598874680533e-06,
+      "loss": 1.4141,
+      "step": 17326
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.725570900096405e-06,
+      "loss": 1.0947,
+      "step": 17327
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.724542939201737e-06,
+      "loss": 1.2383,
+      "step": 17328
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72351499200757e-06,
+      "loss": 1.1279,
+      "step": 17329
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.722487058524945e-06,
+      "loss": 1.1157,
+      "step": 17330
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.7214591387649e-06,
+      "loss": 0.9795,
+      "step": 17331
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72043123273848e-06,
+      "loss": 1.3477,
+      "step": 17332
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.719403340456726e-06,
+      "loss": 1.1885,
+      "step": 17333
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.718375461930681e-06,
+      "loss": 1.0859,
+      "step": 17334
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.717347597171385e-06,
+      "loss": 1.2148,
+      "step": 17335
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.716319746189877e-06,
+      "loss": 0.8228,
+      "step": 17336
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.7152919089972e-06,
+      "loss": 0.8257,
+      "step": 17337
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.714264085604387e-06,
+      "loss": 1.3174,
+      "step": 17338
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.713236276022494e-06,
+      "loss": 1.2393,
+      "step": 17339
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.71220848026255e-06,
+      "loss": 0.8936,
+      "step": 17340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.7111806983356e-06,
+      "loss": 1.0625,
+      "step": 17341
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.710152930252678e-06,
+      "loss": 1.1426,
+      "step": 17342
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709125176024828e-06,
+      "loss": 0.917,
+      "step": 17343
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.70809743566309e-06,
+      "loss": 1.3223,
+      "step": 17344
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.707069709178506e-06,
+      "loss": 0.8066,
+      "step": 17345
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.706041996582111e-06,
+      "loss": 1.0459,
+      "step": 17346
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.705014297884945e-06,
+      "loss": 1.2373,
+      "step": 17347
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.703986613098046e-06,
+      "loss": 0.958,
+      "step": 17348
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.702958942232459e-06,
+      "loss": 1.0435,
+      "step": 17349
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.70193128529922e-06,
+      "loss": 1.1865,
+      "step": 17350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.700903642309365e-06,
+      "loss": 0.8633,
+      "step": 17351
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.699876013273935e-06,
+      "loss": 1.2432,
+      "step": 17352
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.698848398203965e-06,
+      "loss": 1.3086,
+      "step": 17353
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.697820797110499e-06,
+      "loss": 0.9458,
+      "step": 17354
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.696793210004572e-06,
+      "loss": 0.8247,
+      "step": 17355
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69576563689722e-06,
+      "loss": 0.8203,
+      "step": 17356
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.694738077799487e-06,
+      "loss": 0.9395,
+      "step": 17357
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.693710532722402e-06,
+      "loss": 1.0645,
+      "step": 17358
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69268300167701e-06,
+      "loss": 1.0386,
+      "step": 17359
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.691655484674343e-06,
+      "loss": 1.2002,
+      "step": 17360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.690627981725445e-06,
+      "loss": 0.8994,
+      "step": 17361
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.689600492841345e-06,
+      "loss": 1.1602,
+      "step": 17362
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.688573018033087e-06,
+      "loss": 1.3193,
+      "step": 17363
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.687545557311698e-06,
+      "loss": 1.2207,
+      "step": 17364
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.686518110688227e-06,
+      "loss": 1.1719,
+      "step": 17365
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.685490678173703e-06,
+      "loss": 0.8433,
+      "step": 17366
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.684463259779164e-06,
+      "loss": 0.918,
+      "step": 17367
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.683435855515646e-06,
+      "loss": 0.7734,
+      "step": 17368
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.682408465394184e-06,
+      "loss": 1.1094,
+      "step": 17369
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.681381089425816e-06,
+      "loss": 0.8984,
+      "step": 17370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.680353727621575e-06,
+      "loss": 1.0098,
+      "step": 17371
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.679326379992502e-06,
+      "loss": 0.9912,
+      "step": 17372
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.678299046549626e-06,
+      "loss": 0.8613,
+      "step": 17373
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.677271727303983e-06,
+      "loss": 1.0757,
+      "step": 17374
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.676244422266611e-06,
+      "loss": 1.0322,
+      "step": 17375
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.675217131448545e-06,
+      "loss": 1.3076,
+      "step": 17376
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.674189854860818e-06,
+      "loss": 0.6831,
+      "step": 17377
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.673162592514466e-06,
+      "loss": 1.084,
+      "step": 17378
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.672135344420519e-06,
+      "loss": 1.272,
+      "step": 17379
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.67110811059002e-06,
+      "loss": 1.0039,
+      "step": 17380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.670080891033997e-06,
+      "loss": 0.9219,
+      "step": 17381
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.669053685763483e-06,
+      "loss": 1.2285,
+      "step": 17382
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668026494789517e-06,
+      "loss": 1.3994,
+      "step": 17383
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.666999318123124e-06,
+      "loss": 0.8696,
+      "step": 17384
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.665972155775349e-06,
+      "loss": 1.0908,
+      "step": 17385
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.664945007757218e-06,
+      "loss": 1.3496,
+      "step": 17386
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.663917874079768e-06,
+      "loss": 0.9946,
+      "step": 17387
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.66289075475403e-06,
+      "loss": 1.1582,
+      "step": 17388
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.661863649791032e-06,
+      "loss": 0.9502,
+      "step": 17389
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660836559201814e-06,
+      "loss": 0.8887,
+      "step": 17390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.659809482997412e-06,
+      "loss": 1.0996,
+      "step": 17391
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.658782421188847e-06,
+      "loss": 1.3418,
+      "step": 17392
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.65775537378716e-06,
+      "loss": 0.8369,
+      "step": 17393
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.656728340803382e-06,
+      "loss": 0.8457,
+      "step": 17394
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.655701322248537e-06,
+      "loss": 1.1777,
+      "step": 17395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.654674318133668e-06,
+      "loss": 0.8643,
+      "step": 17396
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.653647328469802e-06,
+      "loss": 0.8716,
+      "step": 17397
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.652620353267973e-06,
+      "loss": 0.8213,
+      "step": 17398
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.651593392539206e-06,
+      "loss": 1.3301,
+      "step": 17399
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.650566446294535e-06,
+      "loss": 0.5981,
+      "step": 17400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649539514544996e-06,
+      "loss": 1.0908,
+      "step": 17401
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.648512597301613e-06,
+      "loss": 0.9062,
+      "step": 17402
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.647485694575425e-06,
+      "loss": 1.2139,
+      "step": 17403
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.646458806377454e-06,
+      "loss": 1.1074,
+      "step": 17404
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.645431932718733e-06,
+      "loss": 1.1221,
+      "step": 17405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.644405073610296e-06,
+      "loss": 1.0088,
+      "step": 17406
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.643378229063172e-06,
+      "loss": 1.0547,
+      "step": 17407
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.642351399088387e-06,
+      "loss": 1.2754,
+      "step": 17408
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.641324583696976e-06,
+      "loss": 1.0605,
+      "step": 17409
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.64029778289996e-06,
+      "loss": 0.8357,
+      "step": 17410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63927099670838e-06,
+      "loss": 0.8306,
+      "step": 17411
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.638244225133258e-06,
+      "loss": 0.9917,
+      "step": 17412
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.637217468185628e-06,
+      "loss": 1.1299,
+      "step": 17413
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.636190725876514e-06,
+      "loss": 1.0693,
+      "step": 17414
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.635163998216946e-06,
+      "loss": 1.0801,
+      "step": 17415
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.634137285217956e-06,
+      "loss": 1.3525,
+      "step": 17416
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.633110586890568e-06,
+      "loss": 1.1406,
+      "step": 17417
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.632083903245816e-06,
+      "loss": 1.0186,
+      "step": 17418
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.631057234294724e-06,
+      "loss": 0.6587,
+      "step": 17419
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.630030580048319e-06,
+      "loss": 1.0869,
+      "step": 17420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.629003940517632e-06,
+      "loss": 0.9824,
+      "step": 17421
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.627977315713694e-06,
+      "loss": 0.9258,
+      "step": 17422
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.626950705647526e-06,
+      "loss": 1.083,
+      "step": 17423
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.625924110330157e-06,
+      "loss": 1.0713,
+      "step": 17424
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62489752977262e-06,
+      "loss": 1.0723,
+      "step": 17425
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62387096398593e-06,
+      "loss": 1.2402,
+      "step": 17426
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.622844412981125e-06,
+      "loss": 0.8623,
+      "step": 17427
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.621817876769229e-06,
+      "loss": 0.8115,
+      "step": 17428
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.620791355361271e-06,
+      "loss": 0.999,
+      "step": 17429
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.619764848768272e-06,
+      "loss": 0.7168,
+      "step": 17430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.618738357001256e-06,
+      "loss": 1.2734,
+      "step": 17431
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61771188007126e-06,
+      "loss": 1.0371,
+      "step": 17432
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.616685417989305e-06,
+      "loss": 0.8506,
+      "step": 17433
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.615658970766414e-06,
+      "loss": 1.0205,
+      "step": 17434
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.614632538413616e-06,
+      "loss": 0.8721,
+      "step": 17435
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.613606120941932e-06,
+      "loss": 0.8721,
+      "step": 17436
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.612579718362395e-06,
+      "loss": 1.0898,
+      "step": 17437
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.611553330686024e-06,
+      "loss": 0.9971,
+      "step": 17438
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.610526957923847e-06,
+      "loss": 1.0176,
+      "step": 17439
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60950060008689e-06,
+      "loss": 0.9463,
+      "step": 17440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.608474257186168e-06,
+      "loss": 1.168,
+      "step": 17441
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.607447929232721e-06,
+      "loss": 0.9463,
+      "step": 17442
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.606421616237564e-06,
+      "loss": 1.0332,
+      "step": 17443
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.605395318211725e-06,
+      "loss": 1.3008,
+      "step": 17444
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.604369035166225e-06,
+      "loss": 1.2031,
+      "step": 17445
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.603342767112086e-06,
+      "loss": 0.9727,
+      "step": 17446
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.602316514060338e-06,
+      "loss": 0.9717,
+      "step": 17447
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.601290276022e-06,
+      "loss": 1.1846,
+      "step": 17448
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600264053008102e-06,
+      "loss": 1.2725,
+      "step": 17449
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.599237845029657e-06,
+      "loss": 1.1104,
+      "step": 17450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.598211652097695e-06,
+      "loss": 1.0352,
+      "step": 17451
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.597185474223237e-06,
+      "loss": 1.3252,
+      "step": 17452
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.596159311417311e-06,
+      "loss": 1.0742,
+      "step": 17453
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.595133163690932e-06,
+      "loss": 0.9404,
+      "step": 17454
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.594107031055125e-06,
+      "loss": 0.8545,
+      "step": 17455
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.593080913520916e-06,
+      "loss": 1.0928,
+      "step": 17456
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.592054811099319e-06,
+      "loss": 0.9478,
+      "step": 17457
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.591028723801364e-06,
+      "loss": 1.4453,
+      "step": 17458
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.59000265163807e-06,
+      "loss": 1.042,
+      "step": 17459
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.588976594620462e-06,
+      "loss": 1.0898,
+      "step": 17460
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.587950552759556e-06,
+      "loss": 0.7954,
+      "step": 17461
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.586924526066371e-06,
+      "loss": 1.1201,
+      "step": 17462
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.585898514551937e-06,
+      "loss": 1.2422,
+      "step": 17463
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.584872518227274e-06,
+      "loss": 1.168,
+      "step": 17464
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.583846537103397e-06,
+      "loss": 1.0645,
+      "step": 17465
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.582820571191328e-06,
+      "loss": 1.0479,
+      "step": 17466
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.581794620502088e-06,
+      "loss": 0.9717,
+      "step": 17467
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.580768685046704e-06,
+      "loss": 1.6064,
+      "step": 17468
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.579742764836186e-06,
+      "loss": 0.7153,
+      "step": 17469
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.578716859881561e-06,
+      "loss": 1.1172,
+      "step": 17470
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.577690970193847e-06,
+      "loss": 0.7803,
+      "step": 17471
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57666509578406e-06,
+      "loss": 1.127,
+      "step": 17472
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.575639236663226e-06,
+      "loss": 1.2773,
+      "step": 17473
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.574613392842361e-06,
+      "loss": 0.8921,
+      "step": 17474
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.573587564332486e-06,
+      "loss": 1.0469,
+      "step": 17475
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.572561751144616e-06,
+      "loss": 1.0225,
+      "step": 17476
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.571535953289773e-06,
+      "loss": 1.042,
+      "step": 17477
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.570510170778975e-06,
+      "loss": 1.0391,
+      "step": 17478
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.569484403623245e-06,
+      "loss": 0.72,
+      "step": 17479
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.568458651833592e-06,
+      "loss": 1.0781,
+      "step": 17480
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.567432915421044e-06,
+      "loss": 1.0195,
+      "step": 17481
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.566407194396612e-06,
+      "loss": 0.9595,
+      "step": 17482
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.565381488771317e-06,
+      "loss": 0.8369,
+      "step": 17483
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.564355798556178e-06,
+      "loss": 0.8184,
+      "step": 17484
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.56333012376221e-06,
+      "loss": 1.2168,
+      "step": 17485
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.562304464400433e-06,
+      "loss": 0.8159,
+      "step": 17486
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.561278820481861e-06,
+      "loss": 0.7832,
+      "step": 17487
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.560253192017512e-06,
+      "loss": 0.9619,
+      "step": 17488
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.559227579018404e-06,
+      "loss": 1.165,
+      "step": 17489
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.558201981495555e-06,
+      "loss": 1.1846,
+      "step": 17490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55717639945998e-06,
+      "loss": 0.8135,
+      "step": 17491
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.556150832922696e-06,
+      "loss": 0.8506,
+      "step": 17492
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.555125281894713e-06,
+      "loss": 1.0098,
+      "step": 17493
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.554099746387058e-06,
+      "loss": 0.9502,
+      "step": 17494
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.553074226410743e-06,
+      "loss": 0.9893,
+      "step": 17495
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.552048721976783e-06,
+      "loss": 0.9111,
+      "step": 17496
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.551023233096192e-06,
+      "loss": 0.7261,
+      "step": 17497
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.549997759779981e-06,
+      "loss": 1.1582,
+      "step": 17498
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.548972302039178e-06,
+      "loss": 1.3535,
+      "step": 17499
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.547946859884788e-06,
+      "loss": 0.8748,
+      "step": 17500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546921433327831e-06,
+      "loss": 0.7886,
+      "step": 17501
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54589602237932e-06,
+      "loss": 0.811,
+      "step": 17502
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.544870627050264e-06,
+      "loss": 0.6567,
+      "step": 17503
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.543845247351687e-06,
+      "loss": 0.7427,
+      "step": 17504
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.542819883294598e-06,
+      "loss": 0.9463,
+      "step": 17505
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.541794534890014e-06,
+      "loss": 0.9937,
+      "step": 17506
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.540769202148946e-06,
+      "loss": 0.8267,
+      "step": 17507
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.539743885082406e-06,
+      "loss": 1.3965,
+      "step": 17508
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.538718583701411e-06,
+      "loss": 1.2461,
+      "step": 17509
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.537693298016978e-06,
+      "loss": 1.1348,
+      "step": 17510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.53666802804011e-06,
+      "loss": 1.0449,
+      "step": 17511
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.535642773781831e-06,
+      "loss": 0.9893,
+      "step": 17512
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.534617535253144e-06,
+      "loss": 1.043,
+      "step": 17513
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.533592312465073e-06,
+      "loss": 0.9419,
+      "step": 17514
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.532567105428619e-06,
+      "loss": 0.9512,
+      "step": 17515
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.531541914154803e-06,
+      "loss": 1.4717,
+      "step": 17516
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.530516738654636e-06,
+      "loss": 0.7505,
+      "step": 17517
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529491578939122e-06,
+      "loss": 0.9033,
+      "step": 17518
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.528466435019281e-06,
+      "loss": 0.9629,
+      "step": 17519
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.527441306906122e-06,
+      "loss": 1.1348,
+      "step": 17520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.52641619461066e-06,
+      "loss": 1.0342,
+      "step": 17521
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.525391098143902e-06,
+      "loss": 1.208,
+      "step": 17522
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.524366017516861e-06,
+      "loss": 0.7793,
+      "step": 17523
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.523340952740544e-06,
+      "loss": 1.2314,
+      "step": 17524
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.522315903825973e-06,
+      "loss": 1.1504,
+      "step": 17525
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.521290870784146e-06,
+      "loss": 1.0811,
+      "step": 17526
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.52026585362608e-06,
+      "loss": 0.8242,
+      "step": 17527
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.519240852362788e-06,
+      "loss": 1.0781,
+      "step": 17528
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.518215867005268e-06,
+      "loss": 1.1709,
+      "step": 17529
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.517190897564547e-06,
+      "loss": 0.7334,
+      "step": 17530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.516165944051622e-06,
+      "loss": 0.813,
+      "step": 17531
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.515141006477511e-06,
+      "loss": 1.209,
+      "step": 17532
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.514116084853217e-06,
+      "loss": 0.9448,
+      "step": 17533
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.513091179189749e-06,
+      "loss": 1.1528,
+      "step": 17534
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.512066289498122e-06,
+      "loss": 0.8271,
+      "step": 17535
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.511041415789344e-06,
+      "loss": 1.1807,
+      "step": 17536
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.510016558074419e-06,
+      "loss": 0.783,
+      "step": 17537
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.508991716364359e-06,
+      "loss": 1.4727,
+      "step": 17538
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50796689067017e-06,
+      "loss": 1.1543,
+      "step": 17539
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.506942081002865e-06,
+      "loss": 0.9922,
+      "step": 17540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.505917287373452e-06,
+      "loss": 1.1226,
+      "step": 17541
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50489250979293e-06,
+      "loss": 1.2627,
+      "step": 17542
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50386774827232e-06,
+      "loss": 1.1924,
+      "step": 17543
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.502843002822616e-06,
+      "loss": 0.7061,
+      "step": 17544
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.501818273454837e-06,
+      "loss": 0.9111,
+      "step": 17545
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.500793560179984e-06,
+      "loss": 1.1162,
+      "step": 17546
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.499768863009066e-06,
+      "loss": 1.0244,
+      "step": 17547
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.498744181953091e-06,
+      "loss": 1.2227,
+      "step": 17548
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.497719517023063e-06,
+      "loss": 1.5361,
+      "step": 17549
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.496694868229987e-06,
+      "loss": 1.1279,
+      "step": 17550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.495670235584876e-06,
+      "loss": 1.2617,
+      "step": 17551
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.494645619098734e-06,
+      "loss": 0.9121,
+      "step": 17552
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.493621018782563e-06,
+      "loss": 1.1055,
+      "step": 17553
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49259643464737e-06,
+      "loss": 1.4634,
+      "step": 17554
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.491571866704162e-06,
+      "loss": 0.7056,
+      "step": 17555
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49054731496395e-06,
+      "loss": 1.1914,
+      "step": 17556
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.489522779437732e-06,
+      "loss": 0.8062,
+      "step": 17557
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.488498260136513e-06,
+      "loss": 0.8545,
+      "step": 17558
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.487473757071303e-06,
+      "loss": 0.9785,
+      "step": 17559
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.4864492702531e-06,
+      "loss": 1.0176,
+      "step": 17560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.485424799692917e-06,
+      "loss": 0.8838,
+      "step": 17561
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.484400345401751e-06,
+      "loss": 0.9636,
+      "step": 17562
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.483375907390613e-06,
+      "loss": 0.9863,
+      "step": 17563
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.482351485670504e-06,
+      "loss": 1.0176,
+      "step": 17564
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.481327080252423e-06,
+      "loss": 0.6479,
+      "step": 17565
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.480302691147382e-06,
+      "loss": 1.1255,
+      "step": 17566
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.479278318366381e-06,
+      "loss": 0.8008,
+      "step": 17567
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.478253961920423e-06,
+      "loss": 1.126,
+      "step": 17568
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.477229621820514e-06,
+      "loss": 0.9336,
+      "step": 17569
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47620529807765e-06,
+      "loss": 0.9375,
+      "step": 17570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.475180990702846e-06,
+      "loss": 1.2793,
+      "step": 17571
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.474156699707093e-06,
+      "loss": 0.771,
+      "step": 17572
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.473132425101399e-06,
+      "loss": 0.9863,
+      "step": 17573
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.472108166896766e-06,
+      "loss": 1.2666,
+      "step": 17574
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47108392510419e-06,
+      "loss": 1.0771,
+      "step": 17575
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.470059699734687e-06,
+      "loss": 1.168,
+      "step": 17576
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.469035490799247e-06,
+      "loss": 0.9707,
+      "step": 17577
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46801129830888e-06,
+      "loss": 1.0098,
+      "step": 17578
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.466987122274578e-06,
+      "loss": 0.74,
+      "step": 17579
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.465962962707349e-06,
+      "loss": 0.9478,
+      "step": 17580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.464938819618189e-06,
+      "loss": 1.4199,
+      "step": 17581
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.463914693018106e-06,
+      "loss": 1.1123,
+      "step": 17582
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.462890582918097e-06,
+      "loss": 0.8789,
+      "step": 17583
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.461866489329161e-06,
+      "loss": 0.9473,
+      "step": 17584
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.460842412262305e-06,
+      "loss": 1.3818,
+      "step": 17585
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.459818351728514e-06,
+      "loss": 1.126,
+      "step": 17586
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.458794307738808e-06,
+      "loss": 0.9214,
+      "step": 17587
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.457770280304173e-06,
+      "loss": 1.2021,
+      "step": 17588
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.456746269435618e-06,
+      "loss": 1.334,
+      "step": 17589
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.455722275144134e-06,
+      "loss": 0.8311,
+      "step": 17590
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.454698297440724e-06,
+      "loss": 0.9707,
+      "step": 17591
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.453674336336387e-06,
+      "loss": 1.0273,
+      "step": 17592
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.452650391842123e-06,
+      "loss": 1.0537,
+      "step": 17593
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.451626463968932e-06,
+      "loss": 1.2939,
+      "step": 17594
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45060255272781e-06,
+      "loss": 0.9429,
+      "step": 17595
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.449578658129753e-06,
+      "loss": 1.1836,
+      "step": 17596
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.448554780185766e-06,
+      "loss": 1.0923,
+      "step": 17597
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.447530918906846e-06,
+      "loss": 1.1621,
+      "step": 17598
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.446507074303985e-06,
+      "loss": 0.9688,
+      "step": 17599
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.445483246388185e-06,
+      "loss": 1.1787,
+      "step": 17600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.44445943517044e-06,
+      "loss": 0.96,
+      "step": 17601
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.443435640661757e-06,
+      "loss": 1.2168,
+      "step": 17602
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.442411862873123e-06,
+      "loss": 0.8123,
+      "step": 17603
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.44138810181554e-06,
+      "loss": 0.6802,
+      "step": 17604
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.440364357500004e-06,
+      "loss": 1.1348,
+      "step": 17605
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.439340629937508e-06,
+      "loss": 1.1582,
+      "step": 17606
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.438316919139055e-06,
+      "loss": 0.7827,
+      "step": 17607
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.437293225115639e-06,
+      "loss": 0.9482,
+      "step": 17608
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.436269547878255e-06,
+      "loss": 0.9072,
+      "step": 17609
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.435245887437898e-06,
+      "loss": 1.4033,
+      "step": 17610
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.434222243805566e-06,
+      "loss": 1.1064,
+      "step": 17611
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43319861699225e-06,
+      "loss": 1.334,
+      "step": 17612
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.432175007008954e-06,
+      "loss": 0.6157,
+      "step": 17613
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.431151413866667e-06,
+      "loss": 1.0049,
+      "step": 17614
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.430127837576385e-06,
+      "loss": 1.1543,
+      "step": 17615
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.429104278149106e-06,
+      "loss": 1.0918,
+      "step": 17616
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.428080735595816e-06,
+      "loss": 1.2217,
+      "step": 17617
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.427057209927519e-06,
+      "loss": 0.96,
+      "step": 17618
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.426033701155206e-06,
+      "loss": 0.9424,
+      "step": 17619
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.425010209289872e-06,
+      "loss": 0.8232,
+      "step": 17620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42398673434251e-06,
+      "loss": 1.0156,
+      "step": 17621
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42296327632411e-06,
+      "loss": 1.4038,
+      "step": 17622
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.421939835245672e-06,
+      "loss": 0.7529,
+      "step": 17623
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.420916411118188e-06,
+      "loss": 1.0908,
+      "step": 17624
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41989300395265e-06,
+      "loss": 0.7998,
+      "step": 17625
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41886961376005e-06,
+      "loss": 1.1406,
+      "step": 17626
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.417846240551382e-06,
+      "loss": 1.0654,
+      "step": 17627
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41682288433764e-06,
+      "loss": 1.1211,
+      "step": 17628
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.415799545129816e-06,
+      "loss": 1.0117,
+      "step": 17629
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.414776222938901e-06,
+      "loss": 0.9287,
+      "step": 17630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.413752917775888e-06,
+      "loss": 1.1553,
+      "step": 17631
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.412729629651764e-06,
+      "loss": 0.8926,
+      "step": 17632
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.411706358577533e-06,
+      "loss": 0.8223,
+      "step": 17633
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.410683104564175e-06,
+      "loss": 1.0005,
+      "step": 17634
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.40965986762269e-06,
+      "loss": 1.001,
+      "step": 17635
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.408636647764062e-06,
+      "loss": 1.1123,
+      "step": 17636
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.40761344499928e-06,
+      "loss": 1.0078,
+      "step": 17637
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.406590259339346e-06,
+      "loss": 1.0391,
+      "step": 17638
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.405567090795243e-06,
+      "loss": 0.6133,
+      "step": 17639
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.404543939377965e-06,
+      "loss": 1.126,
+      "step": 17640
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.403520805098497e-06,
+      "loss": 0.8936,
+      "step": 17641
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.402497687967837e-06,
+      "loss": 0.9888,
+      "step": 17642
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.401474587996963e-06,
+      "loss": 1.4648,
+      "step": 17643
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.400451505196878e-06,
+      "loss": 0.8506,
+      "step": 17644
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.399428439578564e-06,
+      "loss": 1.2754,
+      "step": 17645
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.398405391153013e-06,
+      "loss": 0.748,
+      "step": 17646
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.397382359931214e-06,
+      "loss": 1.2314,
+      "step": 17647
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396359345924154e-06,
+      "loss": 0.9951,
+      "step": 17648
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.395336349142821e-06,
+      "loss": 0.8605,
+      "step": 17649
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.394313369598211e-06,
+      "loss": 1.3379,
+      "step": 17650
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.393290407301306e-06,
+      "loss": 0.8408,
+      "step": 17651
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.392267462263095e-06,
+      "loss": 1.0312,
+      "step": 17652
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.391244534494565e-06,
+      "loss": 1.2207,
+      "step": 17653
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.390221624006706e-06,
+      "loss": 1.1143,
+      "step": 17654
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.38919873081051e-06,
+      "loss": 0.7119,
+      "step": 17655
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.388175854916958e-06,
+      "loss": 0.9492,
+      "step": 17656
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.387152996337036e-06,
+      "loss": 0.8779,
+      "step": 17657
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.386130155081736e-06,
+      "loss": 1.1963,
+      "step": 17658
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.385107331162045e-06,
+      "loss": 1.1768,
+      "step": 17659
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.384084524588949e-06,
+      "loss": 0.8965,
+      "step": 17660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.383061735373431e-06,
+      "loss": 0.9082,
+      "step": 17661
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.382038963526487e-06,
+      "loss": 1.3809,
+      "step": 17662
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.381016209059089e-06,
+      "loss": 1.1855,
+      "step": 17663
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.379993471982232e-06,
+      "loss": 1.0332,
+      "step": 17664
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.378970752306903e-06,
+      "loss": 1.1162,
+      "step": 17665
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.377948050044087e-06,
+      "loss": 1.0049,
+      "step": 17666
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.376925365204765e-06,
+      "loss": 1.0508,
+      "step": 17667
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.375902697799926e-06,
+      "loss": 0.9551,
+      "step": 17668
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.374880047840552e-06,
+      "loss": 0.9336,
+      "step": 17669
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.373857415337634e-06,
+      "loss": 1.0801,
+      "step": 17670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37283480030215e-06,
+      "loss": 0.9902,
+      "step": 17671
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37181220274509e-06,
+      "loss": 1.1025,
+      "step": 17672
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.370789622677436e-06,
+      "loss": 1.2783,
+      "step": 17673
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.369767060110165e-06,
+      "loss": 0.9453,
+      "step": 17674
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.368744515054274e-06,
+      "loss": 1.1289,
+      "step": 17675
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.367721987520741e-06,
+      "loss": 1.0762,
+      "step": 17676
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.36669947752055e-06,
+      "loss": 0.8486,
+      "step": 17677
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.365676985064684e-06,
+      "loss": 1.1006,
+      "step": 17678
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.364654510164121e-06,
+      "loss": 0.6455,
+      "step": 17679
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.363632052829853e-06,
+      "loss": 0.8291,
+      "step": 17680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.36260961307286e-06,
+      "loss": 1.1484,
+      "step": 17681
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.361587190904124e-06,
+      "loss": 0.6597,
+      "step": 17682
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.360564786334626e-06,
+      "loss": 1.084,
+      "step": 17683
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.359542399375347e-06,
+      "loss": 0.9512,
+      "step": 17684
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.358520030037273e-06,
+      "loss": 1.4062,
+      "step": 17685
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.357497678331388e-06,
+      "loss": 1.082,
+      "step": 17686
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35647534426867e-06,
+      "loss": 1.1338,
+      "step": 17687
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.355453027860099e-06,
+      "loss": 1.2627,
+      "step": 17688
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.354430729116653e-06,
+      "loss": 1.2051,
+      "step": 17689
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.353408448049327e-06,
+      "loss": 1.1523,
+      "step": 17690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35238618466909e-06,
+      "loss": 1.3926,
+      "step": 17691
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.351363938986925e-06,
+      "loss": 0.7988,
+      "step": 17692
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.350341711013817e-06,
+      "loss": 1.457,
+      "step": 17693
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.349319500760739e-06,
+      "loss": 0.9482,
+      "step": 17694
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.348297308238677e-06,
+      "loss": 0.8682,
+      "step": 17695
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.347275133458608e-06,
+      "loss": 1.2217,
+      "step": 17696
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.346252976431518e-06,
+      "loss": 0.9619,
+      "step": 17697
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34523083716838e-06,
+      "loss": 1.2178,
+      "step": 17698
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.344208715680173e-06,
+      "loss": 1.019,
+      "step": 17699
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.343186611977876e-06,
+      "loss": 1.4062,
+      "step": 17700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.342164526072476e-06,
+      "loss": 0.8428,
+      "step": 17701
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.341142457974945e-06,
+      "loss": 1.1465,
+      "step": 17702
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.340120407696261e-06,
+      "loss": 1.2754,
+      "step": 17703
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.339098375247409e-06,
+      "loss": 1.0117,
+      "step": 17704
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33807636063936e-06,
+      "loss": 0.7686,
+      "step": 17705
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.337054363883095e-06,
+      "loss": 0.9629,
+      "step": 17706
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.336032384989592e-06,
+      "loss": 1.3848,
+      "step": 17707
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33501042396983e-06,
+      "loss": 0.5593,
+      "step": 17708
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.333988480834786e-06,
+      "loss": 1.2373,
+      "step": 17709
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.332966555595432e-06,
+      "loss": 1.0898,
+      "step": 17710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.331944648262754e-06,
+      "loss": 1.2832,
+      "step": 17711
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.330922758847724e-06,
+      "loss": 0.9131,
+      "step": 17712
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.329900887361319e-06,
+      "loss": 0.873,
+      "step": 17713
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.328879033814516e-06,
+      "loss": 0.6953,
+      "step": 17714
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32785719821829e-06,
+      "loss": 1.1006,
+      "step": 17715
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.326835380583621e-06,
+      "loss": 0.8486,
+      "step": 17716
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.325813580921483e-06,
+      "loss": 0.999,
+      "step": 17717
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.324791799242851e-06,
+      "loss": 1.0381,
+      "step": 17718
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.323770035558702e-06,
+      "loss": 0.8213,
+      "step": 17719
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.322748289880005e-06,
+      "loss": 1.1719,
+      "step": 17720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.321726562217747e-06,
+      "loss": 1.1094,
+      "step": 17721
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.320704852582894e-06,
+      "loss": 0.8828,
+      "step": 17722
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.319683160986428e-06,
+      "loss": 1.0449,
+      "step": 17723
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.318661487439313e-06,
+      "loss": 1.1709,
+      "step": 17724
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.317639831952532e-06,
+      "loss": 1.0879,
+      "step": 17725
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.316618194537054e-06,
+      "loss": 1.1982,
+      "step": 17726
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.315596575203858e-06,
+      "loss": 1.084,
+      "step": 17727
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.31457497396392e-06,
+      "loss": 1.2305,
+      "step": 17728
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.313553390828206e-06,
+      "loss": 0.9404,
+      "step": 17729
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.312531825807696e-06,
+      "loss": 1.1191,
+      "step": 17730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.311510278913352e-06,
+      "loss": 0.8374,
+      "step": 17731
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.310488750156164e-06,
+      "loss": 0.8281,
+      "step": 17732
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.309467239547093e-06,
+      "loss": 0.8809,
+      "step": 17733
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.308445747097117e-06,
+      "loss": 1.25,
+      "step": 17734
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.307424272817203e-06,
+      "loss": 0.8506,
+      "step": 17735
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.306402816718327e-06,
+      "loss": 0.8433,
+      "step": 17736
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.305381378811462e-06,
+      "loss": 0.7793,
+      "step": 17737
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.304359959107579e-06,
+      "loss": 0.9751,
+      "step": 17738
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30333855761765e-06,
+      "loss": 0.9756,
+      "step": 17739
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.302317174352647e-06,
+      "loss": 0.9229,
+      "step": 17740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.301295809323534e-06,
+      "loss": 1.2471,
+      "step": 17741
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.300274462541295e-06,
+      "loss": 0.9541,
+      "step": 17742
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.299253134016894e-06,
+      "loss": 0.8843,
+      "step": 17743
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.298231823761301e-06,
+      "loss": 0.9141,
+      "step": 17744
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.297210531785489e-06,
+      "loss": 0.9541,
+      "step": 17745
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.296189258100423e-06,
+      "loss": 1.293,
+      "step": 17746
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.295168002717082e-06,
+      "loss": 0.8945,
+      "step": 17747
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.294146765646428e-06,
+      "loss": 1.0508,
+      "step": 17748
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.293125546899437e-06,
+      "loss": 1.1641,
+      "step": 17749
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.292104346487076e-06,
+      "loss": 1.1694,
+      "step": 17750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.291083164420308e-06,
+      "loss": 1.0869,
+      "step": 17751
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.290062000710112e-06,
+      "loss": 1.0713,
+      "step": 17752
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.289040855367452e-06,
+      "loss": 1.4824,
+      "step": 17753
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.288019728403301e-06,
+      "loss": 1.1289,
+      "step": 17754
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28699861982862e-06,
+      "loss": 1.1719,
+      "step": 17755
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.285977529654383e-06,
+      "loss": 0.9316,
+      "step": 17756
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.284956457891558e-06,
+      "loss": 1.2871,
+      "step": 17757
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.283935404551114e-06,
+      "loss": 0.708,
+      "step": 17758
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.282914369644013e-06,
+      "loss": 0.8213,
+      "step": 17759
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.281893353181228e-06,
+      "loss": 1.2451,
+      "step": 17760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280872355173726e-06,
+      "loss": 1.2441,
+      "step": 17761
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.279851375632469e-06,
+      "loss": 1.1904,
+      "step": 17762
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.27883041456843e-06,
+      "loss": 1.6895,
+      "step": 17763
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.277809471992573e-06,
+      "loss": 0.8066,
+      "step": 17764
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.276788547915868e-06,
+      "loss": 1.1729,
+      "step": 17765
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.275767642349278e-06,
+      "loss": 1.1079,
+      "step": 17766
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.274746755303765e-06,
+      "loss": 0.8003,
+      "step": 17767
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.273725886790302e-06,
+      "loss": 0.8916,
+      "step": 17768
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.272705036819856e-06,
+      "loss": 1.1943,
+      "step": 17769
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.271684205403388e-06,
+      "loss": 1.0664,
+      "step": 17770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.270663392551864e-06,
+      "loss": 0.553,
+      "step": 17771
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.269642598276248e-06,
+      "loss": 1.0967,
+      "step": 17772
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.268621822587508e-06,
+      "loss": 0.7788,
+      "step": 17773
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26760106549661e-06,
+      "loss": 1.2178,
+      "step": 17774
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.266580327014515e-06,
+      "loss": 1.0967,
+      "step": 17775
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26555960715219e-06,
+      "loss": 0.8154,
+      "step": 17776
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.264538905920593e-06,
+      "loss": 1.0215,
+      "step": 17777
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.263518223330698e-06,
+      "loss": 0.9258,
+      "step": 17778
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.262497559393461e-06,
+      "loss": 0.9932,
+      "step": 17779
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.261476914119852e-06,
+      "loss": 1.2529,
+      "step": 17780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.260456287520827e-06,
+      "loss": 0.9106,
+      "step": 17781
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.259435679607351e-06,
+      "loss": 1.0098,
+      "step": 17782
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.258415090390392e-06,
+      "loss": 1.1748,
+      "step": 17783
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25739451988091e-06,
+      "loss": 1.1963,
+      "step": 17784
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.256373968089868e-06,
+      "loss": 1.1094,
+      "step": 17785
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.255353435028228e-06,
+      "loss": 1.1797,
+      "step": 17786
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.254332920706947e-06,
+      "loss": 0.8955,
+      "step": 17787
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.253312425136996e-06,
+      "loss": 0.8428,
+      "step": 17788
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.252291948329334e-06,
+      "loss": 1.0781,
+      "step": 17789
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.251271490294919e-06,
+      "loss": 1.0469,
+      "step": 17790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.250251051044715e-06,
+      "loss": 1.1377,
+      "step": 17791
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.249230630589686e-06,
+      "loss": 0.8286,
+      "step": 17792
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.248210228940782e-06,
+      "loss": 0.729,
+      "step": 17793
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.247189846108978e-06,
+      "loss": 1.0947,
+      "step": 17794
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.246169482105227e-06,
+      "loss": 1.1621,
+      "step": 17795
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.245149136940492e-06,
+      "loss": 0.7383,
+      "step": 17796
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24412881062573e-06,
+      "loss": 1.0469,
+      "step": 17797
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2431085031719e-06,
+      "loss": 1.3369,
+      "step": 17798
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.242088214589967e-06,
+      "loss": 0.9873,
+      "step": 17799
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.241067944890892e-06,
+      "loss": 0.8486,
+      "step": 17800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.240047694085627e-06,
+      "loss": 0.9727,
+      "step": 17801
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.239027462185134e-06,
+      "loss": 0.8232,
+      "step": 17802
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.238007249200372e-06,
+      "loss": 1.3125,
+      "step": 17803
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.236987055142303e-06,
+      "loss": 1.1699,
+      "step": 17804
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.235966880021881e-06,
+      "loss": 0.8936,
+      "step": 17805
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.234946723850068e-06,
+      "loss": 0.9419,
+      "step": 17806
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.233926586637821e-06,
+      "loss": 0.9805,
+      "step": 17807
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.232906468396094e-06,
+      "loss": 0.8604,
+      "step": 17808
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23188636913585e-06,
+      "loss": 1.0352,
+      "step": 17809
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.230866288868045e-06,
+      "loss": 1.5107,
+      "step": 17810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.229846227603638e-06,
+      "loss": 1.0859,
+      "step": 17811
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.228826185353582e-06,
+      "loss": 1.168,
+      "step": 17812
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.227806162128832e-06,
+      "loss": 0.9551,
+      "step": 17813
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.226786157940352e-06,
+      "loss": 1.2422,
+      "step": 17814
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.225766172799099e-06,
+      "loss": 1.3066,
+      "step": 17815
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.224746206716023e-06,
+      "loss": 1.1875,
+      "step": 17816
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.223726259702082e-06,
+      "loss": 0.7803,
+      "step": 17817
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.222706331768231e-06,
+      "loss": 0.9673,
+      "step": 17818
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.221686422925429e-06,
+      "loss": 1.2236,
+      "step": 17819
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22066653318463e-06,
+      "loss": 1.1343,
+      "step": 17820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.21964666255679e-06,
+      "loss": 0.9316,
+      "step": 17821
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.218626811052865e-06,
+      "loss": 0.8945,
+      "step": 17822
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.217606978683806e-06,
+      "loss": 0.9072,
+      "step": 17823
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.216587165460566e-06,
+      "loss": 1.1572,
+      "step": 17824
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.215567371394106e-06,
+      "loss": 0.9492,
+      "step": 17825
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.214547596495378e-06,
+      "loss": 0.7046,
+      "step": 17826
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.213527840775338e-06,
+      "loss": 0.8701,
+      "step": 17827
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.212508104244934e-06,
+      "loss": 0.8975,
+      "step": 17828
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.211488386915123e-06,
+      "loss": 1.1777,
+      "step": 17829
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.210468688796858e-06,
+      "loss": 1.1592,
+      "step": 17830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.209449009901096e-06,
+      "loss": 0.7175,
+      "step": 17831
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208429350238783e-06,
+      "loss": 0.8945,
+      "step": 17832
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.207409709820879e-06,
+      "loss": 1.1455,
+      "step": 17833
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.206390088658326e-06,
+      "loss": 0.8047,
+      "step": 17834
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.205370486762092e-06,
+      "loss": 0.8486,
+      "step": 17835
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.204350904143118e-06,
+      "loss": 0.6592,
+      "step": 17836
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.203331340812358e-06,
+      "loss": 1.0361,
+      "step": 17837
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.202311796780766e-06,
+      "loss": 1.1133,
+      "step": 17838
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.201292272059289e-06,
+      "loss": 1.2715,
+      "step": 17839
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.200272766658882e-06,
+      "loss": 0.7539,
+      "step": 17840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.199253280590497e-06,
+      "loss": 1.166,
+      "step": 17841
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.198233813865084e-06,
+      "loss": 1.019,
+      "step": 17842
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.197214366493594e-06,
+      "loss": 0.9546,
+      "step": 17843
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.196194938486972e-06,
+      "loss": 1.5215,
+      "step": 17844
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.195175529856177e-06,
+      "loss": 0.8281,
+      "step": 17845
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.194156140612157e-06,
+      "loss": 0.9717,
+      "step": 17846
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.193136770765856e-06,
+      "loss": 1.1035,
+      "step": 17847
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19211742032823e-06,
+      "loss": 0.7793,
+      "step": 17848
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.191098089310224e-06,
+      "loss": 0.9912,
+      "step": 17849
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.190078777722794e-06,
+      "loss": 0.7964,
+      "step": 17850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.189059485576882e-06,
+      "loss": 1.0864,
+      "step": 17851
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.18804021288344e-06,
+      "loss": 1.0166,
+      "step": 17852
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.187020959653417e-06,
+      "loss": 0.9468,
+      "step": 17853
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.186001725897761e-06,
+      "loss": 1.2129,
+      "step": 17854
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.184982511627417e-06,
+      "loss": 1.1855,
+      "step": 17855
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.183963316853339e-06,
+      "loss": 1.3037,
+      "step": 17856
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.182944141586471e-06,
+      "loss": 1.1909,
+      "step": 17857
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.181924985837762e-06,
+      "loss": 1.0977,
+      "step": 17858
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.180905849618156e-06,
+      "loss": 0.8721,
+      "step": 17859
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.179886732938605e-06,
+      "loss": 1.2285,
+      "step": 17860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.178867635810054e-06,
+      "loss": 0.7217,
+      "step": 17861
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17784855824345e-06,
+      "loss": 1.4043,
+      "step": 17862
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17682950024974e-06,
+      "loss": 1.3652,
+      "step": 17863
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17581046183987e-06,
+      "loss": 0.8496,
+      "step": 17864
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.174791443024781e-06,
+      "loss": 1.1348,
+      "step": 17865
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17377244381543e-06,
+      "loss": 0.8672,
+      "step": 17866
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.172753464222753e-06,
+      "loss": 0.9351,
+      "step": 17867
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.171734504257702e-06,
+      "loss": 0.9932,
+      "step": 17868
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.170715563931218e-06,
+      "loss": 0.9121,
+      "step": 17869
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.169696643254243e-06,
+      "loss": 0.9629,
+      "step": 17870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16867774223773e-06,
+      "loss": 1.0278,
+      "step": 17871
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16765886089262e-06,
+      "loss": 0.9932,
+      "step": 17872
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.166639999229859e-06,
+      "loss": 0.9707,
+      "step": 17873
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.165621157260389e-06,
+      "loss": 1.4609,
+      "step": 17874
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.164602334995151e-06,
+      "loss": 0.8584,
+      "step": 17875
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.163583532445095e-06,
+      "loss": 0.6616,
+      "step": 17876
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.162564749621163e-06,
+      "loss": 0.8516,
+      "step": 17877
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.161545986534298e-06,
+      "loss": 1.1621,
+      "step": 17878
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.160527243195443e-06,
+      "loss": 0.5327,
+      "step": 17879
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.159508519615535e-06,
+      "loss": 1.001,
+      "step": 17880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.158489815805527e-06,
+      "loss": 1.1709,
+      "step": 17881
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.157471131776356e-06,
+      "loss": 0.9072,
+      "step": 17882
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.156452467538967e-06,
+      "loss": 1.0103,
+      "step": 17883
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1554338231043e-06,
+      "loss": 0.6631,
+      "step": 17884
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.154415198483296e-06,
+      "loss": 0.8652,
+      "step": 17885
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.153396593686897e-06,
+      "loss": 0.9199,
+      "step": 17886
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.152378008726047e-06,
+      "loss": 1.3242,
+      "step": 17887
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.151359443611689e-06,
+      "loss": 0.96,
+      "step": 17888
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.150340898354757e-06,
+      "loss": 1.1484,
+      "step": 17889
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.149322372966195e-06,
+      "loss": 1.0107,
+      "step": 17890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.148303867456944e-06,
+      "loss": 0.9434,
+      "step": 17891
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.147285381837948e-06,
+      "loss": 0.9692,
+      "step": 17892
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.146266916120143e-06,
+      "loss": 0.9241,
+      "step": 17893
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.145248470314469e-06,
+      "loss": 1.3594,
+      "step": 17894
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.144230044431868e-06,
+      "loss": 1.1836,
+      "step": 17895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.143211638483277e-06,
+      "loss": 0.9219,
+      "step": 17896
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.142193252479636e-06,
+      "loss": 0.8857,
+      "step": 17897
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.141174886431885e-06,
+      "loss": 1.2354,
+      "step": 17898
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.140156540350965e-06,
+      "loss": 0.8237,
+      "step": 17899
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.139138214247811e-06,
+      "loss": 1.2227,
+      "step": 17900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.13811990813336e-06,
+      "loss": 1.2529,
+      "step": 17901
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.137101622018557e-06,
+      "loss": 0.8125,
+      "step": 17902
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.136083355914337e-06,
+      "loss": 0.8921,
+      "step": 17903
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.135065109831635e-06,
+      "loss": 0.8115,
+      "step": 17904
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.13404688378139e-06,
+      "loss": 1.0684,
+      "step": 17905
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.13302867777454e-06,
+      "loss": 0.7305,
+      "step": 17906
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.132010491822024e-06,
+      "loss": 0.8608,
+      "step": 17907
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.130992325934777e-06,
+      "loss": 0.9375,
+      "step": 17908
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.129974180123735e-06,
+      "loss": 1.2637,
+      "step": 17909
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12895605439984e-06,
+      "loss": 0.7358,
+      "step": 17910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.127937948774013e-06,
+      "loss": 1.0322,
+      "step": 17911
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.126919863257212e-06,
+      "loss": 1.041,
+      "step": 17912
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.125901797860358e-06,
+      "loss": 0.8765,
+      "step": 17913
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.124883752594394e-06,
+      "loss": 1.2598,
+      "step": 17914
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.123865727470248e-06,
+      "loss": 1.1152,
+      "step": 17915
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.122847722498861e-06,
+      "loss": 0.876,
+      "step": 17916
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.121829737691165e-06,
+      "loss": 0.8916,
+      "step": 17917
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.120811773058098e-06,
+      "loss": 0.7295,
+      "step": 17918
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.119793828610593e-06,
+      "loss": 1.1221,
+      "step": 17919
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.118775904359584e-06,
+      "loss": 1.1494,
+      "step": 17920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.117758000316007e-06,
+      "loss": 1.043,
+      "step": 17921
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.116740116490787e-06,
+      "loss": 1.0483,
+      "step": 17922
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.115722252894872e-06,
+      "loss": 0.9561,
+      "step": 17923
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11470440953919e-06,
+      "loss": 1.166,
+      "step": 17924
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.113686586434669e-06,
+      "loss": 1.0645,
+      "step": 17925
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11266878359225e-06,
+      "loss": 1.1377,
+      "step": 17926
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.111651001022858e-06,
+      "loss": 0.8545,
+      "step": 17927
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.110633238737429e-06,
+      "loss": 1.1973,
+      "step": 17928
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.109615496746899e-06,
+      "loss": 0.8877,
+      "step": 17929
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.108597775062198e-06,
+      "loss": 1.1582,
+      "step": 17930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.107580073694257e-06,
+      "loss": 1.167,
+      "step": 17931
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.106562392654004e-06,
+      "loss": 0.9219,
+      "step": 17932
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.105544731952377e-06,
+      "loss": 0.9629,
+      "step": 17933
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.104527091600307e-06,
+      "loss": 1.2012,
+      "step": 17934
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.103509471608722e-06,
+      "loss": 1.0762,
+      "step": 17935
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.102491871988553e-06,
+      "loss": 1.043,
+      "step": 17936
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10147429275073e-06,
+      "loss": 0.8374,
+      "step": 17937
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.100456733906189e-06,
+      "loss": 0.96,
+      "step": 17938
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.099439195465855e-06,
+      "loss": 1.1143,
+      "step": 17939
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.098421677440659e-06,
+      "loss": 0.8594,
+      "step": 17940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.097404179841533e-06,
+      "loss": 1.3701,
+      "step": 17941
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.096386702679401e-06,
+      "loss": 1.2939,
+      "step": 17942
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.095369245965197e-06,
+      "loss": 0.6592,
+      "step": 17943
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09435180970985e-06,
+      "loss": 0.8545,
+      "step": 17944
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.093334393924292e-06,
+      "loss": 0.8809,
+      "step": 17945
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.092316998619445e-06,
+      "loss": 1.6104,
+      "step": 17946
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09129962380624e-06,
+      "loss": 1.1582,
+      "step": 17947
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.090282269495604e-06,
+      "loss": 0.5991,
+      "step": 17948
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.089264935698471e-06,
+      "loss": 1.1152,
+      "step": 17949
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.088247622425763e-06,
+      "loss": 0.8076,
+      "step": 17950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.08723032968841e-06,
+      "loss": 0.8145,
+      "step": 17951
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.086213057497338e-06,
+      "loss": 0.8564,
+      "step": 17952
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.085195805863471e-06,
+      "loss": 0.9463,
+      "step": 17953
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.084178574797745e-06,
+      "loss": 0.9492,
+      "step": 17954
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.083161364311079e-06,
+      "loss": 0.9146,
+      "step": 17955
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.082144174414405e-06,
+      "loss": 1.0479,
+      "step": 17956
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.081127005118642e-06,
+      "loss": 1.1494,
+      "step": 17957
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.080109856434722e-06,
+      "loss": 0.6934,
+      "step": 17958
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.079092728373569e-06,
+      "loss": 1.3467,
+      "step": 17959
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.078075620946111e-06,
+      "loss": 0.9229,
+      "step": 17960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07705853416327e-06,
+      "loss": 0.9741,
+      "step": 17961
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.076041468035972e-06,
+      "loss": 1.1035,
+      "step": 17962
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07502442257514e-06,
+      "loss": 0.6973,
+      "step": 17963
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.074007397791704e-06,
+      "loss": 0.917,
+      "step": 17964
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.072990393696587e-06,
+      "loss": 1.0273,
+      "step": 17965
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07197341030071e-06,
+      "loss": 1.2168,
+      "step": 17966
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.070956447615e-06,
+      "loss": 1.3438,
+      "step": 17967
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.069939505650376e-06,
+      "loss": 0.707,
+      "step": 17968
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.06892258441777e-06,
+      "loss": 0.7598,
+      "step": 17969
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.067905683928098e-06,
+      "loss": 0.8652,
+      "step": 17970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.066888804192287e-06,
+      "loss": 0.9805,
+      "step": 17971
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.065871945221261e-06,
+      "loss": 1.3711,
+      "step": 17972
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.064855107025937e-06,
+      "loss": 0.6226,
+      "step": 17973
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.063838289617242e-06,
+      "loss": 0.6226,
+      "step": 17974
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.0628214930061e-06,
+      "loss": 1.1299,
+      "step": 17975
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.06180471720343e-06,
+      "loss": 0.9131,
+      "step": 17976
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.060787962220155e-06,
+      "loss": 1.1182,
+      "step": 17977
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.059771228067197e-06,
+      "loss": 0.7251,
+      "step": 17978
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.058754514755468e-06,
+      "loss": 1.0679,
+      "step": 17979
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.057737822295906e-06,
+      "loss": 0.9766,
+      "step": 17980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05672115069942e-06,
+      "loss": 1.1201,
+      "step": 17981
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.055704499976937e-06,
+      "loss": 1.3115,
+      "step": 17982
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.054687870139375e-06,
+      "loss": 1.0708,
+      "step": 17983
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05367126119765e-06,
+      "loss": 1.4238,
+      "step": 17984
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.052654673162688e-06,
+      "loss": 0.9854,
+      "step": 17985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.051638106045408e-06,
+      "loss": 1.0928,
+      "step": 17986
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.050621559856728e-06,
+      "loss": 1.0674,
+      "step": 17987
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.049605034607568e-06,
+      "loss": 1.1562,
+      "step": 17988
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.048588530308844e-06,
+      "loss": 0.9287,
+      "step": 17989
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.047572046971481e-06,
+      "loss": 0.7178,
+      "step": 17990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.046555584606394e-06,
+      "loss": 0.8857,
+      "step": 17991
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.045539143224504e-06,
+      "loss": 0.8711,
+      "step": 17992
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044522722836725e-06,
+      "loss": 1.0283,
+      "step": 17993
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.043506323453976e-06,
+      "loss": 1.0166,
+      "step": 17994
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.04248994508718e-06,
+      "loss": 0.9512,
+      "step": 17995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.041473587747247e-06,
+      "loss": 0.54,
+      "step": 17996
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.0404572514451e-06,
+      "loss": 1.1597,
+      "step": 17997
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.039440936191656e-06,
+      "loss": 0.8711,
+      "step": 17998
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.038424641997824e-06,
+      "loss": 1.1689,
+      "step": 17999
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.037408368874531e-06,
+      "loss": 1.0889,
+      "step": 18000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036392116832689e-06,
+      "loss": 1.334,
+      "step": 18001
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.035375885883217e-06,
+      "loss": 0.6619,
+      "step": 18002
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.034359676037026e-06,
+      "loss": 1.2549,
+      "step": 18003
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.033343487305035e-06,
+      "loss": 1.0405,
+      "step": 18004
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.032327319698154e-06,
+      "loss": 0.8896,
+      "step": 18005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.031311173227308e-06,
+      "loss": 0.9639,
+      "step": 18006
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.030295047903406e-06,
+      "loss": 0.8535,
+      "step": 18007
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.029278943737363e-06,
+      "loss": 0.6904,
+      "step": 18008
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.028262860740097e-06,
+      "loss": 0.7031,
+      "step": 18009
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.027246798922513e-06,
+      "loss": 0.869,
+      "step": 18010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02623075829554e-06,
+      "loss": 1.123,
+      "step": 18011
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.025214738870081e-06,
+      "loss": 1.1709,
+      "step": 18012
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.024198740657054e-06,
+      "loss": 0.957,
+      "step": 18013
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02318276366737e-06,
+      "loss": 0.9722,
+      "step": 18014
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.022166807911942e-06,
+      "loss": 0.8184,
+      "step": 18015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.021150873401685e-06,
+      "loss": 0.8447,
+      "step": 18016
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.020134960147512e-06,
+      "loss": 0.918,
+      "step": 18017
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.019119068160338e-06,
+      "loss": 1.2651,
+      "step": 18018
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01810319745107e-06,
+      "loss": 0.8813,
+      "step": 18019
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01708734803062e-06,
+      "loss": 0.8179,
+      "step": 18020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.016071519909902e-06,
+      "loss": 1.0205,
+      "step": 18021
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.015055713099832e-06,
+      "loss": 1.0684,
+      "step": 18022
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.014039927611317e-06,
+      "loss": 1.1162,
+      "step": 18023
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.013024163455268e-06,
+      "loss": 1.0908,
+      "step": 18024
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.012008420642592e-06,
+      "loss": 0.8369,
+      "step": 18025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.010992699184208e-06,
+      "loss": 0.8945,
+      "step": 18026
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.009976999091023e-06,
+      "loss": 1.1152,
+      "step": 18027
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.008961320373947e-06,
+      "loss": 1.1436,
+      "step": 18028
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.007945663043893e-06,
+      "loss": 0.7402,
+      "step": 18029
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.006930027111761e-06,
+      "loss": 1.083,
+      "step": 18030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.005914412588472e-06,
+      "loss": 0.98,
+      "step": 18031
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00489881948493e-06,
+      "loss": 1.04,
+      "step": 18032
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.003883247812047e-06,
+      "loss": 1.5488,
+      "step": 18033
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.002867697580727e-06,
+      "loss": 1.1309,
+      "step": 18034
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.001852168801883e-06,
+      "loss": 0.9336,
+      "step": 18035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00083666148642e-06,
+      "loss": 1.124,
+      "step": 18036
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.999821175645251e-06,
+      "loss": 1.0215,
+      "step": 18037
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.998805711289281e-06,
+      "loss": 0.9023,
+      "step": 18038
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.997790268429417e-06,
+      "loss": 0.9082,
+      "step": 18039
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99677484707657e-06,
+      "loss": 0.8877,
+      "step": 18040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99575944724164e-06,
+      "loss": 0.9111,
+      "step": 18041
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.994744068935541e-06,
+      "loss": 0.7363,
+      "step": 18042
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.993728712169177e-06,
+      "loss": 0.9238,
+      "step": 18043
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.992713376953461e-06,
+      "loss": 1.126,
+      "step": 18044
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.991698063299288e-06,
+      "loss": 1.23,
+      "step": 18045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.990682771217568e-06,
+      "loss": 1.0459,
+      "step": 18046
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98966750071921e-06,
+      "loss": 0.958,
+      "step": 18047
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.988652251815123e-06,
+      "loss": 1.0908,
+      "step": 18048
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.987637024516206e-06,
+      "loss": 0.9316,
+      "step": 18049
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.986621818833364e-06,
+      "loss": 0.9395,
+      "step": 18050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.985606634777504e-06,
+      "loss": 1.2939,
+      "step": 18051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.984591472359533e-06,
+      "loss": 1.1523,
+      "step": 18052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98357633159035e-06,
+      "loss": 0.8354,
+      "step": 18053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.982561212480863e-06,
+      "loss": 1.0601,
+      "step": 18054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.981546115041981e-06,
+      "loss": 0.8545,
+      "step": 18055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.980531039284593e-06,
+      "loss": 1.0552,
+      "step": 18056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97951598521962e-06,
+      "loss": 0.8838,
+      "step": 18057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.978500952857952e-06,
+      "loss": 0.9746,
+      "step": 18058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.977485942210504e-06,
+      "loss": 0.8936,
+      "step": 18059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.976470953288167e-06,
+      "loss": 1.2715,
+      "step": 18060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.975455986101848e-06,
+      "loss": 1.0146,
+      "step": 18061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97444104066245e-06,
+      "loss": 0.9482,
+      "step": 18062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97342611698088e-06,
+      "loss": 1.2036,
+      "step": 18063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.972411215068035e-06,
+      "loss": 0.835,
+      "step": 18064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.971396334934815e-06,
+      "loss": 0.8955,
+      "step": 18065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.970381476592124e-06,
+      "loss": 0.8643,
+      "step": 18066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.969366640050859e-06,
+      "loss": 1.1411,
+      "step": 18067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.968351825321932e-06,
+      "loss": 0.8784,
+      "step": 18068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.967337032416232e-06,
+      "loss": 0.8467,
+      "step": 18069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.966322261344666e-06,
+      "loss": 0.9961,
+      "step": 18070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.965307512118135e-06,
+      "loss": 1.1426,
+      "step": 18071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96429278474753e-06,
+      "loss": 0.8447,
+      "step": 18072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.963278079243761e-06,
+      "loss": 1.1797,
+      "step": 18073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.962263395617724e-06,
+      "loss": 1.2578,
+      "step": 18074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96124873388032e-06,
+      "loss": 1.0654,
+      "step": 18075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.960234094042445e-06,
+      "loss": 0.8633,
+      "step": 18076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.959219476114997e-06,
+      "loss": 0.6816,
+      "step": 18077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95820488010888e-06,
+      "loss": 1.1816,
+      "step": 18078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95719030603499e-06,
+      "loss": 0.8462,
+      "step": 18079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.956175753904224e-06,
+      "loss": 1.1289,
+      "step": 18080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95516122372748e-06,
+      "loss": 1.2715,
+      "step": 18081
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.954146715515654e-06,
+      "loss": 1.02,
+      "step": 18082
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95313222927965e-06,
+      "loss": 1.0371,
+      "step": 18083
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952117765030359e-06,
+      "loss": 0.592,
+      "step": 18084
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95110332277868e-06,
+      "loss": 0.7295,
+      "step": 18085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.950088902535514e-06,
+      "loss": 0.9678,
+      "step": 18086
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.949074504311747e-06,
+      "loss": 1.251,
+      "step": 18087
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.948060128118283e-06,
+      "loss": 1.0869,
+      "step": 18088
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.947045773966018e-06,
+      "loss": 1.0547,
+      "step": 18089
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.94603144186585e-06,
+      "loss": 1.0586,
+      "step": 18090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.945017131828666e-06,
+      "loss": 1.3975,
+      "step": 18091
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.944002843865367e-06,
+      "loss": 0.9609,
+      "step": 18092
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.942988577986849e-06,
+      "loss": 0.877,
+      "step": 18093
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.941974334204009e-06,
+      "loss": 1.2285,
+      "step": 18094
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.940960112527734e-06,
+      "loss": 0.9209,
+      "step": 18095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.939945912968922e-06,
+      "loss": 1.126,
+      "step": 18096
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93893173553847e-06,
+      "loss": 1.1553,
+      "step": 18097
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.937917580247265e-06,
+      "loss": 0.8613,
+      "step": 18098
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93690344710621e-06,
+      "loss": 0.9197,
+      "step": 18099
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.935889336126193e-06,
+      "loss": 1.0322,
+      "step": 18100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93487524731811e-06,
+      "loss": 0.8638,
+      "step": 18101
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93386118069285e-06,
+      "loss": 0.9756,
+      "step": 18102
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.932847136261304e-06,
+      "loss": 0.833,
+      "step": 18103
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.931833114034373e-06,
+      "loss": 1.0029,
+      "step": 18104
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.930819114022944e-06,
+      "loss": 0.958,
+      "step": 18105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92980513623791e-06,
+      "loss": 0.9346,
+      "step": 18106
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.928791180690159e-06,
+      "loss": 0.5693,
+      "step": 18107
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.927777247390586e-06,
+      "loss": 0.7925,
+      "step": 18108
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.926763336350084e-06,
+      "loss": 1.0645,
+      "step": 18109
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.925749447579543e-06,
+      "loss": 0.8076,
+      "step": 18110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.924735581089853e-06,
+      "loss": 1.021,
+      "step": 18111
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.923721736891905e-06,
+      "loss": 0.8721,
+      "step": 18112
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.922707914996584e-06,
+      "loss": 1.4355,
+      "step": 18113
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92169411541479e-06,
+      "loss": 1.1143,
+      "step": 18114
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.920680338157407e-06,
+      "loss": 0.7891,
+      "step": 18115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.919666583235325e-06,
+      "loss": 1.0732,
+      "step": 18116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.918652850659435e-06,
+      "loss": 0.9717,
+      "step": 18117
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.917639140440623e-06,
+      "loss": 0.8213,
+      "step": 18118
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.916625452589782e-06,
+      "loss": 1.3115,
+      "step": 18119
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.915611787117797e-06,
+      "loss": 0.9639,
+      "step": 18120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91459814403556e-06,
+      "loss": 0.9331,
+      "step": 18121
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.913584523353958e-06,
+      "loss": 0.8926,
+      "step": 18122
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.912570925083873e-06,
+      "loss": 1.4932,
+      "step": 18123
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.911557349236202e-06,
+      "loss": 0.8799,
+      "step": 18124
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91054379582183e-06,
+      "loss": 1.2344,
+      "step": 18125
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90953026485164e-06,
+      "loss": 1.3281,
+      "step": 18126
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.908516756336523e-06,
+      "loss": 0.8276,
+      "step": 18127
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.907503270287365e-06,
+      "loss": 1.3213,
+      "step": 18128
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90648980671505e-06,
+      "loss": 0.7339,
+      "step": 18129
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.905476365630466e-06,
+      "loss": 1.2695,
+      "step": 18130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.9044629470445e-06,
+      "loss": 0.7539,
+      "step": 18131
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90344955096804e-06,
+      "loss": 0.8486,
+      "step": 18132
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.902436177411963e-06,
+      "loss": 0.8877,
+      "step": 18133
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90142282638716e-06,
+      "loss": 0.876,
+      "step": 18134
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.900409497904518e-06,
+      "loss": 0.833,
+      "step": 18135
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.899396191974922e-06,
+      "loss": 1.1562,
+      "step": 18136
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89838290860925e-06,
+      "loss": 0.8823,
+      "step": 18137
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.897369647818394e-06,
+      "loss": 0.7808,
+      "step": 18138
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89635640961323e-06,
+      "loss": 1.1934,
+      "step": 18139
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.895343194004651e-06,
+      "loss": 1.2734,
+      "step": 18140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.894330001003535e-06,
+      "loss": 0.8271,
+      "step": 18141
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.893316830620766e-06,
+      "loss": 1.0449,
+      "step": 18142
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.892303682867229e-06,
+      "loss": 1.2822,
+      "step": 18143
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.891290557753797e-06,
+      "loss": 0.9395,
+      "step": 18144
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.890277455291371e-06,
+      "loss": 0.9819,
+      "step": 18145
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88926437549082e-06,
+      "loss": 0.6943,
+      "step": 18146
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88825131836303e-06,
+      "loss": 0.9268,
+      "step": 18147
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.887238283918881e-06,
+      "loss": 0.9993,
+      "step": 18148
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.886225272169256e-06,
+      "loss": 1.3115,
+      "step": 18149
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.885212283125037e-06,
+      "loss": 0.9004,
+      "step": 18150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.884199316797107e-06,
+      "loss": 0.7036,
+      "step": 18151
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.883186373196343e-06,
+      "loss": 1.209,
+      "step": 18152
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.882173452333625e-06,
+      "loss": 0.9043,
+      "step": 18153
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.881160554219836e-06,
+      "loss": 1.1709,
+      "step": 18154
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.880147678865856e-06,
+      "loss": 0.9775,
+      "step": 18155
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.879134826282568e-06,
+      "loss": 1.0098,
+      "step": 18156
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.878121996480848e-06,
+      "loss": 1.0977,
+      "step": 18157
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.877109189471576e-06,
+      "loss": 0.9688,
+      "step": 18158
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.876096405265629e-06,
+      "loss": 0.8594,
+      "step": 18159
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.875083643873886e-06,
+      "loss": 0.9629,
+      "step": 18160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.87407090530723e-06,
+      "loss": 0.7075,
+      "step": 18161
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.873058189576537e-06,
+      "loss": 0.8677,
+      "step": 18162
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.872045496692688e-06,
+      "loss": 0.6919,
+      "step": 18163
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.871032826666557e-06,
+      "loss": 0.6099,
+      "step": 18164
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870020179509022e-06,
+      "loss": 0.834,
+      "step": 18165
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86900755523096e-06,
+      "loss": 1.0703,
+      "step": 18166
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.867994953843257e-06,
+      "loss": 1.2119,
+      "step": 18167
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.866982375356777e-06,
+      "loss": 1.252,
+      "step": 18168
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.865969819782404e-06,
+      "loss": 1.0625,
+      "step": 18169
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.864957287131012e-06,
+      "loss": 1.207,
+      "step": 18170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.863944777413481e-06,
+      "loss": 1.2549,
+      "step": 18171
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.862932290640684e-06,
+      "loss": 0.8574,
+      "step": 18172
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.861919826823497e-06,
+      "loss": 0.9805,
+      "step": 18173
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.860907385972797e-06,
+      "loss": 0.7246,
+      "step": 18174
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.859894968099454e-06,
+      "loss": 1.0566,
+      "step": 18175
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.858882573214349e-06,
+      "loss": 0.9541,
+      "step": 18176
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.857870201328355e-06,
+      "loss": 0.7275,
+      "step": 18177
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85685785245235e-06,
+      "loss": 1.1338,
+      "step": 18178
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.855845526597202e-06,
+      "loss": 1.293,
+      "step": 18179
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.854833223773784e-06,
+      "loss": 0.7798,
+      "step": 18180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.853820943992977e-06,
+      "loss": 0.8965,
+      "step": 18181
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.852808687265653e-06,
+      "loss": 0.8721,
+      "step": 18182
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85179645360268e-06,
+      "loss": 1.0137,
+      "step": 18183
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850784243014936e-06,
+      "loss": 1.0293,
+      "step": 18184
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84977205551329e-06,
+      "loss": 1.1572,
+      "step": 18185
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84875989110862e-06,
+      "loss": 0.7817,
+      "step": 18186
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.847747749811794e-06,
+      "loss": 0.8359,
+      "step": 18187
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.846735631633685e-06,
+      "loss": 1.292,
+      "step": 18188
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.845723536585165e-06,
+      "loss": 1.0684,
+      "step": 18189
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.844711464677106e-06,
+      "loss": 1.124,
+      "step": 18190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.843699415920377e-06,
+      "loss": 1.1221,
+      "step": 18191
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84268739032585e-06,
+      "loss": 0.7642,
+      "step": 18192
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.841675387904401e-06,
+      "loss": 0.8589,
+      "step": 18193
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.840663408666893e-06,
+      "loss": 1.0381,
+      "step": 18194
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8396514526242e-06,
+      "loss": 0.8965,
+      "step": 18195
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83863951978719e-06,
+      "loss": 0.8081,
+      "step": 18196
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.837627610166738e-06,
+      "loss": 0.8369,
+      "step": 18197
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.836615723773709e-06,
+      "loss": 1.0713,
+      "step": 18198
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.835603860618973e-06,
+      "loss": 1.2676,
+      "step": 18199
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8345920207134e-06,
+      "loss": 0.7988,
+      "step": 18200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.833580204067855e-06,
+      "loss": 0.916,
+      "step": 18201
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.832568410693212e-06,
+      "loss": 1.4004,
+      "step": 18202
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.831556640600338e-06,
+      "loss": 1.0874,
+      "step": 18203
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.830544893800102e-06,
+      "loss": 1.1553,
+      "step": 18204
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.829533170303367e-06,
+      "loss": 0.6587,
+      "step": 18205
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.828521470121001e-06,
+      "loss": 1.2822,
+      "step": 18206
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.827509793263877e-06,
+      "loss": 0.918,
+      "step": 18207
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.826498139742859e-06,
+      "loss": 0.9038,
+      "step": 18208
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.825486509568816e-06,
+      "loss": 0.7891,
+      "step": 18209
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.824474902752609e-06,
+      "loss": 1.4053,
+      "step": 18210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.823463319305107e-06,
+      "loss": 0.9526,
+      "step": 18211
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.822451759237177e-06,
+      "loss": 1.2461,
+      "step": 18212
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.821440222559688e-06,
+      "loss": 1.2441,
+      "step": 18213
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8204287092835e-06,
+      "loss": 1.3701,
+      "step": 18214
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.81941721941948e-06,
+      "loss": 0.833,
+      "step": 18215
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.81840575297849e-06,
+      "loss": 0.6978,
+      "step": 18216
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.817394309971405e-06,
+      "loss": 1.4072,
+      "step": 18217
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.81638289040908e-06,
+      "loss": 1.1543,
+      "step": 18218
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.815371494302382e-06,
+      "loss": 1.1338,
+      "step": 18219
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.814360121662176e-06,
+      "loss": 1.3896,
+      "step": 18220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.813348772499325e-06,
+      "loss": 0.7769,
+      "step": 18221
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.812337446824688e-06,
+      "loss": 1.1709,
+      "step": 18222
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.811326144649138e-06,
+      "loss": 1.1055,
+      "step": 18223
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.810314865983533e-06,
+      "loss": 0.9824,
+      "step": 18224
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.809303610838733e-06,
+      "loss": 1.1201,
+      "step": 18225
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.808292379225605e-06,
+      "loss": 0.8379,
+      "step": 18226
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.807281171155006e-06,
+      "loss": 1.0381,
+      "step": 18227
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.806269986637806e-06,
+      "loss": 0.96,
+      "step": 18228
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80525882568486e-06,
+      "loss": 0.9756,
+      "step": 18229
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.804247688307032e-06,
+      "loss": 1.0518,
+      "step": 18230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.803236574515186e-06,
+      "loss": 0.8467,
+      "step": 18231
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.802225484320174e-06,
+      "loss": 0.9731,
+      "step": 18232
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.801214417732866e-06,
+      "loss": 1.2373,
+      "step": 18233
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80020337476412e-06,
+      "loss": 0.7622,
+      "step": 18234
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.799192355424797e-06,
+      "loss": 1.0234,
+      "step": 18235
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.798181359725754e-06,
+      "loss": 1.1426,
+      "step": 18236
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79717038767785e-06,
+      "loss": 0.9482,
+      "step": 18237
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.796159439291948e-06,
+      "loss": 1.2578,
+      "step": 18238
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.795148514578911e-06,
+      "loss": 1.0078,
+      "step": 18239
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79413761354959e-06,
+      "loss": 0.9746,
+      "step": 18240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.793126736214845e-06,
+      "loss": 1.0703,
+      "step": 18241
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.792115882585536e-06,
+      "loss": 0.8232,
+      "step": 18242
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.791105052672523e-06,
+      "loss": 1.019,
+      "step": 18243
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.790094246486664e-06,
+      "loss": 1.1211,
+      "step": 18244
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.789083464038814e-06,
+      "loss": 0.6963,
+      "step": 18245
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.788072705339833e-06,
+      "loss": 0.8828,
+      "step": 18246
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.787061970400574e-06,
+      "loss": 1.0845,
+      "step": 18247
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.7860512592319e-06,
+      "loss": 1.2466,
+      "step": 18248
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.785040571844662e-06,
+      "loss": 1.1438,
+      "step": 18249
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.784029908249722e-06,
+      "loss": 1.1045,
+      "step": 18250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.783019268457931e-06,
+      "loss": 1.2812,
+      "step": 18251
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.782008652480148e-06,
+      "loss": 0.8721,
+      "step": 18252
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.780998060327225e-06,
+      "loss": 0.9033,
+      "step": 18253
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.779987492010022e-06,
+      "loss": 0.9316,
+      "step": 18254
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.778976947539396e-06,
+      "loss": 0.9551,
+      "step": 18255
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.777966426926194e-06,
+      "loss": 1.0947,
+      "step": 18256
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.776955930181278e-06,
+      "loss": 1.145,
+      "step": 18257
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.775945457315493e-06,
+      "loss": 1.2324,
+      "step": 18258
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.774935008339705e-06,
+      "loss": 0.7842,
+      "step": 18259
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77392458326476e-06,
+      "loss": 1.418,
+      "step": 18260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.772914182101516e-06,
+      "loss": 0.9707,
+      "step": 18261
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.771903804860826e-06,
+      "loss": 1.0547,
+      "step": 18262
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.770893451553537e-06,
+      "loss": 1.2822,
+      "step": 18263
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.769883122190507e-06,
+      "loss": 1.2344,
+      "step": 18264
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.768872816782588e-06,
+      "loss": 1.0312,
+      "step": 18265
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.767862535340635e-06,
+      "loss": 0.5581,
+      "step": 18266
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.766852277875495e-06,
+      "loss": 0.9092,
+      "step": 18267
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.765842044398019e-06,
+      "loss": 1.0674,
+      "step": 18268
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.764831834919067e-06,
+      "loss": 0.9004,
+      "step": 18269
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.763821649449484e-06,
+      "loss": 0.9961,
+      "step": 18270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.762811488000121e-06,
+      "loss": 1.0723,
+      "step": 18271
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76180135058183e-06,
+      "loss": 1.1641,
+      "step": 18272
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.760791237205459e-06,
+      "loss": 1.249,
+      "step": 18273
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.759781147881864e-06,
+      "loss": 0.8301,
+      "step": 18274
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.758771082621889e-06,
+      "loss": 1.2051,
+      "step": 18275
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.75776104143639e-06,
+      "loss": 1.0469,
+      "step": 18276
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.756751024336211e-06,
+      "loss": 1.1348,
+      "step": 18277
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.7557410313322e-06,
+      "loss": 1.2383,
+      "step": 18278
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.754731062435212e-06,
+      "loss": 0.6106,
+      "step": 18279
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.753721117656092e-06,
+      "loss": 0.8281,
+      "step": 18280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.752711197005692e-06,
+      "loss": 1.2695,
+      "step": 18281
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.751701300494857e-06,
+      "loss": 0.8008,
+      "step": 18282
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.750691428134432e-06,
+      "loss": 0.8154,
+      "step": 18283
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.749681579935269e-06,
+      "loss": 1.0449,
+      "step": 18284
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.748671755908218e-06,
+      "loss": 1.3721,
+      "step": 18285
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747661956064118e-06,
+      "loss": 1.0967,
+      "step": 18286
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.746652180413824e-06,
+      "loss": 1.0801,
+      "step": 18287
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.745642428968178e-06,
+      "loss": 1.0762,
+      "step": 18288
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.744632701738024e-06,
+      "loss": 0.8767,
+      "step": 18289
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.743622998734217e-06,
+      "loss": 1.0811,
+      "step": 18290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.742613319967597e-06,
+      "loss": 0.9692,
+      "step": 18291
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.741603665449011e-06,
+      "loss": 1.1113,
+      "step": 18292
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.740594035189301e-06,
+      "loss": 0.8984,
+      "step": 18293
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.739584429199313e-06,
+      "loss": 1.1514,
+      "step": 18294
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.738574847489897e-06,
+      "loss": 0.9961,
+      "step": 18295
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.737565290071892e-06,
+      "loss": 1.1211,
+      "step": 18296
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.736555756956148e-06,
+      "loss": 0.6669,
+      "step": 18297
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.735546248153503e-06,
+      "loss": 0.8374,
+      "step": 18298
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.7345367636748e-06,
+      "loss": 1.0996,
+      "step": 18299
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.733527303530889e-06,
+      "loss": 0.9141,
+      "step": 18300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.732517867732612e-06,
+      "loss": 0.8423,
+      "step": 18301
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.731508456290807e-06,
+      "loss": 1.0439,
+      "step": 18302
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.730499069216321e-06,
+      "loss": 0.9385,
+      "step": 18303
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.729489706519991e-06,
+      "loss": 0.9824,
+      "step": 18304
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.728480368212669e-06,
+      "loss": 0.9536,
+      "step": 18305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.727471054305187e-06,
+      "loss": 1.0146,
+      "step": 18306
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.726461764808394e-06,
+      "loss": 0.7705,
+      "step": 18307
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72545249973313e-06,
+      "loss": 0.8379,
+      "step": 18308
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.724443259090229e-06,
+      "loss": 1.1816,
+      "step": 18309
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.723434042890541e-06,
+      "loss": 0.9346,
+      "step": 18310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.722424851144902e-06,
+      "loss": 0.7588,
+      "step": 18311
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721415683864155e-06,
+      "loss": 1.0986,
+      "step": 18312
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.720406541059137e-06,
+      "loss": 1.001,
+      "step": 18313
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71939742274069e-06,
+      "loss": 0.7959,
+      "step": 18314
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.718388328919652e-06,
+      "loss": 1.0439,
+      "step": 18315
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.717379259606865e-06,
+      "loss": 1.0459,
+      "step": 18316
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.716370214813167e-06,
+      "loss": 1.0576,
+      "step": 18317
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.715361194549394e-06,
+      "loss": 0.562,
+      "step": 18318
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71435219882639e-06,
+      "loss": 1.041,
+      "step": 18319
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.713343227654986e-06,
+      "loss": 0.7676,
+      "step": 18320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.712334281046025e-06,
+      "loss": 1.1318,
+      "step": 18321
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.711325359010344e-06,
+      "loss": 0.9707,
+      "step": 18322
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.710316461558783e-06,
+      "loss": 0.9634,
+      "step": 18323
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.709307588702174e-06,
+      "loss": 1.2031,
+      "step": 18324
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.708298740451354e-06,
+      "loss": 0.6587,
+      "step": 18325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.707289916817164e-06,
+      "loss": 1.0137,
+      "step": 18326
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.70628111781044e-06,
+      "loss": 0.9756,
+      "step": 18327
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.705272343442018e-06,
+      "loss": 0.8662,
+      "step": 18328
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.704263593722729e-06,
+      "loss": 0.9102,
+      "step": 18329
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.703254868663413e-06,
+      "loss": 1.0439,
+      "step": 18330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.702246168274906e-06,
+      "loss": 1.0654,
+      "step": 18331
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.70123749256804e-06,
+      "loss": 1.0706,
+      "step": 18332
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.700228841553652e-06,
+      "loss": 0.9932,
+      "step": 18333
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.699220215242578e-06,
+      "loss": 0.7354,
+      "step": 18334
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.698211613645643e-06,
+      "loss": 1.126,
+      "step": 18335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.697203036773696e-06,
+      "loss": 0.7852,
+      "step": 18336
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.696194484637561e-06,
+      "loss": 1.0645,
+      "step": 18337
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.695185957248074e-06,
+      "loss": 1.0127,
+      "step": 18338
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.694177454616068e-06,
+      "loss": 0.5625,
+      "step": 18339
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.693168976752376e-06,
+      "loss": 1.2441,
+      "step": 18340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.692160523667826e-06,
+      "loss": 1.021,
+      "step": 18341
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69115209537326e-06,
+      "loss": 0.9531,
+      "step": 18342
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.690143691879505e-06,
+      "loss": 0.7139,
+      "step": 18343
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.689135313197391e-06,
+      "loss": 1.2783,
+      "step": 18344
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.688126959337754e-06,
+      "loss": 1.333,
+      "step": 18345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.687118630311418e-06,
+      "loss": 0.5896,
+      "step": 18346
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.686110326129223e-06,
+      "loss": 0.8164,
+      "step": 18347
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.685102046801997e-06,
+      "loss": 0.7529,
+      "step": 18348
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68409379234057e-06,
+      "loss": 1.0166,
+      "step": 18349
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68308556275577e-06,
+      "loss": 0.915,
+      "step": 18350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.682077358058426e-06,
+      "loss": 1.2715,
+      "step": 18351
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.681069178259373e-06,
+      "loss": 0.8242,
+      "step": 18352
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68006102336944e-06,
+      "loss": 0.6509,
+      "step": 18353
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.679052893399454e-06,
+      "loss": 0.7852,
+      "step": 18354
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.678044788360244e-06,
+      "loss": 1.3555,
+      "step": 18355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.677036708262638e-06,
+      "loss": 0.9121,
+      "step": 18356
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.676028653117465e-06,
+      "loss": 0.8047,
+      "step": 18357
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.675020622935558e-06,
+      "loss": 0.9668,
+      "step": 18358
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.674012617727736e-06,
+      "loss": 1.2002,
+      "step": 18359
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.673004637504833e-06,
+      "loss": 1.3535,
+      "step": 18360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.671996682277673e-06,
+      "loss": 1.0654,
+      "step": 18361
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.670988752057088e-06,
+      "loss": 0.9863,
+      "step": 18362
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6699808468539e-06,
+      "loss": 1.0664,
+      "step": 18363
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.668972966678935e-06,
+      "loss": 0.8916,
+      "step": 18364
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.667965111543025e-06,
+      "loss": 1.1919,
+      "step": 18365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.666957281456989e-06,
+      "loss": 0.8818,
+      "step": 18366
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.665949476431658e-06,
+      "loss": 0.7139,
+      "step": 18367
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664941696477854e-06,
+      "loss": 1.0127,
+      "step": 18368
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.663933941606407e-06,
+      "loss": 1.04,
+      "step": 18369
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.662926211828136e-06,
+      "loss": 0.8057,
+      "step": 18370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66191850715387e-06,
+      "loss": 0.6484,
+      "step": 18371
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.660910827594428e-06,
+      "loss": 0.9985,
+      "step": 18372
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.659903173160642e-06,
+      "loss": 1.0342,
+      "step": 18373
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65889554386333e-06,
+      "loss": 1.4824,
+      "step": 18374
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.657887939713318e-06,
+      "loss": 1.2979,
+      "step": 18375
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.656880360721432e-06,
+      "loss": 0.604,
+      "step": 18376
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.655872806898486e-06,
+      "loss": 1.0068,
+      "step": 18377
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65486527825531e-06,
+      "loss": 1.4336,
+      "step": 18378
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.653857774802725e-06,
+      "loss": 1.3965,
+      "step": 18379
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.652850296551557e-06,
+      "loss": 1.1465,
+      "step": 18380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65184284351262e-06,
+      "loss": 1.2578,
+      "step": 18381
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65083541569674e-06,
+      "loss": 1.1387,
+      "step": 18382
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.649828013114739e-06,
+      "loss": 0.8379,
+      "step": 18383
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.648820635777441e-06,
+      "loss": 0.9346,
+      "step": 18384
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.64781328369566e-06,
+      "loss": 0.9717,
+      "step": 18385
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.646805956880222e-06,
+      "loss": 1.1904,
+      "step": 18386
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.645798655341941e-06,
+      "loss": 0.8564,
+      "step": 18387
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.644791379091644e-06,
+      "loss": 0.7002,
+      "step": 18388
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.64378412814015e-06,
+      "loss": 0.7139,
+      "step": 18389
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.642776902498276e-06,
+      "loss": 0.5728,
+      "step": 18390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.641769702176843e-06,
+      "loss": 1.4678,
+      "step": 18391
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.640762527186663e-06,
+      "loss": 0.666,
+      "step": 18392
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639755377538568e-06,
+      "loss": 1.1514,
+      "step": 18393
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.638748253243366e-06,
+      "loss": 1.2236,
+      "step": 18394
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63774115431188e-06,
+      "loss": 1.2529,
+      "step": 18395
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.636734080754924e-06,
+      "loss": 0.957,
+      "step": 18396
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.635727032583317e-06,
+      "loss": 0.835,
+      "step": 18397
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.634720009807879e-06,
+      "loss": 1.0537,
+      "step": 18398
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.633713012439424e-06,
+      "loss": 0.7764,
+      "step": 18399
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.632706040488773e-06,
+      "loss": 1.2334,
+      "step": 18400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.631699093966737e-06,
+      "loss": 1.0303,
+      "step": 18401
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.630692172884137e-06,
+      "loss": 0.8613,
+      "step": 18402
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.629685277251781e-06,
+      "loss": 0.9077,
+      "step": 18403
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.628678407080497e-06,
+      "loss": 0.7363,
+      "step": 18404
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.627671562381092e-06,
+      "loss": 1.2344,
+      "step": 18405
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.626664743164383e-06,
+      "loss": 1.3779,
+      "step": 18406
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.625657949441187e-06,
+      "loss": 0.7517,
+      "step": 18407
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.624651181222312e-06,
+      "loss": 1.0737,
+      "step": 18408
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.62364443851858e-06,
+      "loss": 1.2432,
+      "step": 18409
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.622637721340801e-06,
+      "loss": 0.9951,
+      "step": 18410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.621631029699792e-06,
+      "loss": 0.7466,
+      "step": 18411
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.620624363606363e-06,
+      "loss": 0.5742,
+      "step": 18412
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.619617723071325e-06,
+      "loss": 0.8008,
+      "step": 18413
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.618611108105498e-06,
+      "loss": 1.0342,
+      "step": 18414
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.617604518719692e-06,
+      "loss": 0.5645,
+      "step": 18415
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.616597954924719e-06,
+      "loss": 1.1377,
+      "step": 18416
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.615591416731389e-06,
+      "loss": 0.769,
+      "step": 18417
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.614584904150514e-06,
+      "loss": 0.832,
+      "step": 18418
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.613578417192912e-06,
+      "loss": 1.0176,
+      "step": 18419
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.612571955869387e-06,
+      "loss": 0.8174,
+      "step": 18420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.611565520190751e-06,
+      "loss": 1.3506,
+      "step": 18421
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6105591101678215e-06,
+      "loss": 0.9297,
+      "step": 18422
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.609552725811398e-06,
+      "loss": 0.8237,
+      "step": 18423
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.608546367132299e-06,
+      "loss": 1.1699,
+      "step": 18424
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.607540034141332e-06,
+      "loss": 1.0737,
+      "step": 18425
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.606533726849309e-06,
+      "loss": 1.2539,
+      "step": 18426
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6055274452670345e-06,
+      "loss": 1.2021,
+      "step": 18427
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.604521189405318e-06,
+      "loss": 0.5737,
+      "step": 18428
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.603514959274975e-06,
+      "loss": 0.8794,
+      "step": 18429
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6025087548868094e-06,
+      "loss": 1.2148,
+      "step": 18430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.601502576251628e-06,
+      "loss": 1.1064,
+      "step": 18431
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.600496423380241e-06,
+      "loss": 1.3271,
+      "step": 18432
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.599490296283458e-06,
+      "loss": 0.7236,
+      "step": 18433
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.598484194972076e-06,
+      "loss": 0.9102,
+      "step": 18434
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.597478119456919e-06,
+      "loss": 1.3525,
+      "step": 18435
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.596472069748782e-06,
+      "loss": 1.4277,
+      "step": 18436
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.595466045858477e-06,
+      "loss": 1.4619,
+      "step": 18437
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.594460047796806e-06,
+      "loss": 0.8555,
+      "step": 18438
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.593454075574574e-06,
+      "loss": 0.9863,
+      "step": 18439
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.592448129202594e-06,
+      "loss": 1.0806,
+      "step": 18440
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.591442208691667e-06,
+      "loss": 1.2344,
+      "step": 18441
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5904363140526e-06,
+      "loss": 1.123,
+      "step": 18442
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5894304452961955e-06,
+      "loss": 0.7949,
+      "step": 18443
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5884246024332575e-06,
+      "loss": 0.7397,
+      "step": 18444
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.587418785474594e-06,
+      "loss": 0.8252,
+      "step": 18445
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5864129944310096e-06,
+      "loss": 1.1719,
+      "step": 18446
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.585407229313304e-06,
+      "loss": 0.8281,
+      "step": 18447
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5844014901322835e-06,
+      "loss": 1.3555,
+      "step": 18448
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.583395776898745e-06,
+      "loss": 1.2656,
+      "step": 18449
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.582390089623505e-06,
+      "loss": 1.2363,
+      "step": 18450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.581384428317355e-06,
+      "loss": 1.1719,
+      "step": 18451
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.580378792991101e-06,
+      "loss": 0.8428,
+      "step": 18452
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.579373183655547e-06,
+      "loss": 1.2373,
+      "step": 18453
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.578367600321489e-06,
+      "loss": 1.2148,
+      "step": 18454
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.577362042999734e-06,
+      "loss": 1.0107,
+      "step": 18455
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.576356511701082e-06,
+      "loss": 1.0205,
+      "step": 18456
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.575351006436335e-06,
+      "loss": 0.9102,
+      "step": 18457
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.574345527216293e-06,
+      "loss": 1.0732,
+      "step": 18458
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.573340074051752e-06,
+      "loss": 0.708,
+      "step": 18459
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.572334646953519e-06,
+      "loss": 0.7554,
+      "step": 18460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.571329245932395e-06,
+      "loss": 0.8052,
+      "step": 18461
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5703238709991725e-06,
+      "loss": 0.835,
+      "step": 18462
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.569318522164654e-06,
+      "loss": 1.3057,
+      "step": 18463
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.568313199439641e-06,
+      "loss": 1.2793,
+      "step": 18464
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.567307902834926e-06,
+      "loss": 0.7881,
+      "step": 18465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.566302632361314e-06,
+      "loss": 1.2588,
+      "step": 18466
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.565297388029602e-06,
+      "loss": 0.9033,
+      "step": 18467
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5642921698505875e-06,
+      "loss": 0.8364,
+      "step": 18468
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.563286977835067e-06,
+      "loss": 1.1973,
+      "step": 18469
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.562281811993835e-06,
+      "loss": 1.1104,
+      "step": 18470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.561276672337696e-06,
+      "loss": 1.4355,
+      "step": 18471
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.560271558877445e-06,
+      "loss": 0.7441,
+      "step": 18472
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.559266471623874e-06,
+      "loss": 0.8218,
+      "step": 18473
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.558261410587781e-06,
+      "loss": 0.981,
+      "step": 18474
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.557256375779962e-06,
+      "loss": 1.0996,
+      "step": 18475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.556251367211218e-06,
+      "loss": 1.085,
+      "step": 18476
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.555246384892336e-06,
+      "loss": 1.1797,
+      "step": 18477
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.554241428834118e-06,
+      "loss": 1.0088,
+      "step": 18478
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.553236499047357e-06,
+      "loss": 0.6182,
+      "step": 18479
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.55223159554284e-06,
+      "loss": 1.5957,
+      "step": 18480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.551226718331375e-06,
+      "loss": 1.0508,
+      "step": 18481
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.550221867423746e-06,
+      "loss": 0.8936,
+      "step": 18482
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.549217042830753e-06,
+      "loss": 1.377,
+      "step": 18483
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5482122445631845e-06,
+      "loss": 0.9346,
+      "step": 18484
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.547207472631832e-06,
+      "loss": 1.1045,
+      "step": 18485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.546202727047495e-06,
+      "loss": 0.9854,
+      "step": 18486
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5451980078209615e-06,
+      "loss": 0.7434,
+      "step": 18487
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.544193314963029e-06,
+      "loss": 1.0938,
+      "step": 18488
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.543188648484483e-06,
+      "loss": 1.1777,
+      "step": 18489
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.542184008396116e-06,
+      "loss": 0.834,
+      "step": 18490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.541179394708724e-06,
+      "loss": 0.8213,
+      "step": 18491
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.540174807433098e-06,
+      "loss": 1.1436,
+      "step": 18492
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.539170246580024e-06,
+      "loss": 0.689,
+      "step": 18493
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.538165712160298e-06,
+      "loss": 1.0664,
+      "step": 18494
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.537161204184706e-06,
+      "loss": 1.1973,
+      "step": 18495
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.536156722664036e-06,
+      "loss": 0.9814,
+      "step": 18496
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5351522676090845e-06,
+      "loss": 1.1113,
+      "step": 18497
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.534147839030638e-06,
+      "loss": 0.9639,
+      "step": 18498
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.533143436939487e-06,
+      "loss": 0.9805,
+      "step": 18499
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.532139061346418e-06,
+      "loss": 1.2529,
+      "step": 18500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.531134712262219e-06,
+      "loss": 0.5449,
+      "step": 18501
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.530130389697682e-06,
+      "loss": 1.0767,
+      "step": 18502
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5291260936635946e-06,
+      "loss": 1.4141,
+      "step": 18503
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.528121824170742e-06,
+      "loss": 0.8838,
+      "step": 18504
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.527117581229912e-06,
+      "loss": 0.8945,
+      "step": 18505
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.526113364851891e-06,
+      "loss": 1.4541,
+      "step": 18506
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.525109175047473e-06,
+      "loss": 0.9302,
+      "step": 18507
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.524105011827436e-06,
+      "loss": 1.0059,
+      "step": 18508
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5231008752025695e-06,
+      "loss": 0.9268,
+      "step": 18509
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.522096765183662e-06,
+      "loss": 1.3633,
+      "step": 18510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.521092681781492e-06,
+      "loss": 1.2168,
+      "step": 18511
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.520088625006854e-06,
+      "loss": 0.9521,
+      "step": 18512
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5190845948705275e-06,
+      "loss": 1.0859,
+      "step": 18513
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.518080591383302e-06,
+      "loss": 1.4453,
+      "step": 18514
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.517076614555957e-06,
+      "loss": 1.3184,
+      "step": 18515
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.516072664399278e-06,
+      "loss": 0.9688,
+      "step": 18516
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.515068740924051e-06,
+      "loss": 1.207,
+      "step": 18517
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514064844141061e-06,
+      "loss": 1.3574,
+      "step": 18518
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.513060974061088e-06,
+      "loss": 1.0859,
+      "step": 18519
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.512057130694917e-06,
+      "loss": 0.9893,
+      "step": 18520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.511053314053328e-06,
+      "loss": 1.2129,
+      "step": 18521
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.510049524147111e-06,
+      "loss": 0.9521,
+      "step": 18522
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.509045760987041e-06,
+      "loss": 1.2031,
+      "step": 18523
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.508042024583903e-06,
+      "loss": 1.2812,
+      "step": 18524
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5070383149484805e-06,
+      "loss": 1.0449,
+      "step": 18525
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50603463209155e-06,
+      "loss": 1.6094,
+      "step": 18526
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.505030976023895e-06,
+      "loss": 0.7671,
+      "step": 18527
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.504027346756298e-06,
+      "loss": 0.7994,
+      "step": 18528
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.503023744299541e-06,
+      "loss": 1.001,
+      "step": 18529
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5020201686644e-06,
+      "loss": 0.8433,
+      "step": 18530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.501016619861657e-06,
+      "loss": 1.084,
+      "step": 18531
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50001309790209e-06,
+      "loss": 1.4697,
+      "step": 18532
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.499009602796483e-06,
+      "loss": 1.1191,
+      "step": 18533
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4980061345556134e-06,
+      "loss": 1.3779,
+      "step": 18534
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.497002693190259e-06,
+      "loss": 1.0068,
+      "step": 18535
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.495999278711198e-06,
+      "loss": 0.8818,
+      "step": 18536
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4949958911292045e-06,
+      "loss": 1.1016,
+      "step": 18537
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.493992530455068e-06,
+      "loss": 0.9961,
+      "step": 18538
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.492989196699556e-06,
+      "loss": 1.1328,
+      "step": 18539
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.49198588987345e-06,
+      "loss": 0.7607,
+      "step": 18540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.490982609987529e-06,
+      "loss": 0.8018,
+      "step": 18541
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.489979357052563e-06,
+      "loss": 1.0938,
+      "step": 18542
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488976131079335e-06,
+      "loss": 1.0273,
+      "step": 18543
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.48797293207862e-06,
+      "loss": 1.0615,
+      "step": 18544
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.486969760061196e-06,
+      "loss": 0.7896,
+      "step": 18545
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.485966615037834e-06,
+      "loss": 1.0605,
+      "step": 18546
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4849634970193085e-06,
+      "loss": 0.7871,
+      "step": 18547
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4839604060164e-06,
+      "loss": 0.6592,
+      "step": 18548
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.482957342039882e-06,
+      "loss": 0.9443,
+      "step": 18549
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.481954305100529e-06,
+      "loss": 0.9551,
+      "step": 18550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.480951295209111e-06,
+      "loss": 1.377,
+      "step": 18551
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.479948312376406e-06,
+      "loss": 0.998,
+      "step": 18552
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4789453566131895e-06,
+      "loss": 1.0117,
+      "step": 18553
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.47794242793023e-06,
+      "loss": 1.0234,
+      "step": 18554
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.476939526338305e-06,
+      "loss": 1.1836,
+      "step": 18555
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.475936651848186e-06,
+      "loss": 0.9561,
+      "step": 18556
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4749338044706424e-06,
+      "loss": 1.0244,
+      "step": 18557
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.473930984216447e-06,
+      "loss": 1.1748,
+      "step": 18558
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.472928191096376e-06,
+      "loss": 0.8438,
+      "step": 18559
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.471925425121199e-06,
+      "loss": 1.0361,
+      "step": 18560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.470922686301687e-06,
+      "loss": 1.2344,
+      "step": 18561
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.46991997464861e-06,
+      "loss": 0.7979,
+      "step": 18562
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.468917290172737e-06,
+      "loss": 1.1064,
+      "step": 18563
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.467914632884848e-06,
+      "loss": 0.8999,
+      "step": 18564
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.466912002795701e-06,
+      "loss": 0.9854,
+      "step": 18565
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.465909399916074e-06,
+      "loss": 0.7139,
+      "step": 18566
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4649068242567345e-06,
+      "loss": 0.8887,
+      "step": 18567
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.463904275828444e-06,
+      "loss": 0.8511,
+      "step": 18568
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.462901754641987e-06,
+      "loss": 0.7119,
+      "step": 18569
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.461899260708121e-06,
+      "loss": 0.7954,
+      "step": 18570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.46089679403762e-06,
+      "loss": 0.98,
+      "step": 18571
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.459894354641246e-06,
+      "loss": 0.8682,
+      "step": 18572
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4588919425297705e-06,
+      "loss": 1.1689,
+      "step": 18573
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.457889557713961e-06,
+      "loss": 0.7026,
+      "step": 18574
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.456887200204588e-06,
+      "loss": 1.0908,
+      "step": 18575
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.455884870012412e-06,
+      "loss": 0.9756,
+      "step": 18576
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.454882567148203e-06,
+      "loss": 0.9141,
+      "step": 18577
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.453880291622726e-06,
+      "loss": 0.7617,
+      "step": 18578
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452878043446749e-06,
+      "loss": 0.8979,
+      "step": 18579
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.451875822631041e-06,
+      "loss": 0.7715,
+      "step": 18580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.450873629186359e-06,
+      "loss": 0.874,
+      "step": 18581
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.449871463123477e-06,
+      "loss": 0.4854,
+      "step": 18582
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.448869324453148e-06,
+      "loss": 0.9597,
+      "step": 18583
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.447867213186151e-06,
+      "loss": 0.9971,
+      "step": 18584
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.446865129333242e-06,
+      "loss": 1.1787,
+      "step": 18585
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.445863072905186e-06,
+      "loss": 0.8262,
+      "step": 18586
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.44486104391275e-06,
+      "loss": 0.6543,
+      "step": 18587
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.443859042366693e-06,
+      "loss": 0.751,
+      "step": 18588
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.442857068277776e-06,
+      "loss": 0.9873,
+      "step": 18589
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.441855121656768e-06,
+      "loss": 1.0518,
+      "step": 18590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.440853202514433e-06,
+      "loss": 0.7217,
+      "step": 18591
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439851310861525e-06,
+      "loss": 1.1572,
+      "step": 18592
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.438849446708816e-06,
+      "loss": 0.9282,
+      "step": 18593
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.437847610067053e-06,
+      "loss": 0.9543,
+      "step": 18594
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.436845800947015e-06,
+      "loss": 0.7891,
+      "step": 18595
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.435844019359451e-06,
+      "loss": 0.7461,
+      "step": 18596
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.434842265315125e-06,
+      "loss": 0.7266,
+      "step": 18597
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.433840538824801e-06,
+      "loss": 1.0371,
+      "step": 18598
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4328388398992315e-06,
+      "loss": 0.9404,
+      "step": 18599
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.431837168549183e-06,
+      "loss": 1.1758,
+      "step": 18600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.430835524785414e-06,
+      "loss": 1.0254,
+      "step": 18601
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.429833908618682e-06,
+      "loss": 1.0498,
+      "step": 18602
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.428832320059747e-06,
+      "loss": 1.0186,
+      "step": 18603
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.427830759119365e-06,
+      "loss": 1.2227,
+      "step": 18604
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4268292258083e-06,
+      "loss": 1.124,
+      "step": 18605
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.425827720137308e-06,
+      "loss": 1.332,
+      "step": 18606
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.424826242117144e-06,
+      "loss": 1.0586,
+      "step": 18607
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.423824791758567e-06,
+      "loss": 0.9268,
+      "step": 18608
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.422823369072334e-06,
+      "loss": 1.0693,
+      "step": 18609
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.421821974069205e-06,
+      "loss": 0.8467,
+      "step": 18610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.420820606759933e-06,
+      "loss": 1.5391,
+      "step": 18611
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.419819267155275e-06,
+      "loss": 0.96,
+      "step": 18612
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.418817955265989e-06,
+      "loss": 1.1348,
+      "step": 18613
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.417816671102824e-06,
+      "loss": 1.3301,
+      "step": 18614
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.416815414676546e-06,
+      "loss": 0.6611,
+      "step": 18615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4158141859979025e-06,
+      "loss": 1.2256,
+      "step": 18616
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.414812985077655e-06,
+      "loss": 0.7695,
+      "step": 18617
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.41381181192655e-06,
+      "loss": 0.875,
+      "step": 18618
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.412810666555344e-06,
+      "loss": 0.9473,
+      "step": 18619
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.411809548974792e-06,
+      "loss": 0.8589,
+      "step": 18620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.410808459195652e-06,
+      "loss": 0.9512,
+      "step": 18621
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.409807397228671e-06,
+      "loss": 1.0293,
+      "step": 18622
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.408806363084604e-06,
+      "loss": 0.9531,
+      "step": 18623
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.407805356774207e-06,
+      "loss": 0.9229,
+      "step": 18624
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.406804378308223e-06,
+      "loss": 0.8906,
+      "step": 18625
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.405803427697417e-06,
+      "loss": 1.0547,
+      "step": 18626
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.404802504952533e-06,
+      "loss": 0.8828,
+      "step": 18627
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4038016100843245e-06,
+      "loss": 1.0908,
+      "step": 18628
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.402800743103542e-06,
+      "loss": 0.8818,
+      "step": 18629
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.401799904020934e-06,
+      "loss": 0.8613,
+      "step": 18630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.400799092847255e-06,
+      "loss": 0.9326,
+      "step": 18631
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3997983095932555e-06,
+      "loss": 1.291,
+      "step": 18632
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.398797554269688e-06,
+      "loss": 1.1582,
+      "step": 18633
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.397796826887295e-06,
+      "loss": 0.7183,
+      "step": 18634
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.396796127456827e-06,
+      "loss": 0.8359,
+      "step": 18635
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3957954559890386e-06,
+      "loss": 1.252,
+      "step": 18636
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.394794812494677e-06,
+      "loss": 0.9224,
+      "step": 18637
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3937941969844886e-06,
+      "loss": 1.1436,
+      "step": 18638
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.392793609469225e-06,
+      "loss": 1.2578,
+      "step": 18639
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.391793049959624e-06,
+      "loss": 0.8945,
+      "step": 18640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.390792518466447e-06,
+      "loss": 1.082,
+      "step": 18641
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.389792015000434e-06,
+      "loss": 1.1523,
+      "step": 18642
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3887915395723344e-06,
+      "loss": 1.1064,
+      "step": 18643
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.387791092192895e-06,
+      "loss": 0.894,
+      "step": 18644
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.386790672872857e-06,
+      "loss": 0.9336,
+      "step": 18645
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3857902816229735e-06,
+      "loss": 1.1328,
+      "step": 18646
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.384789918453988e-06,
+      "loss": 1.3066,
+      "step": 18647
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.383789583376646e-06,
+      "loss": 0.939,
+      "step": 18648
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3827892764016905e-06,
+      "loss": 0.8633,
+      "step": 18649
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.381788997539869e-06,
+      "loss": 1.0488,
+      "step": 18650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3807887468019235e-06,
+      "loss": 1.0176,
+      "step": 18651
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.379788524198603e-06,
+      "loss": 0.9478,
+      "step": 18652
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.378788329740648e-06,
+      "loss": 1.0928,
+      "step": 18653
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.377788163438804e-06,
+      "loss": 1.2041,
+      "step": 18654
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.376788025303814e-06,
+      "loss": 0.7734,
+      "step": 18655
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.375787915346415e-06,
+      "loss": 1.0508,
+      "step": 18656
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.374787833577359e-06,
+      "loss": 1.0195,
+      "step": 18657
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.373787780007387e-06,
+      "loss": 1.1279,
+      "step": 18658
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.372787754647238e-06,
+      "loss": 0.9395,
+      "step": 18659
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.371787757507655e-06,
+      "loss": 1.1025,
+      "step": 18660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.370787788599377e-06,
+      "loss": 1.1855,
+      "step": 18661
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.369787847933149e-06,
+      "loss": 1.0117,
+      "step": 18662
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.368787935519714e-06,
+      "loss": 0.9937,
+      "step": 18663
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.367788051369806e-06,
+      "loss": 1.0918,
+      "step": 18664
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.366788195494171e-06,
+      "loss": 1.3115,
+      "step": 18665
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.365788367903543e-06,
+      "loss": 0.9653,
+      "step": 18666
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.364788568608673e-06,
+      "loss": 0.917,
+      "step": 18667
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3637887976202895e-06,
+      "loss": 1.4258,
+      "step": 18668
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3627890549491355e-06,
+      "loss": 1.1523,
+      "step": 18669
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.361789340605952e-06,
+      "loss": 1.0781,
+      "step": 18670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.36078965460147e-06,
+      "loss": 1.1592,
+      "step": 18671
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.359789996946439e-06,
+      "loss": 1.0977,
+      "step": 18672
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3587903676515895e-06,
+      "loss": 0.9951,
+      "step": 18673
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.357790766727663e-06,
+      "loss": 1.751,
+      "step": 18674
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.356791194185392e-06,
+      "loss": 1.0117,
+      "step": 18675
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.355791650035515e-06,
+      "loss": 1.0781,
+      "step": 18676
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.35479213428877e-06,
+      "loss": 0.8955,
+      "step": 18677
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3537926469558974e-06,
+      "loss": 0.8501,
+      "step": 18678
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.352793188047629e-06,
+      "loss": 1.0225,
+      "step": 18679
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.351793757574699e-06,
+      "loss": 0.8096,
+      "step": 18680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.350794355547848e-06,
+      "loss": 0.7891,
+      "step": 18681
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3497949819778e-06,
+      "loss": 1.0508,
+      "step": 18682
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.348795636875306e-06,
+      "loss": 0.7305,
+      "step": 18683
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.347796320251091e-06,
+      "loss": 0.8682,
+      "step": 18684
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3467970321158884e-06,
+      "loss": 0.959,
+      "step": 18685
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.34579777248044e-06,
+      "loss": 1.0527,
+      "step": 18686
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.344798541355469e-06,
+      "loss": 1.1152,
+      "step": 18687
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.343799338751716e-06,
+      "loss": 0.7256,
+      "step": 18688
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.34280016467991e-06,
+      "loss": 1.1943,
+      "step": 18689
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3418010191507915e-06,
+      "loss": 0.8442,
+      "step": 18690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.340801902175083e-06,
+      "loss": 1.2314,
+      "step": 18691
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.33980281376352e-06,
+      "loss": 0.9312,
+      "step": 18692
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.338803753926837e-06,
+      "loss": 1.126,
+      "step": 18693
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3378047226757665e-06,
+      "loss": 1.2188,
+      "step": 18694
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.336805720021034e-06,
+      "loss": 0.7202,
+      "step": 18695
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.335806745973375e-06,
+      "loss": 0.8423,
+      "step": 18696
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.334807800543516e-06,
+      "loss": 1.4033,
+      "step": 18697
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.333808883742192e-06,
+      "loss": 0.9883,
+      "step": 18698
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.332809995580132e-06,
+      "loss": 0.9561,
+      "step": 18699
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.331811136068063e-06,
+      "loss": 0.9209,
+      "step": 18700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.330812305216719e-06,
+      "loss": 1.0381,
+      "step": 18701
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.329813503036822e-06,
+      "loss": 0.7095,
+      "step": 18702
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.328814729539106e-06,
+      "loss": 1.0059,
+      "step": 18703
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3278159847343e-06,
+      "loss": 1.0249,
+      "step": 18704
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.326817268633132e-06,
+      "loss": 0.8706,
+      "step": 18705
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.325818581246325e-06,
+      "loss": 0.9951,
+      "step": 18706
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.324819922584612e-06,
+      "loss": 1.291,
+      "step": 18707
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3238212926587145e-06,
+      "loss": 1.0781,
+      "step": 18708
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.322822691479368e-06,
+      "loss": 1.1753,
+      "step": 18709
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.321824119057292e-06,
+      "loss": 0.9639,
+      "step": 18710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.320825575403214e-06,
+      "loss": 0.876,
+      "step": 18711
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.319827060527864e-06,
+      "loss": 1.1055,
+      "step": 18712
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.318828574441959e-06,
+      "loss": 1.1768,
+      "step": 18713
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.317830117156235e-06,
+      "loss": 1.249,
+      "step": 18714
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3168316886814115e-06,
+      "loss": 1.0693,
+      "step": 18715
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.315833289028216e-06,
+      "loss": 0.9199,
+      "step": 18716
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.314834918207368e-06,
+      "loss": 1.1348,
+      "step": 18717
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3138365762295915e-06,
+      "loss": 0.834,
+      "step": 18718
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.312838263105616e-06,
+      "loss": 1.0991,
+      "step": 18719
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.311839978846165e-06,
+      "loss": 1.0098,
+      "step": 18720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.310841723461957e-06,
+      "loss": 1.1572,
+      "step": 18721
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.309843496963715e-06,
+      "loss": 0.8828,
+      "step": 18722
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.308845299362164e-06,
+      "loss": 0.8691,
+      "step": 18723
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.307847130668026e-06,
+      "loss": 1.0781,
+      "step": 18724
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3068489908920275e-06,
+      "loss": 1.0254,
+      "step": 18725
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.305850880044881e-06,
+      "loss": 1.1357,
+      "step": 18726
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.304852798137315e-06,
+      "loss": 1.2305,
+      "step": 18727
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.303854745180041e-06,
+      "loss": 1.207,
+      "step": 18728
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3028567211837935e-06,
+      "loss": 1.1875,
+      "step": 18729
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.301858726159284e-06,
+      "loss": 1.0371,
+      "step": 18730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.300860760117234e-06,
+      "loss": 1.0488,
+      "step": 18731
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.299862823068366e-06,
+      "loss": 1.1318,
+      "step": 18732
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.298864915023394e-06,
+      "loss": 1.3477,
+      "step": 18733
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.297867035993041e-06,
+      "loss": 1.3115,
+      "step": 18734
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2968691859880266e-06,
+      "loss": 0.7686,
+      "step": 18735
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.295871365019069e-06,
+      "loss": 1.2344,
+      "step": 18736
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.294873573096885e-06,
+      "loss": 1.1982,
+      "step": 18737
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293875810232195e-06,
+      "loss": 1.0781,
+      "step": 18738
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2928780764357076e-06,
+      "loss": 0.8154,
+      "step": 18739
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.291880371718154e-06,
+      "loss": 1.0391,
+      "step": 18740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.290882696090241e-06,
+      "loss": 1.1514,
+      "step": 18741
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2898850495626905e-06,
+      "loss": 1.3281,
+      "step": 18742
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2888874321462165e-06,
+      "loss": 0.8994,
+      "step": 18743
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.287889843851535e-06,
+      "loss": 1.1055,
+      "step": 18744
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.286892284689361e-06,
+      "loss": 1.3145,
+      "step": 18745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.285894754670413e-06,
+      "loss": 1.2705,
+      "step": 18746
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.284897253805405e-06,
+      "loss": 0.9038,
+      "step": 18747
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.283899782105049e-06,
+      "loss": 0.8408,
+      "step": 18748
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.28290233958006e-06,
+      "loss": 0.8994,
+      "step": 18749
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.281904926241157e-06,
+      "loss": 1.4463,
+      "step": 18750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.280907542099052e-06,
+      "loss": 1.7568,
+      "step": 18751
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.279910187164454e-06,
+      "loss": 0.769,
+      "step": 18752
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2789128614480794e-06,
+      "loss": 0.4558,
+      "step": 18753
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27791556496064e-06,
+      "loss": 1.0806,
+      "step": 18754
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.276918297712853e-06,
+      "loss": 0.9814,
+      "step": 18755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.275921059715425e-06,
+      "loss": 1.0996,
+      "step": 18756
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27492385097907e-06,
+      "loss": 1.0488,
+      "step": 18757
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.273926671514503e-06,
+      "loss": 0.8477,
+      "step": 18758
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.272929521332425e-06,
+      "loss": 1.1406,
+      "step": 18759
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27193240044356e-06,
+      "loss": 0.9258,
+      "step": 18760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.270935308858611e-06,
+      "loss": 1.2803,
+      "step": 18761
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.269938246588293e-06,
+      "loss": 1.4238,
+      "step": 18762
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.26894121364331e-06,
+      "loss": 0.8423,
+      "step": 18763
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.267944210034375e-06,
+      "loss": 0.6377,
+      "step": 18764
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.266947235772197e-06,
+      "loss": 0.8887,
+      "step": 18765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.265950290867489e-06,
+      "loss": 1.0391,
+      "step": 18766
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.264953375330955e-06,
+      "loss": 0.5669,
+      "step": 18767
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.263956489173305e-06,
+      "loss": 0.709,
+      "step": 18768
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.262959632405248e-06,
+      "loss": 1.0264,
+      "step": 18769
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.261962805037486e-06,
+      "loss": 1.0312,
+      "step": 18770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.260966007080737e-06,
+      "loss": 1.1309,
+      "step": 18771
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.259969238545702e-06,
+      "loss": 0.9487,
+      "step": 18772
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.258972499443091e-06,
+      "loss": 1.1152,
+      "step": 18773
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2579757897836045e-06,
+      "loss": 0.9331,
+      "step": 18774
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.256979109577953e-06,
+      "loss": 1.3389,
+      "step": 18775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2559824588368425e-06,
+      "loss": 0.8979,
+      "step": 18776
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.254985837570978e-06,
+      "loss": 0.7344,
+      "step": 18777
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.253989245791069e-06,
+      "loss": 0.9316,
+      "step": 18778
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.252992683507814e-06,
+      "loss": 1.3496,
+      "step": 18779
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.251996150731918e-06,
+      "loss": 0.8184,
+      "step": 18780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.250999647474091e-06,
+      "loss": 1.1963,
+      "step": 18781
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2500031737450345e-06,
+      "loss": 0.874,
+      "step": 18782
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.24900672955545e-06,
+      "loss": 1.1309,
+      "step": 18783
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.248010314916043e-06,
+      "loss": 1.1309,
+      "step": 18784
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.247013929837514e-06,
+      "loss": 1.1963,
+      "step": 18785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2460175743305726e-06,
+      "loss": 0.9209,
+      "step": 18786
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.245021248405914e-06,
+      "loss": 1.043,
+      "step": 18787
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.244024952074244e-06,
+      "loss": 0.8203,
+      "step": 18788
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.243028685346265e-06,
+      "loss": 1.0156,
+      "step": 18789
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.242032448232672e-06,
+      "loss": 1.1035,
+      "step": 18790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.241036240744175e-06,
+      "loss": 0.7549,
+      "step": 18791
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.240040062891472e-06,
+      "loss": 1.085,
+      "step": 18792
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.239043914685262e-06,
+      "loss": 1.0811,
+      "step": 18793
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.238047796136247e-06,
+      "loss": 1.3984,
+      "step": 18794
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.237051707255123e-06,
+      "loss": 0.7456,
+      "step": 18795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.236055648052593e-06,
+      "loss": 1.1055,
+      "step": 18796
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2350596185393595e-06,
+      "loss": 1.166,
+      "step": 18797
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.234063618726117e-06,
+      "loss": 0.9053,
+      "step": 18798
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.233067648623564e-06,
+      "loss": 1.2725,
+      "step": 18799
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2320717082424005e-06,
+      "loss": 0.915,
+      "step": 18800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.231075797593322e-06,
+      "loss": 1.0225,
+      "step": 18801
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.230079916687029e-06,
+      "loss": 0.8906,
+      "step": 18802
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2290840655342195e-06,
+      "loss": 1.4492,
+      "step": 18803
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.228088244145591e-06,
+      "loss": 0.8125,
+      "step": 18804
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.227092452531837e-06,
+      "loss": 0.8594,
+      "step": 18805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.226096690703651e-06,
+      "loss": 0.8945,
+      "step": 18806
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.225100958671738e-06,
+      "loss": 0.9609,
+      "step": 18807
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.224105256446789e-06,
+      "loss": 1.2129,
+      "step": 18808
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.223109584039499e-06,
+      "loss": 0.7549,
+      "step": 18809
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.222113941460564e-06,
+      "loss": 1.0164,
+      "step": 18810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.221118328720677e-06,
+      "loss": 0.8999,
+      "step": 18811
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.220122745830536e-06,
+      "loss": 0.688,
+      "step": 18812
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2191271928008345e-06,
+      "loss": 0.9746,
+      "step": 18813
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.218131669642265e-06,
+      "loss": 0.7139,
+      "step": 18814
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2171361763655225e-06,
+      "loss": 1.2363,
+      "step": 18815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2161407129812925e-06,
+      "loss": 0.9463,
+      "step": 18816
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.215145279500282e-06,
+      "loss": 1.0479,
+      "step": 18817
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.214149875933173e-06,
+      "loss": 1.0684,
+      "step": 18818
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.213154502290664e-06,
+      "loss": 0.7998,
+      "step": 18819
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.21215915858344e-06,
+      "loss": 0.8291,
+      "step": 18820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.211163844822196e-06,
+      "loss": 1.0654,
+      "step": 18821
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.210168561017626e-06,
+      "loss": 0.9072,
+      "step": 18822
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.209173307180417e-06,
+      "loss": 1.126,
+      "step": 18823
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2081780833212644e-06,
+      "loss": 0.9731,
+      "step": 18824
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.207182889450853e-06,
+      "loss": 0.9824,
+      "step": 18825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.206187725579874e-06,
+      "loss": 0.8174,
+      "step": 18826
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.20519259171902e-06,
+      "loss": 0.9082,
+      "step": 18827
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.204197487878981e-06,
+      "loss": 0.8252,
+      "step": 18828
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.203202414070442e-06,
+      "loss": 1.207,
+      "step": 18829
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.202207370304093e-06,
+      "loss": 1.0371,
+      "step": 18830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.201212356590625e-06,
+      "loss": 1.1562,
+      "step": 18831
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.200217372940721e-06,
+      "loss": 0.8818,
+      "step": 18832
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.199222419365073e-06,
+      "loss": 0.8262,
+      "step": 18833
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.198227495874368e-06,
+      "loss": 0.7227,
+      "step": 18834
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.197232602479293e-06,
+      "loss": 1.2549,
+      "step": 18835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.196237739190533e-06,
+      "loss": 0.8945,
+      "step": 18836
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.195242906018773e-06,
+      "loss": 1.292,
+      "step": 18837
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.194248102974704e-06,
+      "loss": 0.876,
+      "step": 18838
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.193253330069012e-06,
+      "loss": 0.8608,
+      "step": 18839
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.192258587312378e-06,
+      "loss": 0.9624,
+      "step": 18840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191263874715488e-06,
+      "loss": 0.8848,
+      "step": 18841
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.190269192289028e-06,
+      "loss": 0.9912,
+      "step": 18842
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.189274540043684e-06,
+      "loss": 1.0596,
+      "step": 18843
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.188279917990139e-06,
+      "loss": 0.8916,
+      "step": 18844
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.187285326139077e-06,
+      "loss": 1.2656,
+      "step": 18845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.186290764501183e-06,
+      "loss": 0.9375,
+      "step": 18846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.185296233087132e-06,
+      "loss": 0.9043,
+      "step": 18847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1843017319076166e-06,
+      "loss": 0.9453,
+      "step": 18848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.183307260973317e-06,
+      "loss": 0.967,
+      "step": 18849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.182312820294916e-06,
+      "loss": 0.8428,
+      "step": 18850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1813184098830915e-06,
+      "loss": 0.7319,
+      "step": 18851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.180324029748526e-06,
+      "loss": 0.7178,
+      "step": 18852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.179329679901904e-06,
+      "loss": 0.6086,
+      "step": 18853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.178335360353907e-06,
+      "loss": 0.8887,
+      "step": 18854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.177341071115211e-06,
+      "loss": 1.0361,
+      "step": 18855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1763468121965e-06,
+      "loss": 0.7686,
+      "step": 18856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1753525836084504e-06,
+      "loss": 0.9385,
+      "step": 18857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.174358385361746e-06,
+      "loss": 0.752,
+      "step": 18858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1733642174670674e-06,
+      "loss": 0.7939,
+      "step": 18859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.172370079935087e-06,
+      "loss": 0.9365,
+      "step": 18860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.171375972776489e-06,
+      "loss": 1.0557,
+      "step": 18861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.170381896001948e-06,
+      "loss": 0.8379,
+      "step": 18862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.169387849622141e-06,
+      "loss": 1.042,
+      "step": 18863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.168393833647752e-06,
+      "loss": 0.7891,
+      "step": 18864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.167399848089456e-06,
+      "loss": 1.1475,
+      "step": 18865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.166405892957926e-06,
+      "loss": 1.3184,
+      "step": 18866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1654119682638425e-06,
+      "loss": 0.7627,
+      "step": 18867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.164418074017879e-06,
+      "loss": 1.0381,
+      "step": 18868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.163424210230716e-06,
+      "loss": 0.9453,
+      "step": 18869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.162430376913027e-06,
+      "loss": 1.457,
+      "step": 18870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.161436574075486e-06,
+      "loss": 1.2217,
+      "step": 18871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.160442801728771e-06,
+      "loss": 0.7495,
+      "step": 18872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.159449059883548e-06,
+      "loss": 0.8628,
+      "step": 18873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.158455348550506e-06,
+      "loss": 1.0791,
+      "step": 18874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.157461667740309e-06,
+      "loss": 1.0029,
+      "step": 18875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.156468017463632e-06,
+      "loss": 1.0615,
+      "step": 18876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.155474397731153e-06,
+      "loss": 1.1279,
+      "step": 18877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1544808085535366e-06,
+      "loss": 0.8789,
+      "step": 18878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.153487249941464e-06,
+      "loss": 1.1353,
+      "step": 18879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.152493721905602e-06,
+      "loss": 1.0029,
+      "step": 18880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.151500224456629e-06,
+      "loss": 0.9854,
+      "step": 18881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.150506757605211e-06,
+      "loss": 0.8262,
+      "step": 18882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.149513321362018e-06,
+      "loss": 0.666,
+      "step": 18883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.148519915737727e-06,
+      "loss": 0.8701,
+      "step": 18884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.147526540743009e-06,
+      "loss": 0.9512,
+      "step": 18885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.146533196388529e-06,
+      "loss": 0.9414,
+      "step": 18886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14553988268496e-06,
+      "loss": 0.8877,
+      "step": 18887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.144546599642968e-06,
+      "loss": 0.6387,
+      "step": 18888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.143553347273233e-06,
+      "loss": 1.4766,
+      "step": 18889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.142560125586413e-06,
+      "loss": 1.2891,
+      "step": 18890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.141566934593182e-06,
+      "loss": 0.918,
+      "step": 18891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14057377430421e-06,
+      "loss": 0.7197,
+      "step": 18892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.139580644730161e-06,
+      "loss": 1.0742,
+      "step": 18893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.138587545881701e-06,
+      "loss": 0.8325,
+      "step": 18894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.137594477769505e-06,
+      "loss": 1.1797,
+      "step": 18895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1366014404042386e-06,
+      "loss": 1.1855,
+      "step": 18896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.135608433796563e-06,
+      "loss": 1.041,
+      "step": 18897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.134615457957148e-06,
+      "loss": 1.166,
+      "step": 18898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.133622512896658e-06,
+      "loss": 1.4072,
+      "step": 18899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.132629598625767e-06,
+      "loss": 0.9639,
+      "step": 18900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.131636715155131e-06,
+      "loss": 0.9473,
+      "step": 18901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.130643862495417e-06,
+      "loss": 1.0537,
+      "step": 18902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.129651040657294e-06,
+      "loss": 0.874,
+      "step": 18903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.128658249651418e-06,
+      "loss": 1.0513,
+      "step": 18904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1276654894884644e-06,
+      "loss": 0.9971,
+      "step": 18905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.126672760179092e-06,
+      "loss": 0.9072,
+      "step": 18906
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.125680061733964e-06,
+      "loss": 0.9268,
+      "step": 18907
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.124687394163743e-06,
+      "loss": 0.668,
+      "step": 18908
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.123694757479089e-06,
+      "loss": 1.2559,
+      "step": 18909
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.122702151690671e-06,
+      "loss": 0.9673,
+      "step": 18910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1217095768091485e-06,
+      "loss": 1.0713,
+      "step": 18911
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.120717032845186e-06,
+      "loss": 0.9277,
+      "step": 18912
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1197245198094386e-06,
+      "loss": 0.8789,
+      "step": 18913
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11873203771257e-06,
+      "loss": 0.7778,
+      "step": 18914
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1177395865652444e-06,
+      "loss": 0.7607,
+      "step": 18915
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.116747166378122e-06,
+      "loss": 0.7144,
+      "step": 18916
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11575477716186e-06,
+      "loss": 0.8911,
+      "step": 18917
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1147624189271215e-06,
+      "loss": 0.9912,
+      "step": 18918
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.113770091684556e-06,
+      "loss": 0.9072,
+      "step": 18919
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.112777795444839e-06,
+      "loss": 0.9863,
+      "step": 18920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1117855302186176e-06,
+      "loss": 0.792,
+      "step": 18921
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.110793296016556e-06,
+      "loss": 1.0562,
+      "step": 18922
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.109801092849311e-06,
+      "loss": 1.041,
+      "step": 18923
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.108808920727537e-06,
+      "loss": 1.0039,
+      "step": 18924
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1078167796618935e-06,
+      "loss": 0.835,
+      "step": 18925
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.10682466966304e-06,
+      "loss": 0.6553,
+      "step": 18926
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.105832590741635e-06,
+      "loss": 1.0967,
+      "step": 18927
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.10484054290833e-06,
+      "loss": 0.9121,
+      "step": 18928
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.103848526173782e-06,
+      "loss": 0.9624,
+      "step": 18929
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.102856540548648e-06,
+      "loss": 0.9375,
+      "step": 18930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101864586043586e-06,
+      "loss": 0.9609,
+      "step": 18931
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.100872662669248e-06,
+      "loss": 0.9961,
+      "step": 18932
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.09988077043629e-06,
+      "loss": 1.1611,
+      "step": 18933
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.098888909355368e-06,
+      "loss": 0.8926,
+      "step": 18934
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.09789707943713e-06,
+      "loss": 1.1602,
+      "step": 18935
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096905280692237e-06,
+      "loss": 0.8682,
+      "step": 18936
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.09591351313134e-06,
+      "loss": 0.833,
+      "step": 18937
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.094921776765095e-06,
+      "loss": 1.1006,
+      "step": 18938
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.093930071604149e-06,
+      "loss": 0.8223,
+      "step": 18939
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.092938397659156e-06,
+      "loss": 0.7227,
+      "step": 18940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.091946754940771e-06,
+      "loss": 0.7803,
+      "step": 18941
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.090955143459648e-06,
+      "loss": 1.0908,
+      "step": 18942
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.089963563226433e-06,
+      "loss": 1.0293,
+      "step": 18943
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.088972014251779e-06,
+      "loss": 1.0869,
+      "step": 18944
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.087980496546336e-06,
+      "loss": 1.124,
+      "step": 18945
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.08698901012076e-06,
+      "loss": 1.2363,
+      "step": 18946
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.085997554985693e-06,
+      "loss": 0.9775,
+      "step": 18947
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.08500613115179e-06,
+      "loss": 0.9019,
+      "step": 18948
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.084014738629702e-06,
+      "loss": 1.4365,
+      "step": 18949
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.083023377430069e-06,
+      "loss": 1.085,
+      "step": 18950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.082032047563553e-06,
+      "loss": 1.3496,
+      "step": 18951
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.081040749040793e-06,
+      "loss": 0.9951,
+      "step": 18952
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.080049481872443e-06,
+      "loss": 1.6641,
+      "step": 18953
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.079058246069146e-06,
+      "loss": 0.7554,
+      "step": 18954
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.07806704164155e-06,
+      "loss": 1.1572,
+      "step": 18955
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.077075868600303e-06,
+      "loss": 0.9961,
+      "step": 18956
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.076084726956054e-06,
+      "loss": 0.9092,
+      "step": 18957
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.075093616719449e-06,
+      "loss": 0.7861,
+      "step": 18958
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.074102537901133e-06,
+      "loss": 0.7412,
+      "step": 18959
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.073111490511753e-06,
+      "loss": 0.7954,
+      "step": 18960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.072120474561945e-06,
+      "loss": 1.0239,
+      "step": 18961
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.071129490062372e-06,
+      "loss": 0.7705,
+      "step": 18962
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.070138537023665e-06,
+      "loss": 1.0425,
+      "step": 18963
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.069147615456476e-06,
+      "loss": 1.0664,
+      "step": 18964
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0681567253714426e-06,
+      "loss": 0.9268,
+      "step": 18965
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0671658667792105e-06,
+      "loss": 0.9043,
+      "step": 18966
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.066175039690426e-06,
+      "loss": 0.6992,
+      "step": 18967
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0651842441157315e-06,
+      "loss": 0.8682,
+      "step": 18968
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0641934800657705e-06,
+      "loss": 0.8628,
+      "step": 18969
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0632027475511824e-06,
+      "loss": 1.1562,
+      "step": 18970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0622120465826085e-06,
+      "loss": 0.9707,
+      "step": 18971
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.061221377170694e-06,
+      "loss": 1.2783,
+      "step": 18972
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.060230739326082e-06,
+      "loss": 0.9521,
+      "step": 18973
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.059240133059408e-06,
+      "loss": 1.0605,
+      "step": 18974
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.058249558381316e-06,
+      "loss": 1.1445,
+      "step": 18975
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.057259015302444e-06,
+      "loss": 1.127,
+      "step": 18976
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0562685038334375e-06,
+      "loss": 1.1416,
+      "step": 18977
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.05527802398493e-06,
+      "loss": 0.9688,
+      "step": 18978
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.054287575767564e-06,
+      "loss": 1.1748,
+      "step": 18979
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.053297159191981e-06,
+      "loss": 0.9609,
+      "step": 18980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.052306774268812e-06,
+      "loss": 1.1104,
+      "step": 18981
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.051316421008702e-06,
+      "loss": 0.8408,
+      "step": 18982
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.050326099422287e-06,
+      "loss": 0.9612,
+      "step": 18983
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.049335809520208e-06,
+      "loss": 1.2012,
+      "step": 18984
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.048345551313096e-06,
+      "loss": 1.0693,
+      "step": 18985
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.047355324811591e-06,
+      "loss": 0.7285,
+      "step": 18986
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.046365130026327e-06,
+      "loss": 1.125,
+      "step": 18987
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.045374966967948e-06,
+      "loss": 1.0088,
+      "step": 18988
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.044384835647083e-06,
+      "loss": 0.7969,
+      "step": 18989
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.04339473607437e-06,
+      "loss": 1.0234,
+      "step": 18990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.042404668260447e-06,
+      "loss": 0.5828,
+      "step": 18991
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.041414632215939e-06,
+      "loss": 1.0,
+      "step": 18992
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0404246279514925e-06,
+      "loss": 1.4014,
+      "step": 18993
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.039434655477735e-06,
+      "loss": 0.999,
+      "step": 18994
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0384447148053035e-06,
+      "loss": 1.2891,
+      "step": 18995
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.03745480594483e-06,
+      "loss": 1.0547,
+      "step": 18996
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.036464928906945e-06,
+      "loss": 0.8198,
+      "step": 18997
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.035475083702288e-06,
+      "loss": 0.8105,
+      "step": 18998
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.034485270341489e-06,
+      "loss": 0.813,
+      "step": 18999
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.033495488835178e-06,
+      "loss": 1.0015,
+      "step": 19000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.032505739193986e-06,
+      "loss": 0.8477,
+      "step": 19001
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.031516021428546e-06,
+      "loss": 1.022,
+      "step": 19002
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0305263355494925e-06,
+      "loss": 0.8306,
+      "step": 19003
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.029536681567455e-06,
+      "loss": 1.0742,
+      "step": 19004
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.028547059493061e-06,
+      "loss": 1.1689,
+      "step": 19005
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0275574693369455e-06,
+      "loss": 0.9814,
+      "step": 19006
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.026567911109727e-06,
+      "loss": 0.9775,
+      "step": 19007
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.025578384822049e-06,
+      "loss": 1.3311,
+      "step": 19008
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.024588890484534e-06,
+      "loss": 1.3555,
+      "step": 19009
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.023599428107815e-06,
+      "loss": 0.8086,
+      "step": 19010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.022609997702512e-06,
+      "loss": 1.0586,
+      "step": 19011
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.021620599279257e-06,
+      "loss": 1.2812,
+      "step": 19012
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.020631232848683e-06,
+      "loss": 0.667,
+      "step": 19013
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.019641898421412e-06,
+      "loss": 1.125,
+      "step": 19014
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.018652596008073e-06,
+      "loss": 0.7676,
+      "step": 19015
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.017663325619292e-06,
+      "loss": 0.978,
+      "step": 19016
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.016674087265697e-06,
+      "loss": 1.5918,
+      "step": 19017
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.015684880957906e-06,
+      "loss": 1.0264,
+      "step": 19018
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.014695706706558e-06,
+      "loss": 0.8506,
+      "step": 19019
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0137065645222695e-06,
+      "loss": 0.71,
+      "step": 19020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.012717454415668e-06,
+      "loss": 1.1133,
+      "step": 19021
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.011728376397378e-06,
+      "loss": 0.8643,
+      "step": 19022
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.01073933047802e-06,
+      "loss": 1.0049,
+      "step": 19023
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.009750316668224e-06,
+      "loss": 0.7793,
+      "step": 19024
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.008761334978611e-06,
+      "loss": 1.082,
+      "step": 19025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0077723854198075e-06,
+      "loss": 0.7939,
+      "step": 19026
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.006783468002432e-06,
+      "loss": 1.04,
+      "step": 19027
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.005794582737104e-06,
+      "loss": 0.5986,
+      "step": 19028
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.004805729634454e-06,
+      "loss": 0.8643,
+      "step": 19029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.003816908705101e-06,
+      "loss": 1.1738,
+      "step": 19030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.002828119959665e-06,
+      "loss": 0.9717,
+      "step": 19031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0018393634087664e-06,
+      "loss": 1.0903,
+      "step": 19032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.000850639063028e-06,
+      "loss": 0.7539,
+      "step": 19033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.999861946933072e-06,
+      "loss": 1.1973,
+      "step": 19034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.998873287029515e-06,
+      "loss": 0.8232,
+      "step": 19035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.997884659362979e-06,
+      "loss": 0.8872,
+      "step": 19036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.996896063944083e-06,
+      "loss": 0.9072,
+      "step": 19037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9959075007834435e-06,
+      "loss": 1.1328,
+      "step": 19038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.994918969891683e-06,
+      "loss": 1.0273,
+      "step": 19039
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9939304712794186e-06,
+      "loss": 0.6914,
+      "step": 19040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.992942004957271e-06,
+      "loss": 1.1279,
+      "step": 19041
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.991953570935853e-06,
+      "loss": 1.2852,
+      "step": 19042
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.990965169225785e-06,
+      "loss": 0.7554,
+      "step": 19043
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.989976799837681e-06,
+      "loss": 1.1953,
+      "step": 19044
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.988988462782165e-06,
+      "loss": 1.1035,
+      "step": 19045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.988000158069845e-06,
+      "loss": 1.0312,
+      "step": 19046
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.987011885711341e-06,
+      "loss": 1.0215,
+      "step": 19047
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.986023645717271e-06,
+      "loss": 0.9565,
+      "step": 19048
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.985035438098241e-06,
+      "loss": 0.9014,
+      "step": 19049
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.984047262864879e-06,
+      "loss": 1.5059,
+      "step": 19050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.983059120027791e-06,
+      "loss": 1.041,
+      "step": 19051
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.982071009597595e-06,
+      "loss": 1.1055,
+      "step": 19052
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.981082931584902e-06,
+      "loss": 1.0254,
+      "step": 19053
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.980094886000323e-06,
+      "loss": 0.918,
+      "step": 19054
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9791068728544785e-06,
+      "loss": 1.1733,
+      "step": 19055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.978118892157978e-06,
+      "loss": 1.1211,
+      "step": 19056
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.977130943921435e-06,
+      "loss": 1.27,
+      "step": 19057
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.97614302815546e-06,
+      "loss": 1.25,
+      "step": 19058
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.975155144870665e-06,
+      "loss": 1.252,
+      "step": 19059
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9741672940776604e-06,
+      "loss": 1.3926,
+      "step": 19060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9731794757870645e-06,
+      "loss": 0.9453,
+      "step": 19061
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9721916900094775e-06,
+      "loss": 1.1162,
+      "step": 19062
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.97120393675552e-06,
+      "loss": 1.2383,
+      "step": 19063
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.970216216035788e-06,
+      "loss": 0.8516,
+      "step": 19064
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.969228527860907e-06,
+      "loss": 0.9033,
+      "step": 19065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.968240872241479e-06,
+      "loss": 0.8555,
+      "step": 19066
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.967253249188113e-06,
+      "loss": 1.0742,
+      "step": 19067
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.96626565871142e-06,
+      "loss": 1.0283,
+      "step": 19068
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.965278100822004e-06,
+      "loss": 0.8682,
+      "step": 19069
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.964290575530475e-06,
+      "loss": 0.9707,
+      "step": 19070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.963303082847444e-06,
+      "loss": 1.2031,
+      "step": 19071
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.962315622783517e-06,
+      "loss": 1.1055,
+      "step": 19072
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.961328195349297e-06,
+      "loss": 0.9404,
+      "step": 19073
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.960340800555393e-06,
+      "loss": 0.9658,
+      "step": 19074
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.95935343841241e-06,
+      "loss": 1.1494,
+      "step": 19075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.958366108930958e-06,
+      "loss": 0.8193,
+      "step": 19076
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.957378812121639e-06,
+      "loss": 1.2686,
+      "step": 19077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.956391547995059e-06,
+      "loss": 1.166,
+      "step": 19078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.955404316561825e-06,
+      "loss": 0.8389,
+      "step": 19079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.954417117832535e-06,
+      "loss": 1.1885,
+      "step": 19080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9534299518177986e-06,
+      "loss": 1.0947,
+      "step": 19081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.95244281852822e-06,
+      "loss": 0.8389,
+      "step": 19082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.951455717974401e-06,
+      "loss": 0.8359,
+      "step": 19083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.950468650166944e-06,
+      "loss": 0.7251,
+      "step": 19084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.949481615116451e-06,
+      "loss": 0.9658,
+      "step": 19085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.948494612833527e-06,
+      "loss": 1.2964,
+      "step": 19086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.947507643328775e-06,
+      "loss": 0.8232,
+      "step": 19087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9465207066127934e-06,
+      "loss": 0.9268,
+      "step": 19088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.945533802696183e-06,
+      "loss": 0.8936,
+      "step": 19089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.944546931589546e-06,
+      "loss": 0.8569,
+      "step": 19090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.943560093303487e-06,
+      "loss": 1.0234,
+      "step": 19091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9425732878485995e-06,
+      "loss": 0.7061,
+      "step": 19092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.941586515235488e-06,
+      "loss": 0.75,
+      "step": 19093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9405997754747525e-06,
+      "loss": 1.1006,
+      "step": 19094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.939613068576984e-06,
+      "loss": 0.7861,
+      "step": 19095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.938626394552795e-06,
+      "loss": 1.2041,
+      "step": 19096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.937639753412773e-06,
+      "loss": 1.0376,
+      "step": 19097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.936653145167521e-06,
+      "loss": 1.0234,
+      "step": 19098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.935666569827635e-06,
+      "loss": 0.7393,
+      "step": 19099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.934680027403711e-06,
+      "loss": 1.1094,
+      "step": 19100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.93369351790635e-06,
+      "loss": 1.0664,
+      "step": 19101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.932707041346147e-06,
+      "loss": 1.0225,
+      "step": 19102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.931720597733701e-06,
+      "loss": 1.0215,
+      "step": 19103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.930734187079602e-06,
+      "loss": 1.0586,
+      "step": 19104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.929747809394451e-06,
+      "loss": 1.0234,
+      "step": 19105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.928761464688836e-06,
+      "loss": 0.7886,
+      "step": 19106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.927775152973364e-06,
+      "loss": 1.334,
+      "step": 19107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.926788874258619e-06,
+      "loss": 1.291,
+      "step": 19108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.925802628555203e-06,
+      "loss": 1.0186,
+      "step": 19109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.924816415873703e-06,
+      "loss": 1.1797,
+      "step": 19110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9238302362247124e-06,
+      "loss": 1.0449,
+      "step": 19111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92284408961883e-06,
+      "loss": 1.4023,
+      "step": 19112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.921857976066647e-06,
+      "loss": 1.3643,
+      "step": 19113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.920871895578757e-06,
+      "loss": 1.166,
+      "step": 19114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.919885848165748e-06,
+      "loss": 0.647,
+      "step": 19115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.918899833838211e-06,
+      "loss": 0.9182,
+      "step": 19116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917913852606743e-06,
+      "loss": 1.2285,
+      "step": 19117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.916927904481934e-06,
+      "loss": 1.0483,
+      "step": 19118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9159419894743705e-06,
+      "loss": 0.71,
+      "step": 19119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.914956107594647e-06,
+      "loss": 1.1113,
+      "step": 19120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.913970258853349e-06,
+      "loss": 1.1699,
+      "step": 19121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.912984443261073e-06,
+      "loss": 1.1699,
+      "step": 19122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.911998660828403e-06,
+      "loss": 1.0518,
+      "step": 19123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9110129115659285e-06,
+      "loss": 0.9365,
+      "step": 19124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9100271954842415e-06,
+      "loss": 1.0391,
+      "step": 19125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.909041512593922e-06,
+      "loss": 1.041,
+      "step": 19126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9080558629055665e-06,
+      "loss": 0.9805,
+      "step": 19127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.907070246429759e-06,
+      "loss": 1.0977,
+      "step": 19128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.906084663177088e-06,
+      "loss": 1.1797,
+      "step": 19129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.905099113158138e-06,
+      "loss": 0.8999,
+      "step": 19130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9041135963834945e-06,
+      "loss": 1.0518,
+      "step": 19131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.903128112863749e-06,
+      "loss": 0.999,
+      "step": 19132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9021426626094854e-06,
+      "loss": 1.0723,
+      "step": 19133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.901157245631285e-06,
+      "loss": 0.9072,
+      "step": 19134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.900171861939735e-06,
+      "loss": 1.0928,
+      "step": 19135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.899186511545424e-06,
+      "loss": 1.1182,
+      "step": 19136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.898201194458927e-06,
+      "loss": 1.1396,
+      "step": 19137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.897215910690837e-06,
+      "loss": 1.2227,
+      "step": 19138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.896230660251732e-06,
+      "loss": 0.7974,
+      "step": 19139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.895245443152202e-06,
+      "loss": 1.3428,
+      "step": 19140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.894260259402822e-06,
+      "loss": 1.1699,
+      "step": 19141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.893275109014175e-06,
+      "loss": 1.1367,
+      "step": 19142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89228999199685e-06,
+      "loss": 1.4688,
+      "step": 19143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.891304908361425e-06,
+      "loss": 0.9463,
+      "step": 19144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.890319858118479e-06,
+      "loss": 0.7153,
+      "step": 19145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.889334841278596e-06,
+      "loss": 0.8535,
+      "step": 19146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.888349857852352e-06,
+      "loss": 1.4297,
+      "step": 19147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.887364907850335e-06,
+      "loss": 1.2705,
+      "step": 19148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.886379991283122e-06,
+      "loss": 1.0811,
+      "step": 19149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88539510816129e-06,
+      "loss": 1.2188,
+      "step": 19150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.884410258495422e-06,
+      "loss": 0.8486,
+      "step": 19151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.883425442296089e-06,
+      "loss": 1.0,
+      "step": 19152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8824406595738815e-06,
+      "loss": 1.3398,
+      "step": 19153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.881455910339369e-06,
+      "loss": 0.916,
+      "step": 19154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.880471194603133e-06,
+      "loss": 1.3301,
+      "step": 19155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.879486512375751e-06,
+      "loss": 1.04,
+      "step": 19156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.878501863667795e-06,
+      "loss": 1.0986,
+      "step": 19157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.877517248489848e-06,
+      "loss": 1.2666,
+      "step": 19158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.876532666852484e-06,
+      "loss": 0.7529,
+      "step": 19159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.875548118766281e-06,
+      "loss": 1.084,
+      "step": 19160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.874563604241811e-06,
+      "loss": 0.7114,
+      "step": 19161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.873579123289648e-06,
+      "loss": 1.0977,
+      "step": 19162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8725946759203745e-06,
+      "loss": 1.2705,
+      "step": 19163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.871610262144561e-06,
+      "loss": 1.1523,
+      "step": 19164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.870625881972779e-06,
+      "loss": 1.0117,
+      "step": 19165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.869641535415605e-06,
+      "loss": 1.0498,
+      "step": 19166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.868657222483614e-06,
+      "loss": 1.2783,
+      "step": 19167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.867672943187373e-06,
+      "loss": 1.4453,
+      "step": 19168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8666886975374625e-06,
+      "loss": 0.7852,
+      "step": 19169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.865704485544451e-06,
+      "loss": 0.7021,
+      "step": 19170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.864720307218913e-06,
+      "loss": 1.0859,
+      "step": 19171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.863736162571415e-06,
+      "loss": 1.1055,
+      "step": 19172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8627520516125314e-06,
+      "loss": 0.9834,
+      "step": 19173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8617679743528334e-06,
+      "loss": 1.0586,
+      "step": 19174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8607839308028944e-06,
+      "loss": 0.8711,
+      "step": 19175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.85979992097328e-06,
+      "loss": 0.9473,
+      "step": 19176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.858815944874562e-06,
+      "loss": 1.1436,
+      "step": 19177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.857832002517307e-06,
+      "loss": 0.9482,
+      "step": 19178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.856848093912092e-06,
+      "loss": 1.3184,
+      "step": 19179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.855864219069477e-06,
+      "loss": 0.9985,
+      "step": 19180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.854880378000035e-06,
+      "loss": 1.4365,
+      "step": 19181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.853896570714336e-06,
+      "loss": 1.3076,
+      "step": 19182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.852912797222938e-06,
+      "loss": 0.8281,
+      "step": 19183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.851929057536422e-06,
+      "loss": 1.1641,
+      "step": 19184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.850945351665346e-06,
+      "loss": 1.0879,
+      "step": 19185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849961679620281e-06,
+      "loss": 1.1885,
+      "step": 19186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.848978041411788e-06,
+      "loss": 1.0908,
+      "step": 19187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8479944370504336e-06,
+      "loss": 0.9639,
+      "step": 19188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8470108665467895e-06,
+      "loss": 0.833,
+      "step": 19189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.846027329911417e-06,
+      "loss": 1.0234,
+      "step": 19190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.845043827154881e-06,
+      "loss": 1.1211,
+      "step": 19191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.844060358287745e-06,
+      "loss": 0.9136,
+      "step": 19192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8430769233205715e-06,
+      "loss": 1.1626,
+      "step": 19193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.84209352226393e-06,
+      "loss": 0.9402,
+      "step": 19194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.841110155128381e-06,
+      "loss": 1.0508,
+      "step": 19195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.840126821924486e-06,
+      "loss": 0.8867,
+      "step": 19196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83914352266281e-06,
+      "loss": 1.0635,
+      "step": 19197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.838160257353912e-06,
+      "loss": 0.8994,
+      "step": 19198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.837177026008353e-06,
+      "loss": 1.1172,
+      "step": 19199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8361938286367e-06,
+      "loss": 0.8086,
+      "step": 19200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8352106652495124e-06,
+      "loss": 0.8735,
+      "step": 19201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.834227535857349e-06,
+      "loss": 0.5801,
+      "step": 19202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.833244440470772e-06,
+      "loss": 0.5986,
+      "step": 19203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.832261379100338e-06,
+      "loss": 1.1348,
+      "step": 19204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8312783517566105e-06,
+      "loss": 0.812,
+      "step": 19205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83029535845015e-06,
+      "loss": 0.9365,
+      "step": 19206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.829312399191511e-06,
+      "loss": 0.6851,
+      "step": 19207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.828329473991256e-06,
+      "loss": 0.8115,
+      "step": 19208
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.827346582859936e-06,
+      "loss": 0.8994,
+      "step": 19209
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.82636372580812e-06,
+      "loss": 0.7773,
+      "step": 19210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.825380902846358e-06,
+      "loss": 1.4473,
+      "step": 19211
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.824398113985209e-06,
+      "loss": 0.9102,
+      "step": 19212
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.823415359235231e-06,
+      "loss": 1.2891,
+      "step": 19213
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.822432638606975e-06,
+      "loss": 0.9219,
+      "step": 19214
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.821449952111004e-06,
+      "loss": 0.9082,
+      "step": 19215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.820467299757871e-06,
+      "loss": 1.1396,
+      "step": 19216
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.819484681558133e-06,
+      "loss": 1.2671,
+      "step": 19217
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.818502097522339e-06,
+      "loss": 0.9126,
+      "step": 19218
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.817519547661047e-06,
+      "loss": 1.5742,
+      "step": 19219
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.816537031984813e-06,
+      "loss": 0.918,
+      "step": 19220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.815554550504193e-06,
+      "loss": 1.1221,
+      "step": 19221
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.814572103229735e-06,
+      "loss": 1.3164,
+      "step": 19222
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.813589690171993e-06,
+      "loss": 0.9209,
+      "step": 19223
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812607311341519e-06,
+      "loss": 0.7891,
+      "step": 19224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.811624966748872e-06,
+      "loss": 0.8257,
+      "step": 19225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8106426564045965e-06,
+      "loss": 1.1289,
+      "step": 19226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.809660380319248e-06,
+      "loss": 1.2783,
+      "step": 19227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.808678138503377e-06,
+      "loss": 0.8232,
+      "step": 19228
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.807695930967533e-06,
+      "loss": 0.9561,
+      "step": 19229
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.806713757722265e-06,
+      "loss": 0.8975,
+      "step": 19230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8057316187781286e-06,
+      "loss": 0.7451,
+      "step": 19231
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.804749514145672e-06,
+      "loss": 1.3096,
+      "step": 19232
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.803767443835441e-06,
+      "loss": 1.1094,
+      "step": 19233
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.802785407857987e-06,
+      "loss": 0.7339,
+      "step": 19234
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.801803406223857e-06,
+      "loss": 1.1533,
+      "step": 19235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.800821438943603e-06,
+      "loss": 0.7832,
+      "step": 19236
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.799839506027771e-06,
+      "loss": 1.1416,
+      "step": 19237
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.798857607486909e-06,
+      "loss": 0.877,
+      "step": 19238
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.797875743331564e-06,
+      "loss": 1.4102,
+      "step": 19239
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796893913572275e-06,
+      "loss": 1.0977,
+      "step": 19240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.795912118219604e-06,
+      "loss": 0.8125,
+      "step": 19241
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.794930357284086e-06,
+      "loss": 1.0049,
+      "step": 19242
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7939486307762725e-06,
+      "loss": 0.8857,
+      "step": 19243
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.792966938706704e-06,
+      "loss": 0.7949,
+      "step": 19244
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7919852810859245e-06,
+      "loss": 1.0752,
+      "step": 19245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.791003657924484e-06,
+      "loss": 1.165,
+      "step": 19246
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790022069232925e-06,
+      "loss": 1.0732,
+      "step": 19247
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.789040515021793e-06,
+      "loss": 0.8545,
+      "step": 19248
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7880589953016275e-06,
+      "loss": 1.2803,
+      "step": 19249
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.78707751008297e-06,
+      "loss": 0.9053,
+      "step": 19250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.786096059376369e-06,
+      "loss": 0.9404,
+      "step": 19251
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7851146431923675e-06,
+      "loss": 0.7759,
+      "step": 19252
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.784133261541501e-06,
+      "loss": 1.0996,
+      "step": 19253
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.783151914434318e-06,
+      "loss": 0.8691,
+      "step": 19254
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.78217060188135e-06,
+      "loss": 1.083,
+      "step": 19255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.78118932389315e-06,
+      "loss": 0.8936,
+      "step": 19256
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.780208080480252e-06,
+      "loss": 0.8892,
+      "step": 19257
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.779226871653195e-06,
+      "loss": 1.1777,
+      "step": 19258
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.778245697422525e-06,
+      "loss": 0.9375,
+      "step": 19259
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.777264557798775e-06,
+      "loss": 1.0244,
+      "step": 19260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.776283452792483e-06,
+      "loss": 0.6411,
+      "step": 19261
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.775302382414194e-06,
+      "loss": 0.8301,
+      "step": 19262
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.774321346674446e-06,
+      "loss": 1.0352,
+      "step": 19263
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.773340345583771e-06,
+      "loss": 0.5112,
+      "step": 19264
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.77235937915271e-06,
+      "loss": 0.8242,
+      "step": 19265
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.771378447391799e-06,
+      "loss": 1.4961,
+      "step": 19266
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.770397550311578e-06,
+      "loss": 1.0166,
+      "step": 19267
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.769416687922581e-06,
+      "loss": 1.3037,
+      "step": 19268
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7684358602353425e-06,
+      "loss": 1.3545,
+      "step": 19269
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.767455067260403e-06,
+      "loss": 0.8359,
+      "step": 19270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.766474309008291e-06,
+      "loss": 1.3076,
+      "step": 19271
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.765493585489547e-06,
+      "loss": 1.248,
+      "step": 19272
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.764512896714705e-06,
+      "loss": 1.4648,
+      "step": 19273
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.763532242694298e-06,
+      "loss": 0.5823,
+      "step": 19274
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76255162343886e-06,
+      "loss": 1.1318,
+      "step": 19275
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.761571038958921e-06,
+      "loss": 1.0068,
+      "step": 19276
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7605904892650195e-06,
+      "loss": 1.0352,
+      "step": 19277
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.759609974367689e-06,
+      "loss": 0.9287,
+      "step": 19278
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.758629494277455e-06,
+      "loss": 1.0684,
+      "step": 19279
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.757649049004856e-06,
+      "loss": 0.9893,
+      "step": 19280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.756668638560417e-06,
+      "loss": 0.9336,
+      "step": 19281
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.755688262954677e-06,
+      "loss": 1.5039,
+      "step": 19282
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.754707922198165e-06,
+      "loss": 1.0283,
+      "step": 19283
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.753727616301406e-06,
+      "loss": 0.9688,
+      "step": 19284
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7527473452749374e-06,
+      "loss": 0.6816,
+      "step": 19285
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.751767109129278e-06,
+      "loss": 0.875,
+      "step": 19286
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.750786907874971e-06,
+      "loss": 0.8374,
+      "step": 19287
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.749806741522535e-06,
+      "loss": 1.0352,
+      "step": 19288
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.748826610082506e-06,
+      "loss": 1.1641,
+      "step": 19289
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.747846513565407e-06,
+      "loss": 0.9609,
+      "step": 19290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.746866451981766e-06,
+      "loss": 0.9004,
+      "step": 19291
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.745886425342111e-06,
+      "loss": 1.2715,
+      "step": 19292
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744906433656972e-06,
+      "loss": 0.9219,
+      "step": 19293
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.743926476936874e-06,
+      "loss": 0.9204,
+      "step": 19294
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.742946555192342e-06,
+      "loss": 1.1299,
+      "step": 19295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.741966668433904e-06,
+      "loss": 1.0664,
+      "step": 19296
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.740986816672079e-06,
+      "loss": 0.8555,
+      "step": 19297
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.740006999917406e-06,
+      "loss": 0.8018,
+      "step": 19298
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7390272181803965e-06,
+      "loss": 1.1787,
+      "step": 19299
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.73804747147158e-06,
+      "loss": 0.7861,
+      "step": 19300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.737067759801485e-06,
+      "loss": 0.9131,
+      "step": 19301
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.736088083180625e-06,
+      "loss": 1.0488,
+      "step": 19302
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.735108441619532e-06,
+      "loss": 1.4097,
+      "step": 19303
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.734128835128726e-06,
+      "loss": 0.9414,
+      "step": 19304
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7331492637187325e-06,
+      "loss": 1.207,
+      "step": 19305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.732169727400068e-06,
+      "loss": 0.9414,
+      "step": 19306
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7311902261832555e-06,
+      "loss": 1.0117,
+      "step": 19307
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.730210760078819e-06,
+      "loss": 1.2119,
+      "step": 19308
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.729231329097281e-06,
+      "loss": 1.3525,
+      "step": 19309
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7282519332491594e-06,
+      "loss": 0.9561,
+      "step": 19310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.727272572544974e-06,
+      "loss": 0.8892,
+      "step": 19311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.726293246995244e-06,
+      "loss": 0.9233,
+      "step": 19312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.725313956610496e-06,
+      "loss": 1.8105,
+      "step": 19313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.724334701401239e-06,
+      "loss": 0.6873,
+      "step": 19314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.723355481377999e-06,
+      "loss": 1.0928,
+      "step": 19315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.722376296551294e-06,
+      "loss": 1.0781,
+      "step": 19316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.721397146931636e-06,
+      "loss": 1.0625,
+      "step": 19317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7204180325295475e-06,
+      "loss": 1.0693,
+      "step": 19318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.719438953355547e-06,
+      "loss": 0.8916,
+      "step": 19319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7184599094201515e-06,
+      "loss": 1.2285,
+      "step": 19320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.717480900733873e-06,
+      "loss": 0.7563,
+      "step": 19321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.716501927307231e-06,
+      "loss": 1.1592,
+      "step": 19322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.715522989150738e-06,
+      "loss": 0.8755,
+      "step": 19323
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7145440862749154e-06,
+      "loss": 0.9238,
+      "step": 19324
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7135652186902745e-06,
+      "loss": 0.6809,
+      "step": 19325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.712586386407329e-06,
+      "loss": 1.2095,
+      "step": 19326
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.711607589436596e-06,
+      "loss": 1.042,
+      "step": 19327
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.710628827788584e-06,
+      "loss": 1.1152,
+      "step": 19328
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.709650101473814e-06,
+      "loss": 1.4336,
+      "step": 19329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.708671410502794e-06,
+      "loss": 1.0879,
+      "step": 19330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70769275488604e-06,
+      "loss": 0.8486,
+      "step": 19331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70671413463406e-06,
+      "loss": 1.2031,
+      "step": 19332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.705735549757369e-06,
+      "loss": 0.998,
+      "step": 19333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.704757000266478e-06,
+      "loss": 0.9365,
+      "step": 19334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.703778486171901e-06,
+      "loss": 1.1167,
+      "step": 19335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.702800007484144e-06,
+      "loss": 1.1914,
+      "step": 19336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70182156421372e-06,
+      "loss": 0.7896,
+      "step": 19337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7008431563711365e-06,
+      "loss": 1.0986,
+      "step": 19338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6998647839669074e-06,
+      "loss": 0.9121,
+      "step": 19339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.698886447011541e-06,
+      "loss": 0.8066,
+      "step": 19340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.697908145515546e-06,
+      "loss": 1.1367,
+      "step": 19341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.69692987948943e-06,
+      "loss": 1.2402,
+      "step": 19342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.695951648943696e-06,
+      "loss": 1.4199,
+      "step": 19343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.694973453888863e-06,
+      "loss": 0.7812,
+      "step": 19344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.693995294335432e-06,
+      "loss": 0.7715,
+      "step": 19345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.69301717029391e-06,
+      "loss": 0.8789,
+      "step": 19346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.692039081774806e-06,
+      "loss": 1.1353,
+      "step": 19347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.691061028788621e-06,
+      "loss": 1.3721,
+      "step": 19348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.690083011345866e-06,
+      "loss": 0.9072,
+      "step": 19349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.689105029457046e-06,
+      "loss": 0.832,
+      "step": 19350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.688127083132666e-06,
+      "loss": 1.2407,
+      "step": 19351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.687149172383228e-06,
+      "loss": 1.4922,
+      "step": 19352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.686171297219242e-06,
+      "loss": 1.1758,
+      "step": 19353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6851934576512e-06,
+      "loss": 1.2354,
+      "step": 19354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.684215653689621e-06,
+      "loss": 1.0605,
+      "step": 19355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.683237885345e-06,
+      "loss": 1.1768,
+      "step": 19356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.682260152627839e-06,
+      "loss": 0.6807,
+      "step": 19357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6812824555486456e-06,
+      "loss": 0.8787,
+      "step": 19358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.680304794117914e-06,
+      "loss": 1.3579,
+      "step": 19359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.679327168346153e-06,
+      "loss": 0.9336,
+      "step": 19360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.678349578243861e-06,
+      "loss": 0.9663,
+      "step": 19361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.677372023821542e-06,
+      "loss": 0.8975,
+      "step": 19362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6763945050896915e-06,
+      "loss": 1.2676,
+      "step": 19363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.67541702205881e-06,
+      "loss": 0.917,
+      "step": 19364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.674439574739402e-06,
+      "loss": 0.6865,
+      "step": 19365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.673462163141966e-06,
+      "loss": 0.8887,
+      "step": 19366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.672484787276996e-06,
+      "loss": 0.8779,
+      "step": 19367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.671507447154994e-06,
+      "loss": 1.3027,
+      "step": 19368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.670530142786458e-06,
+      "loss": 0.876,
+      "step": 19369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.669552874181888e-06,
+      "loss": 0.8828,
+      "step": 19370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.668575641351778e-06,
+      "loss": 0.9092,
+      "step": 19371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.667598444306627e-06,
+      "loss": 1.0005,
+      "step": 19372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6666212830569325e-06,
+      "loss": 1.2363,
+      "step": 19373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6656441576131826e-06,
+      "loss": 0.9062,
+      "step": 19374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.664667067985888e-06,
+      "loss": 0.8057,
+      "step": 19375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.663690014185533e-06,
+      "loss": 1.0547,
+      "step": 19376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.662712996222618e-06,
+      "loss": 0.959,
+      "step": 19377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.661736014107635e-06,
+      "loss": 1.0117,
+      "step": 19378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.660759067851078e-06,
+      "loss": 0.8252,
+      "step": 19379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.659782157463441e-06,
+      "loss": 1.0156,
+      "step": 19380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.658805282955223e-06,
+      "loss": 0.9863,
+      "step": 19381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.657828444336911e-06,
+      "loss": 1.1084,
+      "step": 19382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.656851641619e-06,
+      "loss": 1.2812,
+      "step": 19383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.655874874811984e-06,
+      "loss": 0.7842,
+      "step": 19384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.654898143926347e-06,
+      "loss": 1.459,
+      "step": 19385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.653921448972594e-06,
+      "loss": 1.0703,
+      "step": 19386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.652944789961209e-06,
+      "loss": 1.0879,
+      "step": 19387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.651968166902683e-06,
+      "loss": 1.1318,
+      "step": 19388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6509915798075066e-06,
+      "loss": 0.6377,
+      "step": 19389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6500150286861676e-06,
+      "loss": 0.8662,
+      "step": 19390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.64903851354916e-06,
+      "loss": 1.0444,
+      "step": 19391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.648062034406972e-06,
+      "loss": 1.1338,
+      "step": 19392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.647085591270094e-06,
+      "loss": 0.9375,
+      "step": 19393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.646109184149011e-06,
+      "loss": 0.9121,
+      "step": 19394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.645132813054211e-06,
+      "loss": 1.1309,
+      "step": 19395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.644156477996186e-06,
+      "loss": 1.0254,
+      "step": 19396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.643180178985423e-06,
+      "loss": 0.9102,
+      "step": 19397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.642203916032406e-06,
+      "loss": 1.0361,
+      "step": 19398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.641227689147623e-06,
+      "loss": 1.0439,
+      "step": 19399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6402514983415565e-06,
+      "loss": 1.1074,
+      "step": 19400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.639275343624703e-06,
+      "loss": 1.1143,
+      "step": 19401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.638299225007537e-06,
+      "loss": 0.7993,
+      "step": 19402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.637323142500549e-06,
+      "loss": 1.0576,
+      "step": 19403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.636347096114224e-06,
+      "loss": 1.1162,
+      "step": 19404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.635371085859041e-06,
+      "loss": 1.1572,
+      "step": 19405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.63439511174549e-06,
+      "loss": 1.124,
+      "step": 19406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6334191737840525e-06,
+      "loss": 1.0959,
+      "step": 19407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.632443271985213e-06,
+      "loss": 1.1143,
+      "step": 19408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.631467406359451e-06,
+      "loss": 1.1455,
+      "step": 19409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6304915769172505e-06,
+      "loss": 1.0879,
+      "step": 19410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.629515783669092e-06,
+      "loss": 0.6396,
+      "step": 19411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.628540026625463e-06,
+      "loss": 1.2549,
+      "step": 19412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.627564305796839e-06,
+      "loss": 0.8389,
+      "step": 19413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6265886211937015e-06,
+      "loss": 1.1582,
+      "step": 19414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.625612972826534e-06,
+      "loss": 1.0576,
+      "step": 19415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.62463736070581e-06,
+      "loss": 1.0176,
+      "step": 19416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.623661784842017e-06,
+      "loss": 1.1416,
+      "step": 19417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6226862452456295e-06,
+      "loss": 1.125,
+      "step": 19418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.621710741927131e-06,
+      "loss": 0.9419,
+      "step": 19419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.620735274896994e-06,
+      "loss": 1.4346,
+      "step": 19420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.619759844165698e-06,
+      "loss": 1.0918,
+      "step": 19421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.618784449743723e-06,
+      "loss": 0.8555,
+      "step": 19422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.617809091641549e-06,
+      "loss": 0.9902,
+      "step": 19423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.616833769869646e-06,
+      "loss": 1.019,
+      "step": 19424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.615858484438496e-06,
+      "loss": 0.9795,
+      "step": 19425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.614883235358569e-06,
+      "loss": 1.1006,
+      "step": 19426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.613908022640348e-06,
+      "loss": 1.0713,
+      "step": 19427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6129328462943064e-06,
+      "loss": 1.0439,
+      "step": 19428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6119577063309155e-06,
+      "loss": 1.0576,
+      "step": 19429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.610982602760657e-06,
+      "loss": 0.9863,
+      "step": 19430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.610007535593993e-06,
+      "loss": 1.2129,
+      "step": 19431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60903250484141e-06,
+      "loss": 1.1494,
+      "step": 19432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.608057510513377e-06,
+      "loss": 0.9224,
+      "step": 19433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.607082552620368e-06,
+      "loss": 1.3857,
+      "step": 19434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6061076311728485e-06,
+      "loss": 1.1973,
+      "step": 19435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.605132746181298e-06,
+      "loss": 0.8818,
+      "step": 19436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.604157897656187e-06,
+      "loss": 0.8369,
+      "step": 19437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.603183085607987e-06,
+      "loss": 0.9336,
+      "step": 19438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.602208310047171e-06,
+      "loss": 1.0684,
+      "step": 19439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.601233570984205e-06,
+      "loss": 0.9458,
+      "step": 19440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.600258868429563e-06,
+      "loss": 0.9463,
+      "step": 19441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.599284202393709e-06,
+      "loss": 1.0742,
+      "step": 19442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.598309572887123e-06,
+      "loss": 1.1416,
+      "step": 19443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.597334979920264e-06,
+      "loss": 0.9399,
+      "step": 19444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596360423503607e-06,
+      "loss": 0.6099,
+      "step": 19445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.59538590364762e-06,
+      "loss": 1.1123,
+      "step": 19446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.594411420362765e-06,
+      "loss": 1.1641,
+      "step": 19447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.593436973659515e-06,
+      "loss": 0.8721,
+      "step": 19448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.592462563548337e-06,
+      "loss": 0.9385,
+      "step": 19449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.591488190039699e-06,
+      "loss": 1.0869,
+      "step": 19450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5905138531440625e-06,
+      "loss": 0.9976,
+      "step": 19451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.589539552871893e-06,
+      "loss": 0.998,
+      "step": 19452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.588565289233661e-06,
+      "loss": 1.2344,
+      "step": 19453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.587591062239834e-06,
+      "loss": 0.9609,
+      "step": 19454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.586616871900869e-06,
+      "loss": 1.2188,
+      "step": 19455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.585642718227235e-06,
+      "loss": 0.8638,
+      "step": 19456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5846686012293915e-06,
+      "loss": 1.0625,
+      "step": 19457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.583694520917812e-06,
+      "loss": 0.8896,
+      "step": 19458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.582720477302948e-06,
+      "loss": 0.7783,
+      "step": 19459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.581746470395271e-06,
+      "loss": 0.9219,
+      "step": 19460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.580772500205239e-06,
+      "loss": 1.1348,
+      "step": 19461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.579798566743314e-06,
+      "loss": 0.9512,
+      "step": 19462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.578824670019959e-06,
+      "loss": 0.9766,
+      "step": 19463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.577850810045635e-06,
+      "loss": 0.6211,
+      "step": 19464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.576876986830806e-06,
+      "loss": 1.0762,
+      "step": 19465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.575903200385927e-06,
+      "loss": 0.8926,
+      "step": 19466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.574929450721456e-06,
+      "loss": 0.9473,
+      "step": 19467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.573955737847861e-06,
+      "loss": 1.2031,
+      "step": 19468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.572982061775601e-06,
+      "loss": 0.5342,
+      "step": 19469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.572008422515127e-06,
+      "loss": 1.0303,
+      "step": 19470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.571034820076903e-06,
+      "loss": 1.2119,
+      "step": 19471
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.570061254471386e-06,
+      "loss": 1.1846,
+      "step": 19472
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.569087725709028e-06,
+      "loss": 1.1689,
+      "step": 19473
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.568114233800298e-06,
+      "loss": 1.1514,
+      "step": 19474
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.567140778755645e-06,
+      "loss": 0.9014,
+      "step": 19475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.566167360585529e-06,
+      "loss": 0.8359,
+      "step": 19476
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5651939793004025e-06,
+      "loss": 1.1465,
+      "step": 19477
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.56422063491072e-06,
+      "loss": 1.5273,
+      "step": 19478
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.563247327426942e-06,
+      "loss": 1.2842,
+      "step": 19479
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.562274056859525e-06,
+      "loss": 0.9814,
+      "step": 19480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.561300823218917e-06,
+      "loss": 1.0762,
+      "step": 19481
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5603276265155734e-06,
+      "loss": 1.0786,
+      "step": 19482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.559354466759949e-06,
+      "loss": 0.8398,
+      "step": 19483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5583813439624975e-06,
+      "loss": 0.7319,
+      "step": 19484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.557408258133675e-06,
+      "loss": 0.9395,
+      "step": 19485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5564352092839286e-06,
+      "loss": 1.1226,
+      "step": 19486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5554621974237165e-06,
+      "loss": 1.1162,
+      "step": 19487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.554489222563478e-06,
+      "loss": 0.7388,
+      "step": 19488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.553516284713681e-06,
+      "loss": 0.8887,
+      "step": 19489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.552543383884766e-06,
+      "loss": 1.1855,
+      "step": 19490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.551570520087185e-06,
+      "loss": 0.7659,
+      "step": 19491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.550597693331393e-06,
+      "loss": 0.9712,
+      "step": 19492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.549624903627831e-06,
+      "loss": 1.0137,
+      "step": 19493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.548652150986955e-06,
+      "loss": 0.8701,
+      "step": 19494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.547679435419213e-06,
+      "loss": 0.5562,
+      "step": 19495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.546706756935055e-06,
+      "loss": 1.2598,
+      "step": 19496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.545734115544924e-06,
+      "loss": 0.9355,
+      "step": 19497
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.54476151125927e-06,
+      "loss": 0.8359,
+      "step": 19498
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5437889440885425e-06,
+      "loss": 1.0713,
+      "step": 19499
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.542816414043189e-06,
+      "loss": 1.041,
+      "step": 19500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.541843921133653e-06,
+      "loss": 0.8818,
+      "step": 19501
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.540871465370382e-06,
+      "loss": 0.9668,
+      "step": 19502
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.539899046763822e-06,
+      "loss": 1.0801,
+      "step": 19503
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.538926665324417e-06,
+      "loss": 0.957,
+      "step": 19504
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5379543210626115e-06,
+      "loss": 1.0605,
+      "step": 19505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.536982013988853e-06,
+      "loss": 0.999,
+      "step": 19506
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5360097441135875e-06,
+      "loss": 1.2339,
+      "step": 19507
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.535037511447252e-06,
+      "loss": 0.9443,
+      "step": 19508
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.534065316000291e-06,
+      "loss": 1.042,
+      "step": 19509
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.533093157783152e-06,
+      "loss": 1.0361,
+      "step": 19510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.532121036806279e-06,
+      "loss": 0.8721,
+      "step": 19511
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.531148953080107e-06,
+      "loss": 1.002,
+      "step": 19512
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.530176906615082e-06,
+      "loss": 1.1025,
+      "step": 19513
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.529204897421644e-06,
+      "loss": 0.7109,
+      "step": 19514
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.528232925510236e-06,
+      "loss": 0.6167,
+      "step": 19515
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5272609908912975e-06,
+      "loss": 1.0264,
+      "step": 19516
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5262890935752665e-06,
+      "loss": 1.3555,
+      "step": 19517
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.525317233572588e-06,
+      "loss": 0.8218,
+      "step": 19518
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524345410893692e-06,
+      "loss": 0.7397,
+      "step": 19519
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.52337362554903e-06,
+      "loss": 1.5762,
+      "step": 19520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.522401877549032e-06,
+      "loss": 1.0254,
+      "step": 19521
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.521430166904139e-06,
+      "loss": 0.7305,
+      "step": 19522
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5204584936247865e-06,
+      "loss": 1.1924,
+      "step": 19523
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.519486857721411e-06,
+      "loss": 1.2314,
+      "step": 19524
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.518515259204455e-06,
+      "loss": 0.8081,
+      "step": 19525
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.517543698084352e-06,
+      "loss": 0.752,
+      "step": 19526
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5165721743715405e-06,
+      "loss": 0.915,
+      "step": 19527
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.51560068807645e-06,
+      "loss": 1.2939,
+      "step": 19528
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.514629239209519e-06,
+      "loss": 0.7773,
+      "step": 19529
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.513657827781184e-06,
+      "loss": 1.0947,
+      "step": 19530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.512686453801882e-06,
+      "loss": 0.9756,
+      "step": 19531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.511715117282041e-06,
+      "loss": 1.1455,
+      "step": 19532
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.510743818232101e-06,
+      "loss": 1.1855,
+      "step": 19533
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.509772556662488e-06,
+      "loss": 0.9648,
+      "step": 19534
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.508801332583638e-06,
+      "loss": 1.0308,
+      "step": 19535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5078301460059865e-06,
+      "loss": 0.9082,
+      "step": 19536
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5068589969399646e-06,
+      "loss": 1.043,
+      "step": 19537
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505887885396003e-06,
+      "loss": 1.1289,
+      "step": 19538
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.504916811384532e-06,
+      "loss": 1.4414,
+      "step": 19539
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5039457749159805e-06,
+      "loss": 0.9268,
+      "step": 19540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5029747760007854e-06,
+      "loss": 1.1338,
+      "step": 19541
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5020038146493755e-06,
+      "loss": 1.0234,
+      "step": 19542
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.501032890872177e-06,
+      "loss": 0.791,
+      "step": 19543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.500062004679619e-06,
+      "loss": 0.8765,
+      "step": 19544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4990911560821316e-06,
+      "loss": 0.7417,
+      "step": 19545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4981203450901464e-06,
+      "loss": 0.7207,
+      "step": 19546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.497149571714087e-06,
+      "loss": 0.917,
+      "step": 19547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4961788359643844e-06,
+      "loss": 0.8682,
+      "step": 19548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.495208137851465e-06,
+      "loss": 0.8594,
+      "step": 19549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.494237477385753e-06,
+      "loss": 0.936,
+      "step": 19550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.493266854577678e-06,
+      "loss": 0.9702,
+      "step": 19551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.492296269437666e-06,
+      "loss": 0.7603,
+      "step": 19552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.491325721976142e-06,
+      "loss": 1.062,
+      "step": 19553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.490355212203532e-06,
+      "loss": 0.8359,
+      "step": 19554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.489384740130256e-06,
+      "loss": 0.8867,
+      "step": 19555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4884143057667445e-06,
+      "loss": 1.0684,
+      "step": 19556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.487443909123422e-06,
+      "loss": 1.1064,
+      "step": 19557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4864735502107075e-06,
+      "loss": 0.9248,
+      "step": 19558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.485503229039028e-06,
+      "loss": 1.0449,
+      "step": 19559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4845329456188e-06,
+      "loss": 0.8774,
+      "step": 19560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.483562699960456e-06,
+      "loss": 1.0332,
+      "step": 19561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.482592492074411e-06,
+      "loss": 0.8809,
+      "step": 19562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4816223219710884e-06,
+      "loss": 0.9526,
+      "step": 19563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.48065218966091e-06,
+      "loss": 0.9795,
+      "step": 19564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.479682095154296e-06,
+      "loss": 1.1074,
+      "step": 19565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.478712038461663e-06,
+      "loss": 0.9502,
+      "step": 19566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.477742019593438e-06,
+      "loss": 0.7524,
+      "step": 19567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.476772038560039e-06,
+      "loss": 0.7676,
+      "step": 19568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.475802095371882e-06,
+      "loss": 0.5588,
+      "step": 19569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4748321900393865e-06,
+      "loss": 1.0586,
+      "step": 19570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.473862322572968e-06,
+      "loss": 0.8052,
+      "step": 19571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.472892492983053e-06,
+      "loss": 1.124,
+      "step": 19572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.471922701280055e-06,
+      "loss": 0.9502,
+      "step": 19573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.470952947474389e-06,
+      "loss": 1.0488,
+      "step": 19574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.469983231576475e-06,
+      "loss": 1.0737,
+      "step": 19575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.469013553596722e-06,
+      "loss": 0.9321,
+      "step": 19576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.468043913545556e-06,
+      "loss": 0.8779,
+      "step": 19577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.467074311433385e-06,
+      "loss": 0.9182,
+      "step": 19578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4661047472706315e-06,
+      "loss": 1.1147,
+      "step": 19579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.465135221067704e-06,
+      "loss": 1.0527,
+      "step": 19580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.464165732835015e-06,
+      "loss": 0.7646,
+      "step": 19581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.463196282582984e-06,
+      "loss": 1.1172,
+      "step": 19582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.462226870322023e-06,
+      "loss": 0.8086,
+      "step": 19583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.461257496062546e-06,
+      "loss": 0.9619,
+      "step": 19584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.460288159814962e-06,
+      "loss": 1.2646,
+      "step": 19585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4593188615896855e-06,
+      "loss": 1.2793,
+      "step": 19586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.458349601397128e-06,
+      "loss": 1.0752,
+      "step": 19587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.457380379247704e-06,
+      "loss": 0.875,
+      "step": 19588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.456411195151819e-06,
+      "loss": 1.0107,
+      "step": 19589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.455442049119888e-06,
+      "loss": 0.9434,
+      "step": 19590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.454472941162317e-06,
+      "loss": 0.7383,
+      "step": 19591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.453503871289521e-06,
+      "loss": 0.8037,
+      "step": 19592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4525348395119055e-06,
+      "loss": 1.0674,
+      "step": 19593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.451565845839882e-06,
+      "loss": 1.043,
+      "step": 19594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45059689028386e-06,
+      "loss": 0.8044,
+      "step": 19595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.449627972854242e-06,
+      "loss": 0.7646,
+      "step": 19596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.448659093561438e-06,
+      "loss": 1.207,
+      "step": 19597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.447690252415859e-06,
+      "loss": 0.9482,
+      "step": 19598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.446721449427911e-06,
+      "loss": 0.8359,
+      "step": 19599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.445752684607998e-06,
+      "loss": 1.3057,
+      "step": 19600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.444783957966525e-06,
+      "loss": 1.2168,
+      "step": 19601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4438152695139e-06,
+      "loss": 1.0225,
+      "step": 19602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.442846619260532e-06,
+      "loss": 0.9053,
+      "step": 19603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.44187800721682e-06,
+      "loss": 0.7793,
+      "step": 19604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.44090943339317e-06,
+      "loss": 0.8467,
+      "step": 19605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43994089779999e-06,
+      "loss": 0.7607,
+      "step": 19606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.438972400447676e-06,
+      "loss": 1.1084,
+      "step": 19607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.438003941346638e-06,
+      "loss": 1.5146,
+      "step": 19608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.437035520507276e-06,
+      "loss": 0.6157,
+      "step": 19609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.436067137939994e-06,
+      "loss": 0.8406,
+      "step": 19610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.435098793655192e-06,
+      "loss": 0.9717,
+      "step": 19611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.434130487663269e-06,
+      "loss": 0.8584,
+      "step": 19612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.433162219974633e-06,
+      "loss": 0.7549,
+      "step": 19613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4321939905996845e-06,
+      "loss": 0.6655,
+      "step": 19614
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.431225799548817e-06,
+      "loss": 1.0098,
+      "step": 19615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.430257646832434e-06,
+      "loss": 0.9492,
+      "step": 19616
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.429289532460935e-06,
+      "loss": 0.8047,
+      "step": 19617
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.428321456444722e-06,
+      "loss": 0.8867,
+      "step": 19618
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.427353418794191e-06,
+      "loss": 0.9092,
+      "step": 19619
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.426385419519741e-06,
+      "loss": 0.6519,
+      "step": 19620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.425417458631769e-06,
+      "loss": 1.1182,
+      "step": 19621
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.424449536140668e-06,
+      "loss": 1.2578,
+      "step": 19622
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4234816520568464e-06,
+      "loss": 1.0186,
+      "step": 19623
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.422513806390691e-06,
+      "loss": 0.9932,
+      "step": 19624
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.421545999152606e-06,
+      "loss": 1.1855,
+      "step": 19625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.420578230352978e-06,
+      "loss": 0.7646,
+      "step": 19626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.419610500002209e-06,
+      "loss": 0.9707,
+      "step": 19627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.418642808110689e-06,
+      "loss": 1.1953,
+      "step": 19628
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.417675154688818e-06,
+      "loss": 1.208,
+      "step": 19629
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.416707539746991e-06,
+      "loss": 1.0664,
+      "step": 19630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.415739963295596e-06,
+      "loss": 1.0781,
+      "step": 19631
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4147724253450315e-06,
+      "loss": 1.04,
+      "step": 19632
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.413804925905683e-06,
+      "loss": 0.752,
+      "step": 19633
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.412837464987951e-06,
+      "loss": 0.8921,
+      "step": 19634
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.411870042602224e-06,
+      "loss": 0.9512,
+      "step": 19635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.410902658758895e-06,
+      "loss": 1.1152,
+      "step": 19636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4099353134683575e-06,
+      "loss": 1.0127,
+      "step": 19637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.408968006740994e-06,
+      "loss": 0.7842,
+      "step": 19638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.408000738587203e-06,
+      "loss": 1.0957,
+      "step": 19639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.407033509017372e-06,
+      "loss": 0.8672,
+      "step": 19640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.406066318041892e-06,
+      "loss": 0.8821,
+      "step": 19641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40509916567115e-06,
+      "loss": 0.8232,
+      "step": 19642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.404132051915534e-06,
+      "loss": 0.7949,
+      "step": 19643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.403164976785434e-06,
+      "loss": 1.335,
+      "step": 19644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.402197940291243e-06,
+      "loss": 1.3232,
+      "step": 19645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.401230942443339e-06,
+      "loss": 0.9404,
+      "step": 19646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.400263983252115e-06,
+      "loss": 0.9248,
+      "step": 19647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.399297062727954e-06,
+      "loss": 0.9902,
+      "step": 19648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.398330180881249e-06,
+      "loss": 0.9746,
+      "step": 19649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.39736333772238e-06,
+      "loss": 0.6919,
+      "step": 19650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.396396533261735e-06,
+      "loss": 1.126,
+      "step": 19651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.395429767509698e-06,
+      "loss": 0.9854,
+      "step": 19652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.394463040476651e-06,
+      "loss": 1.458,
+      "step": 19653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.393496352172983e-06,
+      "loss": 1.1221,
+      "step": 19654
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.392529702609076e-06,
+      "loss": 0.873,
+      "step": 19655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3915630917953165e-06,
+      "loss": 1.0449,
+      "step": 19656
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3905965197420805e-06,
+      "loss": 1.1797,
+      "step": 19657
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.389629986459756e-06,
+      "loss": 0.8628,
+      "step": 19658
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3886634919587195e-06,
+      "loss": 0.7834,
+      "step": 19659
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.387697036249363e-06,
+      "loss": 1.2168,
+      "step": 19660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.386730619342058e-06,
+      "loss": 1.3818,
+      "step": 19661
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.38576424124719e-06,
+      "loss": 1.417,
+      "step": 19662
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.384797901975139e-06,
+      "loss": 0.8896,
+      "step": 19663
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.383831601536279e-06,
+      "loss": 0.9795,
+      "step": 19664
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.382865339941002e-06,
+      "loss": 0.6929,
+      "step": 19665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.381899117199678e-06,
+      "loss": 0.9395,
+      "step": 19666
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.38093293332269e-06,
+      "loss": 1.4316,
+      "step": 19667
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.379966788320413e-06,
+      "loss": 0.8057,
+      "step": 19668
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.379000682203224e-06,
+      "loss": 1.2197,
+      "step": 19669
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.378034614981507e-06,
+      "loss": 1.1367,
+      "step": 19670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.377068586665634e-06,
+      "loss": 0.9932,
+      "step": 19671
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.376102597265985e-06,
+      "loss": 0.8662,
+      "step": 19672
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3751366467929324e-06,
+      "loss": 0.9453,
+      "step": 19673
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.374170735256853e-06,
+      "loss": 1.1484,
+      "step": 19674
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.373204862668123e-06,
+      "loss": 0.9893,
+      "step": 19675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.372239029037122e-06,
+      "loss": 1.1084,
+      "step": 19676
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.371273234374218e-06,
+      "loss": 1.0996,
+      "step": 19677
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.37030747868979e-06,
+      "loss": 1.0376,
+      "step": 19678
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.369341761994202e-06,
+      "loss": 1.0967,
+      "step": 19679
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.368376084297841e-06,
+      "loss": 0.9521,
+      "step": 19680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.367410445611073e-06,
+      "loss": 1.5742,
+      "step": 19681
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36644484594427e-06,
+      "loss": 0.9346,
+      "step": 19682
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36547928530781e-06,
+      "loss": 0.6807,
+      "step": 19683
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.364513763712054e-06,
+      "loss": 1.0947,
+      "step": 19684
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.363548281167381e-06,
+      "loss": 1.0127,
+      "step": 19685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.362582837684161e-06,
+      "loss": 0.916,
+      "step": 19686
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.361617433272767e-06,
+      "loss": 1.1475,
+      "step": 19687
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.360652067943562e-06,
+      "loss": 1.0898,
+      "step": 19688
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.35968674170692e-06,
+      "loss": 1.0303,
+      "step": 19689
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3587214545732066e-06,
+      "loss": 0.832,
+      "step": 19690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.357756206552798e-06,
+      "loss": 1.1484,
+      "step": 19691
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.356790997656057e-06,
+      "loss": 1.1484,
+      "step": 19692
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.355825827893353e-06,
+      "loss": 1.0254,
+      "step": 19693
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.354860697275054e-06,
+      "loss": 0.9346,
+      "step": 19694
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.353895605811522e-06,
+      "loss": 1.1807,
+      "step": 19695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.35293055351313e-06,
+      "loss": 1.0596,
+      "step": 19696
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.351965540390244e-06,
+      "loss": 1.2422,
+      "step": 19697
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.351000566453227e-06,
+      "loss": 0.9814,
+      "step": 19698
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.350035631712446e-06,
+      "loss": 1.3125,
+      "step": 19699
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.349070736178261e-06,
+      "loss": 0.6357,
+      "step": 19700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.348105879861044e-06,
+      "loss": 1.3867,
+      "step": 19701
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.347141062771158e-06,
+      "loss": 1.0381,
+      "step": 19702
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.346176284918962e-06,
+      "loss": 1.0986,
+      "step": 19703
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.345211546314823e-06,
+      "loss": 0.791,
+      "step": 19704
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3442468469691e-06,
+      "loss": 1.0176,
+      "step": 19705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.343282186892163e-06,
+      "loss": 1.0459,
+      "step": 19706
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.342317566094367e-06,
+      "loss": 0.752,
+      "step": 19707
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.341352984586077e-06,
+      "loss": 0.9194,
+      "step": 19708
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.340388442377654e-06,
+      "loss": 0.9453,
+      "step": 19709
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.339423939479452e-06,
+      "loss": 0.9009,
+      "step": 19710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.338459475901844e-06,
+      "loss": 1.3613,
+      "step": 19711
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.337495051655181e-06,
+      "loss": 1.1826,
+      "step": 19712
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.336530666749826e-06,
+      "loss": 0.9395,
+      "step": 19713
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.335566321196135e-06,
+      "loss": 0.9355,
+      "step": 19714
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3346020150044695e-06,
+      "loss": 1.2549,
+      "step": 19715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.333637748185183e-06,
+      "loss": 0.875,
+      "step": 19716
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.33267352074864e-06,
+      "loss": 0.5845,
+      "step": 19717
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.331709332705196e-06,
+      "loss": 1.1182,
+      "step": 19718
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.330745184065205e-06,
+      "loss": 1.1973,
+      "step": 19719
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.329781074839029e-06,
+      "loss": 0.729,
+      "step": 19720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.328817005037012e-06,
+      "loss": 1.2969,
+      "step": 19721
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.327852974669526e-06,
+      "loss": 0.7356,
+      "step": 19722
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.326888983746915e-06,
+      "loss": 0.998,
+      "step": 19723
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.32592503227954e-06,
+      "loss": 0.9941,
+      "step": 19724
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3249611202777506e-06,
+      "loss": 0.7397,
+      "step": 19725
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.323997247751901e-06,
+      "loss": 0.9185,
+      "step": 19726
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.323033414712349e-06,
+      "loss": 1.1533,
+      "step": 19727
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3220696211694445e-06,
+      "loss": 1.374,
+      "step": 19728
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.321105867133544e-06,
+      "loss": 1.1445,
+      "step": 19729
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3201421526149945e-06,
+      "loss": 1.1738,
+      "step": 19730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.319178477624148e-06,
+      "loss": 0.7256,
+      "step": 19731
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.31821484217136e-06,
+      "loss": 0.7954,
+      "step": 19732
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.317251246266982e-06,
+      "loss": 1.1709,
+      "step": 19733
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.316287689921361e-06,
+      "loss": 1.2344,
+      "step": 19734
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3153241731448475e-06,
+      "loss": 0.9746,
+      "step": 19735
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.314360695947791e-06,
+      "loss": 1.1836,
+      "step": 19736
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.313397258340546e-06,
+      "loss": 0.8311,
+      "step": 19737
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.312433860333456e-06,
+      "loss": 0.9258,
+      "step": 19738
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.311470501936869e-06,
+      "loss": 0.8926,
+      "step": 19739
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.310507183161139e-06,
+      "loss": 0.9141,
+      "step": 19740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.309543904016604e-06,
+      "loss": 0.7402,
+      "step": 19741
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.308580664513621e-06,
+      "loss": 0.8516,
+      "step": 19742
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.307617464662531e-06,
+      "loss": 0.8276,
+      "step": 19743
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.306654304473685e-06,
+      "loss": 0.8276,
+      "step": 19744
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.305691183957423e-06,
+      "loss": 0.8066,
+      "step": 19745
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.304728103124094e-06,
+      "loss": 1.0254,
+      "step": 19746
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3037650619840396e-06,
+      "loss": 1.3184,
+      "step": 19747
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3028020605476105e-06,
+      "loss": 1.1172,
+      "step": 19748
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.301839098825148e-06,
+      "loss": 0.9756,
+      "step": 19749
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.300876176826993e-06,
+      "loss": 0.6316,
+      "step": 19750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.299913294563495e-06,
+      "loss": 1.2324,
+      "step": 19751
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.29895045204499e-06,
+      "loss": 0.9512,
+      "step": 19752
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.297987649281825e-06,
+      "loss": 0.7954,
+      "step": 19753
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.297024886284342e-06,
+      "loss": 1.1934,
+      "step": 19754
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.296062163062882e-06,
+      "loss": 0.834,
+      "step": 19755
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.295099479627784e-06,
+      "loss": 1.127,
+      "step": 19756
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.29413683598939e-06,
+      "loss": 0.9014,
+      "step": 19757
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.293174232158041e-06,
+      "loss": 1.0537,
+      "step": 19758
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.292211668144081e-06,
+      "loss": 0.9263,
+      "step": 19759
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.291249143957843e-06,
+      "loss": 0.9873,
+      "step": 19760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.290286659609668e-06,
+      "loss": 1.3501,
+      "step": 19761
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2893242151098936e-06,
+      "loss": 1.1436,
+      "step": 19762
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.288361810468861e-06,
+      "loss": 0.7686,
+      "step": 19763
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.287399445696909e-06,
+      "loss": 1.1387,
+      "step": 19764
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.286437120804369e-06,
+      "loss": 0.8643,
+      "step": 19765
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.285474835801586e-06,
+      "loss": 0.624,
+      "step": 19766
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.284512590698886e-06,
+      "loss": 0.8926,
+      "step": 19767
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.283550385506616e-06,
+      "loss": 0.6328,
+      "step": 19768
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.282588220235105e-06,
+      "loss": 1.1611,
+      "step": 19769
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2816260948946905e-06,
+      "loss": 0.666,
+      "step": 19770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.280664009495709e-06,
+      "loss": 1.1055,
+      "step": 19771
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.279701964048488e-06,
+      "loss": 0.9922,
+      "step": 19772
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.278739958563369e-06,
+      "loss": 0.8779,
+      "step": 19773
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.277777993050682e-06,
+      "loss": 0.8379,
+      "step": 19774
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.276816067520763e-06,
+      "loss": 1.0342,
+      "step": 19775
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.275854181983941e-06,
+      "loss": 0.9478,
+      "step": 19776
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.274892336450551e-06,
+      "loss": 0.8916,
+      "step": 19777
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.273930530930919e-06,
+      "loss": 1.0928,
+      "step": 19778
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.272968765435384e-06,
+      "loss": 0.9795,
+      "step": 19779
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2720070399742735e-06,
+      "loss": 0.7148,
+      "step": 19780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.271045354557918e-06,
+      "loss": 1.1182,
+      "step": 19781
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.27008370919665e-06,
+      "loss": 0.9307,
+      "step": 19782
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.269122103900793e-06,
+      "loss": 0.5869,
+      "step": 19783
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2681605386806835e-06,
+      "loss": 0.9731,
+      "step": 19784
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.267199013546645e-06,
+      "loss": 0.3525,
+      "step": 19785
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.26623752850901e-06,
+      "loss": 0.8599,
+      "step": 19786
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.265276083578103e-06,
+      "loss": 0.9287,
+      "step": 19787
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.26431467876425e-06,
+      "loss": 1.3262,
+      "step": 19788
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.263353314077785e-06,
+      "loss": 1.335,
+      "step": 19789
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2623919895290305e-06,
+      "loss": 1.0859,
+      "step": 19790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.261430705128311e-06,
+      "loss": 1.3887,
+      "step": 19791
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.260469460885954e-06,
+      "loss": 1.0801,
+      "step": 19792
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.259508256812284e-06,
+      "loss": 0.7002,
+      "step": 19793
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2585470929176285e-06,
+      "loss": 1.1992,
+      "step": 19794
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.25758596921231e-06,
+      "loss": 1.4844,
+      "step": 19795
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.256624885706652e-06,
+      "loss": 0.77,
+      "step": 19796
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.255663842410982e-06,
+      "loss": 1.1084,
+      "step": 19797
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.254702839335612e-06,
+      "loss": 0.6182,
+      "step": 19798
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.25374187649088e-06,
+      "loss": 0.5972,
+      "step": 19799
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.252780953887099e-06,
+      "loss": 1.2148,
+      "step": 19800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2518200715345935e-06,
+      "loss": 0.8096,
+      "step": 19801
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.250859229443684e-06,
+      "loss": 1.2275,
+      "step": 19802
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.24989842762469e-06,
+      "loss": 1.0342,
+      "step": 19803
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.248937666087935e-06,
+      "loss": 0.9297,
+      "step": 19804
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.247976944843741e-06,
+      "loss": 0.6943,
+      "step": 19805
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.247016263902424e-06,
+      "loss": 0.9385,
+      "step": 19806
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.246055623274302e-06,
+      "loss": 1.1367,
+      "step": 19807
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.245095022969698e-06,
+      "loss": 0.9424,
+      "step": 19808
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2441344629989245e-06,
+      "loss": 1.0928,
+      "step": 19809
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.243173943372307e-06,
+      "loss": 1.1934,
+      "step": 19810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.24221346410016e-06,
+      "loss": 1.1064,
+      "step": 19811
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.241253025192801e-06,
+      "loss": 1.209,
+      "step": 19812
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.240292626660544e-06,
+      "loss": 1.0742,
+      "step": 19813
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.239332268513705e-06,
+      "loss": 1.0566,
+      "step": 19814
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.238371950762603e-06,
+      "loss": 0.9023,
+      "step": 19815
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.237411673417553e-06,
+      "loss": 1.1699,
+      "step": 19816
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.236451436488871e-06,
+      "loss": 0.7222,
+      "step": 19817
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.235491239986869e-06,
+      "loss": 1.2637,
+      "step": 19818
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.234531083921858e-06,
+      "loss": 1.0171,
+      "step": 19819
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.233570968304159e-06,
+      "loss": 1.1338,
+      "step": 19820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.232610893144082e-06,
+      "loss": 1.0127,
+      "step": 19821
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231650858451939e-06,
+      "loss": 0.6035,
+      "step": 19822
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2306908642380444e-06,
+      "loss": 1.1426,
+      "step": 19823
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.229730910512703e-06,
+      "loss": 1.1104,
+      "step": 19824
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.228770997286236e-06,
+      "loss": 1.1084,
+      "step": 19825
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2278111245689495e-06,
+      "loss": 1.4141,
+      "step": 19826
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.226851292371156e-06,
+      "loss": 0.8906,
+      "step": 19827
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.225891500703165e-06,
+      "loss": 0.7175,
+      "step": 19828
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.224931749575283e-06,
+      "loss": 0.853,
+      "step": 19829
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.223972038997823e-06,
+      "loss": 0.8311,
+      "step": 19830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.223012368981093e-06,
+      "loss": 0.8115,
+      "step": 19831
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.222052739535403e-06,
+      "loss": 0.9277,
+      "step": 19832
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.221093150671058e-06,
+      "loss": 1.0244,
+      "step": 19833
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.220133602398365e-06,
+      "loss": 0.8369,
+      "step": 19834
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.219174094727634e-06,
+      "loss": 0.8115,
+      "step": 19835
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2182146276691745e-06,
+      "loss": 0.7783,
+      "step": 19836
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.217255201233286e-06,
+      "loss": 1.1729,
+      "step": 19837
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.216295815430277e-06,
+      "loss": 0.7817,
+      "step": 19838
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.215336470270457e-06,
+      "loss": 0.9893,
+      "step": 19839
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2143771657641205e-06,
+      "loss": 0.792,
+      "step": 19840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.213417901921581e-06,
+      "loss": 1.2627,
+      "step": 19841
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2124586787531415e-06,
+      "loss": 1.0977,
+      "step": 19842
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.211499496269105e-06,
+      "loss": 1.3555,
+      "step": 19843
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.210540354479773e-06,
+      "loss": 0.8799,
+      "step": 19844
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.209581253395447e-06,
+      "loss": 1.1968,
+      "step": 19845
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.208622193026435e-06,
+      "loss": 1.1348,
+      "step": 19846
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.207663173383034e-06,
+      "loss": 0.7559,
+      "step": 19847
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.206704194475548e-06,
+      "loss": 1.0723,
+      "step": 19848
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.205745256314275e-06,
+      "loss": 0.748,
+      "step": 19849
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.204786358909516e-06,
+      "loss": 1.0156,
+      "step": 19850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.203827502271578e-06,
+      "loss": 0.8164,
+      "step": 19851
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.202868686410751e-06,
+      "loss": 1.123,
+      "step": 19852
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.201909911337339e-06,
+      "loss": 1.0576,
+      "step": 19853
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.200951177061643e-06,
+      "loss": 0.9453,
+      "step": 19854
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.199992483593953e-06,
+      "loss": 1.2969,
+      "step": 19855
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1990338309445785e-06,
+      "loss": 0.8389,
+      "step": 19856
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.198075219123809e-06,
+      "loss": 0.9834,
+      "step": 19857
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1971166481419455e-06,
+      "loss": 1.0942,
+      "step": 19858
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196158118009281e-06,
+      "loss": 0.8503,
+      "step": 19859
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.195199628736112e-06,
+      "loss": 1.1855,
+      "step": 19860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.194241180332736e-06,
+      "loss": 1.2832,
+      "step": 19861
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.19328277280945e-06,
+      "loss": 0.9165,
+      "step": 19862
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1923244061765486e-06,
+      "loss": 1.0684,
+      "step": 19863
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.191366080444323e-06,
+      "loss": 1.1836,
+      "step": 19864
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.190407795623066e-06,
+      "loss": 1.1709,
+      "step": 19865
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.189449551723076e-06,
+      "loss": 0.9048,
+      "step": 19866
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.188491348754646e-06,
+      "loss": 1.0107,
+      "step": 19867
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.187533186728065e-06,
+      "loss": 1.0889,
+      "step": 19868
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.186575065653628e-06,
+      "loss": 1.1621,
+      "step": 19869
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.185616985541625e-06,
+      "loss": 1.1113,
+      "step": 19870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.184658946402345e-06,
+      "loss": 1.4053,
+      "step": 19871
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.183700948246085e-06,
+      "loss": 0.8174,
+      "step": 19872
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.182742991083133e-06,
+      "loss": 0.7456,
+      "step": 19873
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.181785074923778e-06,
+      "loss": 0.9565,
+      "step": 19874
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.18082719977831e-06,
+      "loss": 1.4746,
+      "step": 19875
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.179869365657016e-06,
+      "loss": 1.1885,
+      "step": 19876
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.178911572570189e-06,
+      "loss": 0.9688,
+      "step": 19877
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.177953820528117e-06,
+      "loss": 1.2588,
+      "step": 19878
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.176996109541084e-06,
+      "loss": 0.8799,
+      "step": 19879
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.176038439619379e-06,
+      "loss": 1.623,
+      "step": 19880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.175080810773289e-06,
+      "loss": 0.9258,
+      "step": 19881
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.174123223013102e-06,
+      "loss": 0.8721,
+      "step": 19882
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.173165676349103e-06,
+      "loss": 1.1462,
+      "step": 19883
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.172208170791578e-06,
+      "loss": 0.6553,
+      "step": 19884
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.171250706350813e-06,
+      "loss": 0.7256,
+      "step": 19885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.170293283037088e-06,
+      "loss": 0.9336,
+      "step": 19886
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.169335900860691e-06,
+      "loss": 0.8311,
+      "step": 19887
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1683785598319066e-06,
+      "loss": 0.9736,
+      "step": 19888
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.16742125996102e-06,
+      "loss": 1.0923,
+      "step": 19889
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.166464001258307e-06,
+      "loss": 1.4902,
+      "step": 19890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.165506783734055e-06,
+      "loss": 1.3936,
+      "step": 19891
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.164549607398547e-06,
+      "loss": 0.9209,
+      "step": 19892
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.163592472262066e-06,
+      "loss": 1.2178,
+      "step": 19893
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.162635378334887e-06,
+      "loss": 0.8882,
+      "step": 19894
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.161678325627295e-06,
+      "loss": 1.2471,
+      "step": 19895
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.160721314149567e-06,
+      "loss": 0.8823,
+      "step": 19896
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.159764343911988e-06,
+      "loss": 0.8135,
+      "step": 19897
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1588074149248365e-06,
+      "loss": 0.9971,
+      "step": 19898
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1578505271983876e-06,
+      "loss": 0.9619,
+      "step": 19899
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.156893680742924e-06,
+      "loss": 0.8599,
+      "step": 19900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.155936875568721e-06,
+      "loss": 1.1973,
+      "step": 19901
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.154980111686054e-06,
+      "loss": 1.1934,
+      "step": 19902
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.154023389105206e-06,
+      "loss": 0.9883,
+      "step": 19903
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.153066707836453e-06,
+      "loss": 1.0146,
+      "step": 19904
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.152110067890067e-06,
+      "loss": 1.064,
+      "step": 19905
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.151153469276326e-06,
+      "loss": 0.8096,
+      "step": 19906
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.150196912005504e-06,
+      "loss": 1.0361,
+      "step": 19907
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.149240396087881e-06,
+      "loss": 1.2109,
+      "step": 19908
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.148283921533729e-06,
+      "loss": 1.4961,
+      "step": 19909
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.147327488353319e-06,
+      "loss": 0.8975,
+      "step": 19910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.14637109655693e-06,
+      "loss": 0.7829,
+      "step": 19911
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.145414746154825e-06,
+      "loss": 1.3096,
+      "step": 19912
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1444584371572924e-06,
+      "loss": 1.0352,
+      "step": 19913
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.143502169574591e-06,
+      "loss": 0.7847,
+      "step": 19914
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.142545943417e-06,
+      "loss": 0.9717,
+      "step": 19915
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.141589758694791e-06,
+      "loss": 1.0117,
+      "step": 19916
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.140633615418227e-06,
+      "loss": 0.7568,
+      "step": 19917
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.139677513597588e-06,
+      "loss": 0.8604,
+      "step": 19918
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.138721453243139e-06,
+      "loss": 0.8442,
+      "step": 19919
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.137765434365153e-06,
+      "loss": 0.8018,
+      "step": 19920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1368094569738955e-06,
+      "loss": 0.6143,
+      "step": 19921
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.135853521079636e-06,
+      "loss": 1.207,
+      "step": 19922
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.134897626692644e-06,
+      "loss": 1.1348,
+      "step": 19923
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.13394177382319e-06,
+      "loss": 0.8281,
+      "step": 19924
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.132985962481537e-06,
+      "loss": 1.1387,
+      "step": 19925
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.132030192677953e-06,
+      "loss": 1.2178,
+      "step": 19926
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.131074464422702e-06,
+      "loss": 1.0488,
+      "step": 19927
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.130118777726059e-06,
+      "loss": 0.7988,
+      "step": 19928
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.12916313259828e-06,
+      "loss": 0.8945,
+      "step": 19929
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.128207529049634e-06,
+      "loss": 0.9014,
+      "step": 19930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.127251967090391e-06,
+      "loss": 1.2852,
+      "step": 19931
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.126296446730805e-06,
+      "loss": 1.1567,
+      "step": 19932
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.125340967981143e-06,
+      "loss": 0.7754,
+      "step": 19933
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.124385530851673e-06,
+      "loss": 0.9299,
+      "step": 19934
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.123430135352657e-06,
+      "loss": 0.8247,
+      "step": 19935
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.122474781494354e-06,
+      "loss": 0.8633,
+      "step": 19936
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.12151946928703e-06,
+      "loss": 1.0781,
+      "step": 19937
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.120564198740939e-06,
+      "loss": 0.9121,
+      "step": 19938
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.119608969866353e-06,
+      "loss": 1.1016,
+      "step": 19939
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.118653782673526e-06,
+      "loss": 1.0967,
+      "step": 19940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.117698637172718e-06,
+      "loss": 1.0527,
+      "step": 19941
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.116743533374194e-06,
+      "loss": 1.1211,
+      "step": 19942
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.115788471288204e-06,
+      "loss": 1.0801,
+      "step": 19943
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.114833450925017e-06,
+      "loss": 0.9473,
+      "step": 19944
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.113878472294888e-06,
+      "loss": 1.2715,
+      "step": 19945
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.112923535408074e-06,
+      "loss": 0.9854,
+      "step": 19946
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.111968640274832e-06,
+      "loss": 0.7666,
+      "step": 19947
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1110137869054175e-06,
+      "loss": 1.1953,
+      "step": 19948
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.110058975310091e-06,
+      "loss": 0.9814,
+      "step": 19949
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.109104205499111e-06,
+      "loss": 1.084,
+      "step": 19950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.108149477482727e-06,
+      "loss": 0.8643,
+      "step": 19951
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.107194791271197e-06,
+      "loss": 0.9502,
+      "step": 19952
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.106240146874775e-06,
+      "loss": 1.0381,
+      "step": 19953
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.105285544303717e-06,
+      "loss": 1.0273,
+      "step": 19954
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.104330983568278e-06,
+      "loss": 1.3379,
+      "step": 19955
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.10337646467871e-06,
+      "loss": 1.1074,
+      "step": 19956
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.102421987645268e-06,
+      "loss": 1.0215,
+      "step": 19957
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.101467552478195e-06,
+      "loss": 0.8896,
+      "step": 19958
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.100513159187758e-06,
+      "loss": 0.8271,
+      "step": 19959
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.099558807784199e-06,
+      "loss": 0.9883,
+      "step": 19960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.098604498277774e-06,
+      "loss": 1.3672,
+      "step": 19961
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.097650230678731e-06,
+      "loss": 1.5098,
+      "step": 19962
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.09669600499732e-06,
+      "loss": 0.8691,
+      "step": 19963
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.095741821243791e-06,
+      "loss": 0.8223,
+      "step": 19964
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.094787679428397e-06,
+      "loss": 1.124,
+      "step": 19965
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0938335795613854e-06,
+      "loss": 1.4619,
+      "step": 19966
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.092879521653003e-06,
+      "loss": 1.0723,
+      "step": 19967
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0919255057135005e-06,
+      "loss": 1.3594,
+      "step": 19968
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.090971531753118e-06,
+      "loss": 1.2109,
+      "step": 19969
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0900175997821165e-06,
+      "loss": 1.4922,
+      "step": 19970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.089063709810732e-06,
+      "loss": 1.1475,
+      "step": 19971
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.088109861849214e-06,
+      "loss": 1.21,
+      "step": 19972
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.08715605590781e-06,
+      "loss": 1.167,
+      "step": 19973
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0862022919967615e-06,
+      "loss": 0.7441,
+      "step": 19974
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.085248570126317e-06,
+      "loss": 0.998,
+      "step": 19975
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.084294890306719e-06,
+      "loss": 1.0254,
+      "step": 19976
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.083341252548215e-06,
+      "loss": 1.0205,
+      "step": 19977
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.082387656861044e-06,
+      "loss": 1.3564,
+      "step": 19978
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0814341032554495e-06,
+      "loss": 0.9458,
+      "step": 19979
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.080480591741677e-06,
+      "loss": 0.7725,
+      "step": 19980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.079527122329971e-06,
+      "loss": 1.0645,
+      "step": 19981
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.078573695030568e-06,
+      "loss": 0.8379,
+      "step": 19982
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.077620309853711e-06,
+      "loss": 1.0293,
+      "step": 19983
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.076666966809641e-06,
+      "loss": 1.2627,
+      "step": 19984
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.075713665908601e-06,
+      "loss": 1.3242,
+      "step": 19985
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.074760407160828e-06,
+      "loss": 1.0654,
+      "step": 19986
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.073807190576562e-06,
+      "loss": 0.6919,
+      "step": 19987
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.072854016166045e-06,
+      "loss": 0.6483,
+      "step": 19988
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.071900883939505e-06,
+      "loss": 0.8584,
+      "step": 19989
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.070947793907197e-06,
+      "loss": 0.9287,
+      "step": 19990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.069994746079347e-06,
+      "loss": 0.8398,
+      "step": 19991
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0690417404661985e-06,
+      "loss": 1.0393,
+      "step": 19992
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.068088777077983e-06,
+      "loss": 1.209,
+      "step": 19993
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.067135855924937e-06,
+      "loss": 0.7437,
+      "step": 19994
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.066182977017299e-06,
+      "loss": 1.085,
+      "step": 19995
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.065230140365308e-06,
+      "loss": 0.9482,
+      "step": 19996
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.064277345979193e-06,
+      "loss": 0.9414,
+      "step": 19997
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.063324593869189e-06,
+      "loss": 0.6768,
+      "step": 19998
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0623718840455355e-06,
+      "loss": 1.0034,
+      "step": 19999
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.061419216518456e-06,
+      "loss": 0.9512,
+      "step": 20000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060466591298197e-06,
+      "loss": 0.7773,
+      "step": 20001
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.059514008394982e-06,
+      "loss": 1.0908,
+      "step": 20002
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.058561467819048e-06,
+      "loss": 0.6821,
+      "step": 20003
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.057608969580622e-06,
+      "loss": 1.2568,
+      "step": 20004
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0566565136899355e-06,
+      "loss": 0.7222,
+      "step": 20005
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.055704100157225e-06,
+      "loss": 1.251,
+      "step": 20006
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.054751728992717e-06,
+      "loss": 1.1182,
+      "step": 20007
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0537994002066454e-06,
+      "loss": 0.9189,
+      "step": 20008
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.052847113809235e-06,
+      "loss": 1.0952,
+      "step": 20009
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0518948698107125e-06,
+      "loss": 1.1055,
+      "step": 20010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0509426682213155e-06,
+      "loss": 1.21,
+      "step": 20011
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.049990509051269e-06,
+      "loss": 1.209,
+      "step": 20012
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.049038392310795e-06,
+      "loss": 1.1729,
+      "step": 20013
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.048086318010128e-06,
+      "loss": 1.3242,
+      "step": 20014
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.047134286159488e-06,
+      "loss": 1.0664,
+      "step": 20015
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.04618229676911e-06,
+      "loss": 0.8311,
+      "step": 20016
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.045230349849214e-06,
+      "loss": 0.9912,
+      "step": 20017
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.044278445410025e-06,
+      "loss": 1.0518,
+      "step": 20018
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0433265834617725e-06,
+      "loss": 1.0547,
+      "step": 20019
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0423747640146735e-06,
+      "loss": 1.1221,
+      "step": 20020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0414229870789595e-06,
+      "loss": 0.8101,
+      "step": 20021
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.040471252664853e-06,
+      "loss": 1.0176,
+      "step": 20022
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0395195607825765e-06,
+      "loss": 1.0928,
+      "step": 20023
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.038567911442348e-06,
+      "loss": 1.3994,
+      "step": 20024
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.037616304654397e-06,
+      "loss": 0.7549,
+      "step": 20025
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.036664740428937e-06,
+      "loss": 0.9004,
+      "step": 20026
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0357132187762e-06,
+      "loss": 1.166,
+      "step": 20027
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.034761739706399e-06,
+      "loss": 1.0,
+      "step": 20028
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.033810303229757e-06,
+      "loss": 1.0771,
+      "step": 20029
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.032858909356495e-06,
+      "loss": 0.9395,
+      "step": 20030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.031907558096828e-06,
+      "loss": 1.3945,
+      "step": 20031
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0309562494609786e-06,
+      "loss": 1.1855,
+      "step": 20032
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.030004983459167e-06,
+      "loss": 0.9316,
+      "step": 20033
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0290537601016105e-06,
+      "loss": 0.6802,
+      "step": 20034
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.028102579398524e-06,
+      "loss": 0.7632,
+      "step": 20035
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.027151441360124e-06,
+      "loss": 1.0728,
+      "step": 20036
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.026200345996632e-06,
+      "loss": 1.1211,
+      "step": 20037
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0252492933182645e-06,
+      "loss": 0.9966,
+      "step": 20038
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.024298283335233e-06,
+      "loss": 1.2236,
+      "step": 20039
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.023347316057755e-06,
+      "loss": 0.7881,
+      "step": 20040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.022396391496044e-06,
+      "loss": 0.7812,
+      "step": 20041
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.021445509660315e-06,
+      "loss": 1.3613,
+      "step": 20042
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.020494670560788e-06,
+      "loss": 1.5117,
+      "step": 20043
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.019543874207667e-06,
+      "loss": 1.4336,
+      "step": 20044
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.018593120611174e-06,
+      "loss": 0.9097,
+      "step": 20045
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.017642409781509e-06,
+      "loss": 1.209,
+      "step": 20046
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0166917417289e-06,
+      "loss": 0.6978,
+      "step": 20047
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.015741116463547e-06,
+      "loss": 1.0928,
+      "step": 20048
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.014790533995669e-06,
+      "loss": 1.0781,
+      "step": 20049
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.013839994335471e-06,
+      "loss": 1.1602,
+      "step": 20050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.012889497493164e-06,
+      "loss": 1.1934,
+      "step": 20051
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.01193904347896e-06,
+      "loss": 0.8813,
+      "step": 20052
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.010988632303068e-06,
+      "loss": 0.7274,
+      "step": 20053
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0100382639757e-06,
+      "loss": 1.0664,
+      "step": 20054
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.009087938507059e-06,
+      "loss": 1.1973,
+      "step": 20055
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.008137655907358e-06,
+      "loss": 1.0117,
+      "step": 20056
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.007187416186795e-06,
+      "loss": 0.9619,
+      "step": 20057
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.006237219355591e-06,
+      "loss": 0.9004,
+      "step": 20058
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0052870654239424e-06,
+      "loss": 0.7598,
+      "step": 20059
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.004336954402059e-06,
+      "loss": 1.1216,
+      "step": 20060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0033868863001496e-06,
+      "loss": 0.7271,
+      "step": 20061
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.002436861128411e-06,
+      "loss": 0.9678,
+      "step": 20062
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.001486878897056e-06,
+      "loss": 0.8232,
+      "step": 20063
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0005369396162874e-06,
+      "loss": 1.4551,
+      "step": 20064
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.999587043296309e-06,
+      "loss": 1.0498,
+      "step": 20065
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9986371899473205e-06,
+      "loss": 0.9893,
+      "step": 20066
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.997687379579527e-06,
+      "loss": 1.0078,
+      "step": 20067
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.996737612203131e-06,
+      "loss": 0.9678,
+      "step": 20068
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.995787887828339e-06,
+      "loss": 1.0234,
+      "step": 20069
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.994838206465347e-06,
+      "loss": 0.8291,
+      "step": 20070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9938885681243575e-06,
+      "loss": 0.9873,
+      "step": 20071
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.992938972815569e-06,
+      "loss": 0.8315,
+      "step": 20072
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.991989420549189e-06,
+      "loss": 1.0869,
+      "step": 20073
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.99103991133541e-06,
+      "loss": 1.0332,
+      "step": 20074
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.990090445184434e-06,
+      "loss": 1.0542,
+      "step": 20075
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.989141022106462e-06,
+      "loss": 0.7168,
+      "step": 20076
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.988191642111686e-06,
+      "loss": 1.0449,
+      "step": 20077
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.98724230521031e-06,
+      "loss": 1.3711,
+      "step": 20078
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.986293011412529e-06,
+      "loss": 0.9766,
+      "step": 20079
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.985343760728541e-06,
+      "loss": 0.6836,
+      "step": 20080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9843945531685425e-06,
+      "loss": 1.3066,
+      "step": 20081
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.983445388742726e-06,
+      "loss": 0.8477,
+      "step": 20082
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.98249626746129e-06,
+      "loss": 0.667,
+      "step": 20083
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.981547189334433e-06,
+      "loss": 0.9854,
+      "step": 20084
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.980598154372344e-06,
+      "loss": 1.1514,
+      "step": 20085
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.979649162585219e-06,
+      "loss": 0.7456,
+      "step": 20086
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.978700213983255e-06,
+      "loss": 1.1309,
+      "step": 20087
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.977751308576636e-06,
+      "loss": 1.1621,
+      "step": 20088
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.976802446375567e-06,
+      "loss": 1.2754,
+      "step": 20089
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.975853627390233e-06,
+      "loss": 0.7588,
+      "step": 20090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.974904851630829e-06,
+      "loss": 0.8818,
+      "step": 20091
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.973956119107542e-06,
+      "loss": 0.9727,
+      "step": 20092
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.973007429830564e-06,
+      "loss": 1.584,
+      "step": 20093
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9720587838100905e-06,
+      "loss": 0.9961,
+      "step": 20094
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.97111018105631e-06,
+      "loss": 0.8652,
+      "step": 20095
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.970161621579407e-06,
+      "loss": 1.0654,
+      "step": 20096
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.969213105389575e-06,
+      "loss": 0.7002,
+      "step": 20097
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.968264632496998e-06,
+      "loss": 1.0879,
+      "step": 20098
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.967316202911872e-06,
+      "loss": 0.9502,
+      "step": 20099
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.966367816644381e-06,
+      "loss": 0.9961,
+      "step": 20100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.96541947370471e-06,
+      "loss": 1.3247,
+      "step": 20101
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.964471174103048e-06,
+      "loss": 0.96,
+      "step": 20102
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9635229178495755e-06,
+      "loss": 0.75,
+      "step": 20103
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.962574704954489e-06,
+      "loss": 0.9639,
+      "step": 20104
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.961626535427968e-06,
+      "loss": 0.9409,
+      "step": 20105
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.960678409280196e-06,
+      "loss": 1.4336,
+      "step": 20106
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9597303265213615e-06,
+      "loss": 1.2158,
+      "step": 20107
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.958782287161642e-06,
+      "loss": 1.334,
+      "step": 20108
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.957834291211227e-06,
+      "loss": 1.1738,
+      "step": 20109
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.956886338680298e-06,
+      "loss": 0.8242,
+      "step": 20110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.955938429579039e-06,
+      "loss": 1.1621,
+      "step": 20111
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.954990563917628e-06,
+      "loss": 1.0039,
+      "step": 20112
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.954042741706249e-06,
+      "loss": 1.0254,
+      "step": 20113
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.953094962955081e-06,
+      "loss": 1.0576,
+      "step": 20114
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.952147227674311e-06,
+      "loss": 0.9312,
+      "step": 20115
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.951199535874113e-06,
+      "loss": 1.0137,
+      "step": 20116
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.950251887564669e-06,
+      "loss": 0.7715,
+      "step": 20117
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.949304282756159e-06,
+      "loss": 1.0254,
+      "step": 20118
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.948356721458757e-06,
+      "loss": 1.0371,
+      "step": 20119
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.947409203682648e-06,
+      "loss": 0.7222,
+      "step": 20120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.946461729438007e-06,
+      "loss": 1.2812,
+      "step": 20121
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.945514298735014e-06,
+      "loss": 1.0664,
+      "step": 20122
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.94456691158384e-06,
+      "loss": 0.9121,
+      "step": 20123
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.943619567994665e-06,
+      "loss": 1.3101,
+      "step": 20124
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.942672267977665e-06,
+      "loss": 0.8857,
+      "step": 20125
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.941725011543018e-06,
+      "loss": 1.1914,
+      "step": 20126
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.940777798700896e-06,
+      "loss": 0.5811,
+      "step": 20127
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.939830629461474e-06,
+      "loss": 0.9287,
+      "step": 20128
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.938883503834924e-06,
+      "loss": 0.8555,
+      "step": 20129
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.937936421831425e-06,
+      "loss": 0.8081,
+      "step": 20130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.936989383461147e-06,
+      "loss": 1.0391,
+      "step": 20131
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.936042388734262e-06,
+      "loss": 0.9756,
+      "step": 20132
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.935095437660946e-06,
+      "loss": 0.7891,
+      "step": 20133
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934148530251362e-06,
+      "loss": 0.6274,
+      "step": 20134
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.933201666515693e-06,
+      "loss": 0.7275,
+      "step": 20135
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.932254846464103e-06,
+      "loss": 0.8545,
+      "step": 20136
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.931308070106765e-06,
+      "loss": 1.0273,
+      "step": 20137
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.930361337453847e-06,
+      "loss": 1.1445,
+      "step": 20138
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.929414648515517e-06,
+      "loss": 1.2871,
+      "step": 20139
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.928468003301947e-06,
+      "loss": 0.9177,
+      "step": 20140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9275214018233055e-06,
+      "loss": 0.689,
+      "step": 20141
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.926574844089761e-06,
+      "loss": 0.8643,
+      "step": 20142
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.925628330111478e-06,
+      "loss": 1.0332,
+      "step": 20143
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.924681859898628e-06,
+      "loss": 0.9536,
+      "step": 20144
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9237354334613685e-06,
+      "loss": 1.3613,
+      "step": 20145
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.922789050809878e-06,
+      "loss": 0.999,
+      "step": 20146
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.921842711954315e-06,
+      "loss": 0.9028,
+      "step": 20147
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.920896416904849e-06,
+      "loss": 0.8364,
+      "step": 20148
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9199501656716374e-06,
+      "loss": 0.8223,
+      "step": 20149
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.919003958264848e-06,
+      "loss": 0.5806,
+      "step": 20150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.918057794694649e-06,
+      "loss": 0.9722,
+      "step": 20151
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9171116749712e-06,
+      "loss": 0.8896,
+      "step": 20152
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.916165599104667e-06,
+      "loss": 0.916,
+      "step": 20153
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.915219567105205e-06,
+      "loss": 1.002,
+      "step": 20154
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.91427357898298e-06,
+      "loss": 0.9902,
+      "step": 20155
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.913327634748156e-06,
+      "loss": 1.3809,
+      "step": 20156
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.912381734410894e-06,
+      "loss": 1.0063,
+      "step": 20157
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.911435877981351e-06,
+      "loss": 1.2754,
+      "step": 20158
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.910490065469688e-06,
+      "loss": 0.7441,
+      "step": 20159
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.909544296886063e-06,
+      "loss": 1.1738,
+      "step": 20160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.908598572240643e-06,
+      "loss": 1.0918,
+      "step": 20161
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.907652891543576e-06,
+      "loss": 1.123,
+      "step": 20162
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906707254805028e-06,
+      "loss": 1.1338,
+      "step": 20163
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.905761662035155e-06,
+      "loss": 1.043,
+      "step": 20164
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.904816113244107e-06,
+      "loss": 0.9092,
+      "step": 20165
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.903870608442051e-06,
+      "loss": 1.0117,
+      "step": 20166
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.902925147639138e-06,
+      "loss": 1.1367,
+      "step": 20167
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.901979730845527e-06,
+      "loss": 1.2793,
+      "step": 20168
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9010343580713695e-06,
+      "loss": 1.2695,
+      "step": 20169
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.900089029326819e-06,
+      "loss": 0.604,
+      "step": 20170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.899143744622036e-06,
+      "loss": 0.8091,
+      "step": 20171
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.898198503967173e-06,
+      "loss": 0.9258,
+      "step": 20172
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.897253307372379e-06,
+      "loss": 1.0,
+      "step": 20173
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.896308154847811e-06,
+      "loss": 0.7207,
+      "step": 20174
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.895363046403622e-06,
+      "loss": 1.1182,
+      "step": 20175
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8944179820499555e-06,
+      "loss": 1.1006,
+      "step": 20176
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.893472961796974e-06,
+      "loss": 1.0889,
+      "step": 20177
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.892527985654824e-06,
+      "loss": 1.0527,
+      "step": 20178
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8915830536336585e-06,
+      "loss": 1.623,
+      "step": 20179
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.890638165743624e-06,
+      "loss": 1.4219,
+      "step": 20180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.889693321994869e-06,
+      "loss": 1.0181,
+      "step": 20181
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.888748522397547e-06,
+      "loss": 1.1089,
+      "step": 20182
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.887803766961807e-06,
+      "loss": 1.1445,
+      "step": 20183
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.886859055697793e-06,
+      "loss": 1.0928,
+      "step": 20184
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8859143886156565e-06,
+      "loss": 0.8047,
+      "step": 20185
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.88496976572554e-06,
+      "loss": 0.9692,
+      "step": 20186
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.884025187037596e-06,
+      "loss": 1.1934,
+      "step": 20187
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.88308065256197e-06,
+      "loss": 1.4043,
+      "step": 20188
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.882136162308805e-06,
+      "loss": 0.8916,
+      "step": 20189
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.881191716288248e-06,
+      "loss": 1.0928,
+      "step": 20190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8802473145104385e-06,
+      "loss": 1.2451,
+      "step": 20191
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.879302956985531e-06,
+      "loss": 1.168,
+      "step": 20192
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.878358643723664e-06,
+      "loss": 1.2773,
+      "step": 20193
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.877414374734982e-06,
+      "loss": 1.1484,
+      "step": 20194
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.876470150029627e-06,
+      "loss": 1.2559,
+      "step": 20195
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.875525969617737e-06,
+      "loss": 1.0068,
+      "step": 20196
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8745818335094625e-06,
+      "loss": 1.1123,
+      "step": 20197
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.873637741714941e-06,
+      "loss": 1.4355,
+      "step": 20198
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.872693694244318e-06,
+      "loss": 1.1875,
+      "step": 20199
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8717496911077265e-06,
+      "loss": 0.9023,
+      "step": 20200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870805732315308e-06,
+      "loss": 0.918,
+      "step": 20201
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.869861817877207e-06,
+      "loss": 0.854,
+      "step": 20202
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.868917947803563e-06,
+      "loss": 1.0518,
+      "step": 20203
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.86797412210451e-06,
+      "loss": 1.0459,
+      "step": 20204
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8670303407901854e-06,
+      "loss": 0.9443,
+      "step": 20205
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.866086603870733e-06,
+      "loss": 1.0986,
+      "step": 20206
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.865142911356284e-06,
+      "loss": 0.8574,
+      "step": 20207
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.864199263256979e-06,
+      "loss": 1.1699,
+      "step": 20208
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.863255659582954e-06,
+      "loss": 0.8701,
+      "step": 20209
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.862312100344345e-06,
+      "loss": 1.1504,
+      "step": 20210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.861368585551286e-06,
+      "loss": 0.8262,
+      "step": 20211
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.860425115213908e-06,
+      "loss": 0.9141,
+      "step": 20212
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.859481689342354e-06,
+      "loss": 1.2148,
+      "step": 20213
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.858538307946755e-06,
+      "loss": 1.2178,
+      "step": 20214
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.857594971037242e-06,
+      "loss": 1.0176,
+      "step": 20215
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.856651678623948e-06,
+      "loss": 1.3877,
+      "step": 20216
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.855708430717004e-06,
+      "loss": 0.8447,
+      "step": 20217
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854765227326551e-06,
+      "loss": 0.668,
+      "step": 20218
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.853822068462711e-06,
+      "loss": 0.7778,
+      "step": 20219
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852878954135619e-06,
+      "loss": 1.2256,
+      "step": 20220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8519358843554065e-06,
+      "loss": 0.8857,
+      "step": 20221
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.850992859132197e-06,
+      "loss": 0.9209,
+      "step": 20222
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.85004987847613e-06,
+      "loss": 1.5098,
+      "step": 20223
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.849106942397327e-06,
+      "loss": 0.8125,
+      "step": 20224
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.848164050905922e-06,
+      "loss": 1.3008,
+      "step": 20225
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8472212040120405e-06,
+      "loss": 0.8213,
+      "step": 20226
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.846278401725807e-06,
+      "loss": 0.8525,
+      "step": 20227
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.845335644057355e-06,
+      "loss": 0.8535,
+      "step": 20228
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.844392931016809e-06,
+      "loss": 0.96,
+      "step": 20229
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.843450262614294e-06,
+      "loss": 0.9219,
+      "step": 20230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.842507638859938e-06,
+      "loss": 0.7891,
+      "step": 20231
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.841565059763861e-06,
+      "loss": 0.9248,
+      "step": 20232
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.840622525336194e-06,
+      "loss": 0.9287,
+      "step": 20233
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.839680035587061e-06,
+      "loss": 0.7395,
+      "step": 20234
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.838737590526582e-06,
+      "loss": 0.5977,
+      "step": 20235
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.837795190164884e-06,
+      "loss": 1.2129,
+      "step": 20236
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.836852834512085e-06,
+      "loss": 1.043,
+      "step": 20237
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8359105235783105e-06,
+      "loss": 1.1855,
+      "step": 20238
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.834968257373688e-06,
+      "loss": 0.9448,
+      "step": 20239
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.834026035908331e-06,
+      "loss": 0.9395,
+      "step": 20240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833083859192361e-06,
+      "loss": 1.1445,
+      "step": 20241
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8321417272359045e-06,
+      "loss": 0.9453,
+      "step": 20242
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.831199640049074e-06,
+      "loss": 1.0156,
+      "step": 20243
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.830257597641996e-06,
+      "loss": 0.915,
+      "step": 20244
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.829315600024783e-06,
+      "loss": 0.999,
+      "step": 20245
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.828373647207562e-06,
+      "loss": 1.04,
+      "step": 20246
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.827431739200444e-06,
+      "loss": 1.2705,
+      "step": 20247
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.826489876013546e-06,
+      "loss": 0.6846,
+      "step": 20248
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.825548057656988e-06,
+      "loss": 1.375,
+      "step": 20249
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.82460628414089e-06,
+      "loss": 1.1074,
+      "step": 20250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.823664555475366e-06,
+      "loss": 1.1553,
+      "step": 20251
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8227228716705255e-06,
+      "loss": 0.9873,
+      "step": 20252
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.821781232736488e-06,
+      "loss": 0.8701,
+      "step": 20253
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.820839638683373e-06,
+      "loss": 1.1279,
+      "step": 20254
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.819898089521292e-06,
+      "loss": 0.9961,
+      "step": 20255
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8189565852603534e-06,
+      "loss": 0.9141,
+      "step": 20256
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.818015125910679e-06,
+      "loss": 0.6763,
+      "step": 20257
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8170737114823715e-06,
+      "loss": 1.1289,
+      "step": 20258
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.816132341985555e-06,
+      "loss": 0.9561,
+      "step": 20259
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.81519101743033e-06,
+      "loss": 1.1963,
+      "step": 20260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.814249737826816e-06,
+      "loss": 0.9351,
+      "step": 20261
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.813308503185122e-06,
+      "loss": 0.8682,
+      "step": 20262
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.812367313515354e-06,
+      "loss": 1.2793,
+      "step": 20263
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.811426168827627e-06,
+      "loss": 0.9961,
+      "step": 20264
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.81048506913205e-06,
+      "loss": 1.3711,
+      "step": 20265
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.809544014438732e-06,
+      "loss": 1.2285,
+      "step": 20266
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.808603004757776e-06,
+      "loss": 0.9453,
+      "step": 20267
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8076620400992975e-06,
+      "loss": 0.8145,
+      "step": 20268
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8067211204733984e-06,
+      "loss": 1.0586,
+      "step": 20269
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.805780245890192e-06,
+      "loss": 1.2988,
+      "step": 20270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.804839416359775e-06,
+      "loss": 1.29,
+      "step": 20271
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.803898631892266e-06,
+      "loss": 1.1201,
+      "step": 20272
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.802957892497762e-06,
+      "loss": 0.4795,
+      "step": 20273
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.802017198186366e-06,
+      "loss": 0.7451,
+      "step": 20274
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.801076548968188e-06,
+      "loss": 1.1152,
+      "step": 20275
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.800135944853332e-06,
+      "loss": 0.71,
+      "step": 20276
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7991953858519025e-06,
+      "loss": 0.7822,
+      "step": 20277
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.798254871973996e-06,
+      "loss": 1.041,
+      "step": 20278
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.797314403229719e-06,
+      "loss": 1.1108,
+      "step": 20279
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.796373979629179e-06,
+      "loss": 1.3281,
+      "step": 20280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7954336011824744e-06,
+      "loss": 0.9121,
+      "step": 20281
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.794493267899699e-06,
+      "loss": 0.7896,
+      "step": 20282
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.793552979790964e-06,
+      "loss": 1.0645,
+      "step": 20283
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.792612736866361e-06,
+      "loss": 1.0459,
+      "step": 20284
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.791672539135998e-06,
+      "loss": 1.2598,
+      "step": 20285
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.790732386609966e-06,
+      "loss": 0.7617,
+      "step": 20286
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.789792279298373e-06,
+      "loss": 1.0112,
+      "step": 20287
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.788852217211311e-06,
+      "loss": 1.1357,
+      "step": 20288
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.787912200358876e-06,
+      "loss": 1.1426,
+      "step": 20289
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.786972228751169e-06,
+      "loss": 1.2939,
+      "step": 20290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.786032302398287e-06,
+      "loss": 1.0195,
+      "step": 20291
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.78509242131033e-06,
+      "loss": 0.8271,
+      "step": 20292
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.784152585497383e-06,
+      "loss": 1.2305,
+      "step": 20293
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.783212794969548e-06,
+      "loss": 0.5186,
+      "step": 20294
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.782273049736924e-06,
+      "loss": 0.8799,
+      "step": 20295
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.781333349809601e-06,
+      "loss": 1.3066,
+      "step": 20296
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.780393695197668e-06,
+      "loss": 0.7104,
+      "step": 20297
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.779454085911229e-06,
+      "loss": 1.0225,
+      "step": 20298
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.77851452196037e-06,
+      "loss": 0.668,
+      "step": 20299
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777575003355178e-06,
+      "loss": 1.0996,
+      "step": 20300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7766355301057595e-06,
+      "loss": 0.9795,
+      "step": 20301
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.775696102222199e-06,
+      "loss": 1.5684,
+      "step": 20302
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7747567197145845e-06,
+      "loss": 0.8457,
+      "step": 20303
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.773817382593008e-06,
+      "loss": 1.0186,
+      "step": 20304
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7728780908675576e-06,
+      "loss": 0.9277,
+      "step": 20305
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.771938844548331e-06,
+      "loss": 0.9258,
+      "step": 20306
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.770999643645412e-06,
+      "loss": 0.8516,
+      "step": 20307
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.770060488168885e-06,
+      "loss": 0.9805,
+      "step": 20308
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7691213781288445e-06,
+      "loss": 1.25,
+      "step": 20309
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.768182313535373e-06,
+      "loss": 0.8981,
+      "step": 20310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.767243294398559e-06,
+      "loss": 0.7817,
+      "step": 20311
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7663043207284955e-06,
+      "loss": 1.1248,
+      "step": 20312
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.765365392535263e-06,
+      "loss": 1.1602,
+      "step": 20313
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7644265098289485e-06,
+      "loss": 0.9424,
+      "step": 20314
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.763487672619628e-06,
+      "loss": 0.8799,
+      "step": 20315
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7625488809174e-06,
+      "loss": 1.4326,
+      "step": 20316
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.761610134732345e-06,
+      "loss": 0.873,
+      "step": 20317
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.760671434074541e-06,
+      "loss": 0.7524,
+      "step": 20318
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.759732778954078e-06,
+      "loss": 0.5991,
+      "step": 20319
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.758794169381032e-06,
+      "loss": 1.0107,
+      "step": 20320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.757855605365492e-06,
+      "loss": 0.9746,
+      "step": 20321
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.756917086917533e-06,
+      "loss": 1.2061,
+      "step": 20322
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.755978614047244e-06,
+      "loss": 0.7593,
+      "step": 20323
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.755040186764702e-06,
+      "loss": 0.7988,
+      "step": 20324
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.754101805079981e-06,
+      "loss": 1.082,
+      "step": 20325
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.753163469003168e-06,
+      "loss": 0.9121,
+      "step": 20326
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.752225178544344e-06,
+      "loss": 0.9873,
+      "step": 20327
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.751286933713585e-06,
+      "loss": 0.9209,
+      "step": 20328
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.750348734520964e-06,
+      "loss": 1.1465,
+      "step": 20329
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.749410580976569e-06,
+      "loss": 0.9995,
+      "step": 20330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.748472473090466e-06,
+      "loss": 1.4727,
+      "step": 20331
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.747534410872743e-06,
+      "loss": 0.8408,
+      "step": 20332
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7465963943334654e-06,
+      "loss": 1.3711,
+      "step": 20333
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.74565842348272e-06,
+      "loss": 0.6016,
+      "step": 20334
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.744720498330578e-06,
+      "loss": 0.8154,
+      "step": 20335
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.743782618887108e-06,
+      "loss": 1.0156,
+      "step": 20336
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.742844785162389e-06,
+      "loss": 0.9727,
+      "step": 20337
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7419069971664995e-06,
+      "loss": 1.3887,
+      "step": 20338
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.740969254909509e-06,
+      "loss": 0.9072,
+      "step": 20339
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.740031558401486e-06,
+      "loss": 1.248,
+      "step": 20340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.739093907652507e-06,
+      "loss": 1.2031,
+      "step": 20341
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.738156302672646e-06,
+      "loss": 1.4961,
+      "step": 20342
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.737218743471975e-06,
+      "loss": 0.7432,
+      "step": 20343
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.736281230060556e-06,
+      "loss": 1.0918,
+      "step": 20344
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.735343762448471e-06,
+      "loss": 1.1318,
+      "step": 20345
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.734406340645781e-06,
+      "loss": 1.0127,
+      "step": 20346
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.733468964662562e-06,
+      "loss": 0.9902,
+      "step": 20347
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.732531634508875e-06,
+      "loss": 1.0088,
+      "step": 20348
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7315943501948e-06,
+      "loss": 0.7979,
+      "step": 20349
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.730657111730397e-06,
+      "loss": 0.7891,
+      "step": 20350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.729719919125731e-06,
+      "loss": 1.0586,
+      "step": 20351
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7287827723908725e-06,
+      "loss": 1.0366,
+      "step": 20352
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.727845671535895e-06,
+      "loss": 1.1992,
+      "step": 20353
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7269086165708546e-06,
+      "loss": 1.0396,
+      "step": 20354
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.725971607505817e-06,
+      "loss": 1.1611,
+      "step": 20355
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7250346443508505e-06,
+      "loss": 0.959,
+      "step": 20356
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.724097727116022e-06,
+      "loss": 1.2959,
+      "step": 20357
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.723160855811396e-06,
+      "loss": 0.9805,
+      "step": 20358
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.722224030447027e-06,
+      "loss": 1.0654,
+      "step": 20359
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.721287251032988e-06,
+      "loss": 0.7988,
+      "step": 20360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.720350517579339e-06,
+      "loss": 1.1572,
+      "step": 20361
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.719413830096134e-06,
+      "loss": 1.0376,
+      "step": 20362
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.718477188593443e-06,
+      "loss": 1.0615,
+      "step": 20363
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.717540593081329e-06,
+      "loss": 0.9858,
+      "step": 20364
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.716604043569849e-06,
+      "loss": 0.9932,
+      "step": 20365
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.715667540069058e-06,
+      "loss": 0.8672,
+      "step": 20366
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71473108258902e-06,
+      "loss": 1.1729,
+      "step": 20367
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7137946711398e-06,
+      "loss": 0.876,
+      "step": 20368
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.712858305731451e-06,
+      "loss": 0.8765,
+      "step": 20369
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.711921986374027e-06,
+      "loss": 1.0537,
+      "step": 20370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7109857130775935e-06,
+      "loss": 0.749,
+      "step": 20371
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7100494858522e-06,
+      "loss": 0.9692,
+      "step": 20372
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.709113304707912e-06,
+      "loss": 0.9189,
+      "step": 20373
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.708177169654776e-06,
+      "loss": 0.9932,
+      "step": 20374
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.707241080702857e-06,
+      "loss": 0.9932,
+      "step": 20375
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.706305037862203e-06,
+      "loss": 1.1504,
+      "step": 20376
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.705369041142869e-06,
+      "loss": 1.084,
+      "step": 20377
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.704433090554912e-06,
+      "loss": 0.6157,
+      "step": 20378
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7034971861083875e-06,
+      "loss": 1.2925,
+      "step": 20379
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7025613278133476e-06,
+      "loss": 0.959,
+      "step": 20380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.701625515679837e-06,
+      "loss": 0.79,
+      "step": 20381
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.700689749717916e-06,
+      "loss": 0.9062,
+      "step": 20382
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6997540299376376e-06,
+      "loss": 0.9805,
+      "step": 20383
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.698818356349049e-06,
+      "loss": 0.855,
+      "step": 20384
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6978827289622e-06,
+      "loss": 0.918,
+      "step": 20385
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.696947147787145e-06,
+      "loss": 0.6074,
+      "step": 20386
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.696011612833927e-06,
+      "loss": 0.8623,
+      "step": 20387
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.695076124112604e-06,
+      "loss": 0.77,
+      "step": 20388
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6941406816332155e-06,
+      "loss": 1.3545,
+      "step": 20389
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.69320528540582e-06,
+      "loss": 1.3359,
+      "step": 20390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6922699354404576e-06,
+      "loss": 1.0928,
+      "step": 20391
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.691334631747174e-06,
+      "loss": 0.7954,
+      "step": 20392
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.690399374336018e-06,
+      "loss": 0.7754,
+      "step": 20393
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.689464163217042e-06,
+      "loss": 0.8643,
+      "step": 20394
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.688528998400287e-06,
+      "loss": 0.8599,
+      "step": 20395
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.687593879895793e-06,
+      "loss": 1.0488,
+      "step": 20396
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6866588077136145e-06,
+      "loss": 1.2363,
+      "step": 20397
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.685723781863786e-06,
+      "loss": 0.9189,
+      "step": 20398
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.684788802356356e-06,
+      "loss": 0.7036,
+      "step": 20399
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.683853869201372e-06,
+      "loss": 1.1245,
+      "step": 20400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.682918982408873e-06,
+      "loss": 1.2441,
+      "step": 20401
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6819841419889e-06,
+      "loss": 1.3574,
+      "step": 20402
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681049347951487e-06,
+      "loss": 0.9346,
+      "step": 20403
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.680114600306692e-06,
+      "loss": 0.9932,
+      "step": 20404
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.679179899064547e-06,
+      "loss": 1.1885,
+      "step": 20405
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6782452442350935e-06,
+      "loss": 1.0947,
+      "step": 20406
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.677310635828366e-06,
+      "loss": 1.2002,
+      "step": 20407
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6763760738544075e-06,
+      "loss": 1.2021,
+      "step": 20408
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.675441558323262e-06,
+      "loss": 0.7026,
+      "step": 20409
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.674507089244957e-06,
+      "loss": 1.0527,
+      "step": 20410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.673572666629542e-06,
+      "loss": 0.5762,
+      "step": 20411
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.672638290487048e-06,
+      "loss": 1.1982,
+      "step": 20412
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.671703960827508e-06,
+      "loss": 1.1553,
+      "step": 20413
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.670769677660962e-06,
+      "loss": 1.1016,
+      "step": 20414
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.669835440997451e-06,
+      "loss": 1.0166,
+      "step": 20415
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.668901250847004e-06,
+      "loss": 1.1201,
+      "step": 20416
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.667967107219654e-06,
+      "loss": 1.2588,
+      "step": 20417
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.667033010125442e-06,
+      "loss": 0.894,
+      "step": 20418
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.666098959574395e-06,
+      "loss": 1.209,
+      "step": 20419
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.665164955576552e-06,
+      "loss": 0.8428,
+      "step": 20420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.66423099814194e-06,
+      "loss": 1.042,
+      "step": 20421
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.663297087280597e-06,
+      "loss": 0.9238,
+      "step": 20422
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.662363223002554e-06,
+      "loss": 0.6909,
+      "step": 20423
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.661429405317834e-06,
+      "loss": 1.1152,
+      "step": 20424
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.660495634236475e-06,
+      "loss": 1.0986,
+      "step": 20425
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.659561909768509e-06,
+      "loss": 1.082,
+      "step": 20426
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.658628231923965e-06,
+      "loss": 1.3584,
+      "step": 20427
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.657694600712864e-06,
+      "loss": 1.1465,
+      "step": 20428
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.656761016145242e-06,
+      "loss": 0.7905,
+      "step": 20429
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.655827478231131e-06,
+      "loss": 0.6106,
+      "step": 20430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.654893986980552e-06,
+      "loss": 0.7666,
+      "step": 20431
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.653960542403531e-06,
+      "loss": 1.0889,
+      "step": 20432
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.653027144510101e-06,
+      "loss": 0.7939,
+      "step": 20433
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.652093793310283e-06,
+      "loss": 1.3086,
+      "step": 20434
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.651160488814107e-06,
+      "loss": 0.9062,
+      "step": 20435
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.650227231031592e-06,
+      "loss": 0.7832,
+      "step": 20436
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.649294019972771e-06,
+      "loss": 0.998,
+      "step": 20437
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.648360855647663e-06,
+      "loss": 0.8633,
+      "step": 20438
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6474277380662904e-06,
+      "loss": 0.9712,
+      "step": 20439
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.646494667238678e-06,
+      "loss": 0.9932,
+      "step": 20440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.645561643174852e-06,
+      "loss": 0.8633,
+      "step": 20441
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.644628665884833e-06,
+      "loss": 1.5527,
+      "step": 20442
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.643695735378636e-06,
+      "loss": 0.9941,
+      "step": 20443
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.642762851666288e-06,
+      "loss": 0.7354,
+      "step": 20444
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.641830014757814e-06,
+      "loss": 0.9326,
+      "step": 20445
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.640897224663229e-06,
+      "loss": 0.9521,
+      "step": 20446
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.63996448139255e-06,
+      "loss": 0.9673,
+      "step": 20447
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.639031784955802e-06,
+      "loss": 0.9932,
+      "step": 20448
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638099135363002e-06,
+      "loss": 1.2031,
+      "step": 20449
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6371665326241635e-06,
+      "loss": 1.0464,
+      "step": 20450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.636233976749308e-06,
+      "loss": 0.7285,
+      "step": 20451
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6353014677484565e-06,
+      "loss": 0.7373,
+      "step": 20452
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.634369005631621e-06,
+      "loss": 1.1621,
+      "step": 20453
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6334365904088155e-06,
+      "loss": 0.8257,
+      "step": 20454
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.632504222090058e-06,
+      "loss": 1.2012,
+      "step": 20455
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.631571900685369e-06,
+      "loss": 0.9514,
+      "step": 20456
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.630639626204757e-06,
+      "loss": 1.1631,
+      "step": 20457
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.629707398658235e-06,
+      "loss": 1.0615,
+      "step": 20458
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628775218055823e-06,
+      "loss": 1.0674,
+      "step": 20459
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.627843084407527e-06,
+      "loss": 1.0391,
+      "step": 20460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.626910997723366e-06,
+      "loss": 1.1797,
+      "step": 20461
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.625978958013345e-06,
+      "loss": 0.8848,
+      "step": 20462
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.625046965287486e-06,
+      "loss": 1.248,
+      "step": 20463
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.624115019555791e-06,
+      "loss": 1.3438,
+      "step": 20464
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.623183120828272e-06,
+      "loss": 0.7534,
+      "step": 20465
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.62225126911494e-06,
+      "loss": 0.7266,
+      "step": 20466
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6213194644258096e-06,
+      "loss": 0.5732,
+      "step": 20467
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.620387706770884e-06,
+      "loss": 0.8877,
+      "step": 20468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.619455996160172e-06,
+      "loss": 0.9609,
+      "step": 20469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.618524332603681e-06,
+      "loss": 0.9326,
+      "step": 20470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.617592716111426e-06,
+      "loss": 1.2637,
+      "step": 20471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.616661146693408e-06,
+      "loss": 1.1865,
+      "step": 20472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.615729624359632e-06,
+      "loss": 1.2637,
+      "step": 20473
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.614798149120108e-06,
+      "loss": 1.1768,
+      "step": 20474
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.613866720984838e-06,
+      "loss": 1.085,
+      "step": 20475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.612935339963832e-06,
+      "loss": 0.8628,
+      "step": 20476
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6120040060670885e-06,
+      "loss": 1.1895,
+      "step": 20477
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611072719304618e-06,
+      "loss": 0.9707,
+      "step": 20478
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.610141479686422e-06,
+      "loss": 1.5332,
+      "step": 20479
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.609210287222497e-06,
+      "loss": 1.2031,
+      "step": 20480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.608279141922852e-06,
+      "loss": 1.1152,
+      "step": 20481
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.607348043797491e-06,
+      "loss": 0.8806,
+      "step": 20482
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.606416992856413e-06,
+      "loss": 0.8604,
+      "step": 20483
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6054859891096135e-06,
+      "loss": 0.9727,
+      "step": 20484
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.604555032567103e-06,
+      "loss": 1.0122,
+      "step": 20485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.603624123238871e-06,
+      "loss": 0.917,
+      "step": 20486
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.602693261134928e-06,
+      "loss": 1.1221,
+      "step": 20487
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6017624462652634e-06,
+      "loss": 0.9648,
+      "step": 20488
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.600831678639883e-06,
+      "loss": 0.8164,
+      "step": 20489
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.599900958268781e-06,
+      "loss": 0.8809,
+      "step": 20490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.598970285161949e-06,
+      "loss": 0.6162,
+      "step": 20491
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.598039659329396e-06,
+      "loss": 1.3203,
+      "step": 20492
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5971090807811135e-06,
+      "loss": 1.2305,
+      "step": 20493
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.596178549527098e-06,
+      "loss": 0.7915,
+      "step": 20494
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.595248065577339e-06,
+      "loss": 0.7056,
+      "step": 20495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5943176289418345e-06,
+      "loss": 0.9854,
+      "step": 20496
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.593387239630585e-06,
+      "loss": 0.8711,
+      "step": 20497
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5924568976535775e-06,
+      "loss": 1.1377,
+      "step": 20498
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5915266030208105e-06,
+      "loss": 0.6558,
+      "step": 20499
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5905963557422735e-06,
+      "loss": 0.9277,
+      "step": 20500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.589666155827957e-06,
+      "loss": 0.5317,
+      "step": 20501
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.588736003287854e-06,
+      "loss": 1.3428,
+      "step": 20502
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.587805898131962e-06,
+      "loss": 1.0674,
+      "step": 20503
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.586875840370267e-06,
+      "loss": 0.7788,
+      "step": 20504
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.58594583001276e-06,
+      "loss": 0.9639,
+      "step": 20505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.585015867069421e-06,
+      "loss": 1.0986,
+      "step": 20506
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.584085951550258e-06,
+      "loss": 1.0645,
+      "step": 20507
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.583156083465248e-06,
+      "loss": 0.749,
+      "step": 20508
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.582226262824379e-06,
+      "loss": 1.1006,
+      "step": 20509
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.581296489637646e-06,
+      "loss": 1.0771,
+      "step": 20510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5803667639150325e-06,
+      "loss": 0.9458,
+      "step": 20511
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.579437085666519e-06,
+      "loss": 1.1035,
+      "step": 20512
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.578507454902098e-06,
+      "loss": 1.46,
+      "step": 20513
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.577577871631758e-06,
+      "loss": 1.0039,
+      "step": 20514
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.576648335865482e-06,
+      "loss": 0.9023,
+      "step": 20515
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5757188476132494e-06,
+      "loss": 0.7207,
+      "step": 20516
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.574789406885048e-06,
+      "loss": 1.0176,
+      "step": 20517
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.573860013690867e-06,
+      "loss": 1.1455,
+      "step": 20518
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.572930668040683e-06,
+      "loss": 1.0059,
+      "step": 20519
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.572001369944479e-06,
+      "loss": 1.251,
+      "step": 20520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.571072119412242e-06,
+      "loss": 0.9912,
+      "step": 20521
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.570142916453944e-06,
+      "loss": 0.8479,
+      "step": 20522
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.569213761079579e-06,
+      "loss": 0.8066,
+      "step": 20523
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.568284653299117e-06,
+      "loss": 0.8799,
+      "step": 20524
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.567355593122544e-06,
+      "loss": 1.4941,
+      "step": 20525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.56642658055984e-06,
+      "loss": 0.9277,
+      "step": 20526
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5654976156209775e-06,
+      "loss": 0.9014,
+      "step": 20527
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.564568698315939e-06,
+      "loss": 1.3672,
+      "step": 20528
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.563639828654708e-06,
+      "loss": 0.9258,
+      "step": 20529
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5627110066472564e-06,
+      "loss": 0.9512,
+      "step": 20530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.561782232303559e-06,
+      "loss": 0.665,
+      "step": 20531
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.560853505633594e-06,
+      "loss": 0.8945,
+      "step": 20532
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.559924826647345e-06,
+      "loss": 0.6418,
+      "step": 20533
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55899619535478e-06,
+      "loss": 0.8584,
+      "step": 20534
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5580676117658725e-06,
+      "loss": 0.895,
+      "step": 20535
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.557139075890603e-06,
+      "loss": 1.3701,
+      "step": 20536
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5562105877389395e-06,
+      "loss": 0.7549,
+      "step": 20537
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.555282147320862e-06,
+      "loss": 0.999,
+      "step": 20538
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.554353754646336e-06,
+      "loss": 1.002,
+      "step": 20539
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5534254097253425e-06,
+      "loss": 0.8267,
+      "step": 20540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5524971125678495e-06,
+      "loss": 1.21,
+      "step": 20541
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.551568863183823e-06,
+      "loss": 1.042,
+      "step": 20542
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.550640661583238e-06,
+      "loss": 1.1748,
+      "step": 20543
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.54971250777607e-06,
+      "loss": 0.9282,
+      "step": 20544
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.548784401772285e-06,
+      "loss": 0.6523,
+      "step": 20545
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.547856343581848e-06,
+      "loss": 1.1016,
+      "step": 20546
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.546928333214735e-06,
+      "loss": 1.1357,
+      "step": 20547
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.546000370680907e-06,
+      "loss": 0.8672,
+      "step": 20548
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.545072455990341e-06,
+      "loss": 0.9424,
+      "step": 20549
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.544144589152995e-06,
+      "loss": 1.2881,
+      "step": 20550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.543216770178843e-06,
+      "loss": 0.8608,
+      "step": 20551
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.54228899907785e-06,
+      "loss": 1.0049,
+      "step": 20552
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.541361275859976e-06,
+      "loss": 1.0952,
+      "step": 20553
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5404336005351896e-06,
+      "loss": 0.6558,
+      "step": 20554
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.539505973113461e-06,
+      "loss": 1.1953,
+      "step": 20555
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.53857839360475e-06,
+      "loss": 1.541,
+      "step": 20556
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.537650862019015e-06,
+      "loss": 1.0186,
+      "step": 20557
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.536723378366226e-06,
+      "loss": 0.5786,
+      "step": 20558
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5357959426563475e-06,
+      "loss": 0.8989,
+      "step": 20559
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5348685548993374e-06,
+      "loss": 0.7607,
+      "step": 20560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5339412151051555e-06,
+      "loss": 0.8389,
+      "step": 20561
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.53301392328377e-06,
+      "loss": 0.6665,
+      "step": 20562
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.532086679445132e-06,
+      "loss": 0.7568,
+      "step": 20563
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.531159483599211e-06,
+      "loss": 1.1201,
+      "step": 20564
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.530232335755958e-06,
+      "loss": 1.3965,
+      "step": 20565
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.529305235925342e-06,
+      "loss": 0.9102,
+      "step": 20566
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.528378184117315e-06,
+      "loss": 1.3662,
+      "step": 20567
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5274511803418315e-06,
+      "loss": 0.7744,
+      "step": 20568
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.526524224608856e-06,
+      "loss": 1.1904,
+      "step": 20569
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.525597316928344e-06,
+      "loss": 1.4043,
+      "step": 20570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5246704573102535e-06,
+      "loss": 0.7715,
+      "step": 20571
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5237436457645345e-06,
+      "loss": 0.9902,
+      "step": 20572
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.522816882301149e-06,
+      "loss": 0.9954,
+      "step": 20573
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.521890166930045e-06,
+      "loss": 1.4004,
+      "step": 20574
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520963499661186e-06,
+      "loss": 1.0132,
+      "step": 20575
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520036880504515e-06,
+      "loss": 1.1152,
+      "step": 20576
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5191103094699985e-06,
+      "loss": 1.1211,
+      "step": 20577
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.518183786567582e-06,
+      "loss": 0.9561,
+      "step": 20578
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.517257311807212e-06,
+      "loss": 1.2451,
+      "step": 20579
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.516330885198847e-06,
+      "loss": 0.6978,
+      "step": 20580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.515404506752443e-06,
+      "loss": 1.0693,
+      "step": 20581
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.514478176477944e-06,
+      "loss": 1.0674,
+      "step": 20582
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5135518943853e-06,
+      "loss": 1.2246,
+      "step": 20583
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.512625660484463e-06,
+      "loss": 0.5703,
+      "step": 20584
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.511699474785384e-06,
+      "loss": 1.0146,
+      "step": 20585
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.510773337298011e-06,
+      "loss": 1.0732,
+      "step": 20586
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.509847248032288e-06,
+      "loss": 0.9004,
+      "step": 20587
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.508921206998168e-06,
+      "loss": 1.0732,
+      "step": 20588
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.507995214205593e-06,
+      "loss": 0.9941,
+      "step": 20589
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.507069269664512e-06,
+      "loss": 0.9839,
+      "step": 20590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.506143373384876e-06,
+      "loss": 1.083,
+      "step": 20591
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5052175253766275e-06,
+      "loss": 0.7324,
+      "step": 20592
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5042917256497085e-06,
+      "loss": 0.6641,
+      "step": 20593
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.503365974214059e-06,
+      "loss": 1.0117,
+      "step": 20594
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.502440271079639e-06,
+      "loss": 1.0508,
+      "step": 20595
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.50151461625638e-06,
+      "loss": 1.0254,
+      "step": 20596
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.50058900975423e-06,
+      "loss": 1.0957,
+      "step": 20597
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.499663451583125e-06,
+      "loss": 1.0586,
+      "step": 20598
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.498737941753012e-06,
+      "loss": 1.6797,
+      "step": 20599
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.497812480273834e-06,
+      "loss": 1.1562,
+      "step": 20600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4968870671555276e-06,
+      "loss": 0.8809,
+      "step": 20601
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.495961702408039e-06,
+      "loss": 1.2305,
+      "step": 20602
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.495036386041306e-06,
+      "loss": 1.0322,
+      "step": 20603
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.494111118065264e-06,
+      "loss": 0.9106,
+      "step": 20604
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.493185898489849e-06,
+      "loss": 1.2168,
+      "step": 20605
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.492260727325012e-06,
+      "loss": 0.9922,
+      "step": 20606
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4913356045806845e-06,
+      "loss": 0.9023,
+      "step": 20607
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.490410530266798e-06,
+      "loss": 1.2363,
+      "step": 20608
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.489485504393299e-06,
+      "loss": 0.6217,
+      "step": 20609
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.488560526970116e-06,
+      "loss": 0.9404,
+      "step": 20610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.487635598007191e-06,
+      "loss": 1.332,
+      "step": 20611
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.486710717514453e-06,
+      "loss": 0.729,
+      "step": 20612
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4857858855018446e-06,
+      "loss": 1.1719,
+      "step": 20613
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.484861101979294e-06,
+      "loss": 0.6958,
+      "step": 20614
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.483936366956732e-06,
+      "loss": 0.627,
+      "step": 20615
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.483011680444098e-06,
+      "loss": 1.0312,
+      "step": 20616
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.482087042451326e-06,
+      "loss": 1.2363,
+      "step": 20617
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.481162452988344e-06,
+      "loss": 0.7412,
+      "step": 20618
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4802379120650826e-06,
+      "loss": 1.1768,
+      "step": 20619
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.479313419691472e-06,
+      "loss": 1.0723,
+      "step": 20620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.478388975877451e-06,
+      "loss": 1.0635,
+      "step": 20621
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.477464580632945e-06,
+      "loss": 1.1985,
+      "step": 20622
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.476540233967877e-06,
+      "loss": 0.5618,
+      "step": 20623
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.475615935892187e-06,
+      "loss": 1.3066,
+      "step": 20624
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.474691686415794e-06,
+      "loss": 0.9502,
+      "step": 20625
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.473767485548635e-06,
+      "loss": 1.4082,
+      "step": 20626
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.472843333300627e-06,
+      "loss": 0.9883,
+      "step": 20627
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.471919229681706e-06,
+      "loss": 0.8789,
+      "step": 20628
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.470995174701797e-06,
+      "loss": 1.1025,
+      "step": 20629
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.470071168370818e-06,
+      "loss": 0.9277,
+      "step": 20630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.469147210698701e-06,
+      "loss": 1.0137,
+      "step": 20631
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4682233016953735e-06,
+      "loss": 0.958,
+      "step": 20632
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.467299441370756e-06,
+      "loss": 0.6113,
+      "step": 20633
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.466375629734768e-06,
+      "loss": 0.7754,
+      "step": 20634
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.465451866797342e-06,
+      "loss": 1.1436,
+      "step": 20635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4645281525683926e-06,
+      "loss": 0.9648,
+      "step": 20636
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.463604487057848e-06,
+      "loss": 1.1357,
+      "step": 20637
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.462680870275624e-06,
+      "loss": 0.8887,
+      "step": 20638
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.461757302231649e-06,
+      "loss": 1.1523,
+      "step": 20639
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4608337829358395e-06,
+      "loss": 1.0508,
+      "step": 20640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4599103123981116e-06,
+      "loss": 1.2197,
+      "step": 20641
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.458986890628389e-06,
+      "loss": 0.8704,
+      "step": 20642
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.458063517636595e-06,
+      "loss": 1.0029,
+      "step": 20643
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.457140193432644e-06,
+      "loss": 0.9814,
+      "step": 20644
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.456216918026449e-06,
+      "loss": 0.7466,
+      "step": 20645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.455293691427934e-06,
+      "loss": 1.4541,
+      "step": 20646
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.454370513647017e-06,
+      "loss": 0.8418,
+      "step": 20647
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.453447384693613e-06,
+      "loss": 1.0278,
+      "step": 20648
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.452524304577631e-06,
+      "loss": 0.6177,
+      "step": 20649
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.451601273308997e-06,
+      "loss": 1.2158,
+      "step": 20650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.450678290897619e-06,
+      "loss": 1.1367,
+      "step": 20651
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.449755357353415e-06,
+      "loss": 0.8979,
+      "step": 20652
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.448832472686294e-06,
+      "loss": 0.9609,
+      "step": 20653
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.447909636906177e-06,
+      "loss": 1.0293,
+      "step": 20654
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.446986850022971e-06,
+      "loss": 1.228,
+      "step": 20655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.446064112046585e-06,
+      "loss": 0.9688,
+      "step": 20656
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.445141422986937e-06,
+      "loss": 0.9297,
+      "step": 20657
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4442187828539406e-06,
+      "loss": 0.7715,
+      "step": 20658
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.443296191657501e-06,
+      "loss": 1.0276,
+      "step": 20659
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.442373649407526e-06,
+      "loss": 1.2227,
+      "step": 20660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.441451156113928e-06,
+      "loss": 1.0107,
+      "step": 20661
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.440528711786622e-06,
+      "loss": 0.7876,
+      "step": 20662
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.439606316435511e-06,
+      "loss": 0.9775,
+      "step": 20663
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4386839700705e-06,
+      "loss": 1.0215,
+      "step": 20664
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.437761672701505e-06,
+      "loss": 1.0977,
+      "step": 20665
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.436839424338426e-06,
+      "loss": 0.8823,
+      "step": 20666
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.435917224991169e-06,
+      "loss": 1.0444,
+      "step": 20667
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4349950746696416e-06,
+      "loss": 0.8228,
+      "step": 20668
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.434072973383753e-06,
+      "loss": 1.082,
+      "step": 20669
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.433150921143406e-06,
+      "loss": 1.04,
+      "step": 20670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.432228917958501e-06,
+      "loss": 1.4316,
+      "step": 20671
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4313069638389446e-06,
+      "loss": 0.6909,
+      "step": 20672
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.430385058794643e-06,
+      "loss": 0.7861,
+      "step": 20673
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.429463202835498e-06,
+      "loss": 1.0332,
+      "step": 20674
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4285413959714055e-06,
+      "loss": 0.8789,
+      "step": 20675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4276196382122756e-06,
+      "loss": 1.1582,
+      "step": 20676
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.426697929568001e-06,
+      "loss": 1.21,
+      "step": 20677
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.425776270048493e-06,
+      "loss": 0.9424,
+      "step": 20678
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.424854659663641e-06,
+      "loss": 1.1416,
+      "step": 20679
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.423933098423353e-06,
+      "loss": 0.6909,
+      "step": 20680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.423011586337526e-06,
+      "loss": 0.9385,
+      "step": 20681
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.422090123416048e-06,
+      "loss": 1.1963,
+      "step": 20682
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.421168709668834e-06,
+      "loss": 1.0947,
+      "step": 20683
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4202473451057736e-06,
+      "loss": 1.1797,
+      "step": 20684
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.419326029736765e-06,
+      "loss": 1.104,
+      "step": 20685
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.418404763571698e-06,
+      "loss": 0.9253,
+      "step": 20686
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.417483546620476e-06,
+      "loss": 1.0522,
+      "step": 20687
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.416562378892994e-06,
+      "loss": 1.167,
+      "step": 20688
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.415641260399143e-06,
+      "loss": 0.9766,
+      "step": 20689
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.414720191148822e-06,
+      "loss": 0.96,
+      "step": 20690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.413799171151923e-06,
+      "loss": 1.0596,
+      "step": 20691
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.412878200418334e-06,
+      "loss": 0.8086,
+      "step": 20692
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.411957278957953e-06,
+      "loss": 1.1689,
+      "step": 20693
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.411036406780676e-06,
+      "loss": 1.2021,
+      "step": 20694
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.410115583896389e-06,
+      "loss": 0.9824,
+      "step": 20695
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.409194810314985e-06,
+      "loss": 1.1763,
+      "step": 20696
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.40827408604635e-06,
+      "loss": 0.7007,
+      "step": 20697
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.407353411100377e-06,
+      "loss": 0.7256,
+      "step": 20698
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.406432785486961e-06,
+      "loss": 0.6777,
+      "step": 20699
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.405512209215984e-06,
+      "loss": 1.0273,
+      "step": 20700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.40459168229734e-06,
+      "loss": 0.9287,
+      "step": 20701
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.403671204740915e-06,
+      "loss": 0.8652,
+      "step": 20702
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.402750776556591e-06,
+      "loss": 1.0049,
+      "step": 20703
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.401830397754259e-06,
+      "loss": 1.0811,
+      "step": 20704
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.40091006834381e-06,
+      "loss": 1.1553,
+      "step": 20705
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.399989788335125e-06,
+      "loss": 1.1006,
+      "step": 20706
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.399069557738087e-06,
+      "loss": 0.7344,
+      "step": 20707
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.398149376562583e-06,
+      "loss": 1.1055,
+      "step": 20708
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.397229244818503e-06,
+      "loss": 0.8804,
+      "step": 20709
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.396309162515725e-06,
+      "loss": 1.1045,
+      "step": 20710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.395389129664128e-06,
+      "loss": 1.0,
+      "step": 20711
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3944691462736055e-06,
+      "loss": 0.9414,
+      "step": 20712
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.393549212354029e-06,
+      "loss": 1.0508,
+      "step": 20713
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.392629327915289e-06,
+      "loss": 1.0811,
+      "step": 20714
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.391709492967258e-06,
+      "loss": 0.9282,
+      "step": 20715
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.390789707519824e-06,
+      "loss": 0.7456,
+      "step": 20716
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3898699715828665e-06,
+      "loss": 1.252,
+      "step": 20717
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.388950285166256e-06,
+      "loss": 1.2939,
+      "step": 20718
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3880306482798784e-06,
+      "loss": 1.0537,
+      "step": 20719
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.387111060933616e-06,
+      "loss": 1.0234,
+      "step": 20720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.386191523137342e-06,
+      "loss": 0.7402,
+      "step": 20721
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.38527203490093e-06,
+      "loss": 0.998,
+      "step": 20722
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3843525962342606e-06,
+      "loss": 1.2295,
+      "step": 20723
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.383433207147214e-06,
+      "loss": 0.9561,
+      "step": 20724
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.382513867649663e-06,
+      "loss": 0.8154,
+      "step": 20725
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.381594577751477e-06,
+      "loss": 1.1973,
+      "step": 20726
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.38067533746254e-06,
+      "loss": 1.0488,
+      "step": 20727
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3797561467927226e-06,
+      "loss": 0.8174,
+      "step": 20728
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.378837005751894e-06,
+      "loss": 0.9307,
+      "step": 20729
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.377917914349929e-06,
+      "loss": 1.0186,
+      "step": 20730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.376998872596707e-06,
+      "loss": 1.1367,
+      "step": 20731
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.376079880502096e-06,
+      "loss": 1.1299,
+      "step": 20732
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.37516093807596e-06,
+      "loss": 1.1006,
+      "step": 20733
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.37424204532818e-06,
+      "loss": 1.0737,
+      "step": 20734
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3733232022686255e-06,
+      "loss": 1.0371,
+      "step": 20735
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.372404408907163e-06,
+      "loss": 0.9375,
+      "step": 20736
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.371485665253659e-06,
+      "loss": 1.1436,
+      "step": 20737
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.37056697131799e-06,
+      "loss": 0.7817,
+      "step": 20738
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.369648327110018e-06,
+      "loss": 1.0557,
+      "step": 20739
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.368729732639615e-06,
+      "loss": 0.5889,
+      "step": 20740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.367811187916643e-06,
+      "loss": 0.5781,
+      "step": 20741
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3668926929509745e-06,
+      "loss": 0.9446,
+      "step": 20742
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.365974247752475e-06,
+      "loss": 0.9834,
+      "step": 20743
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.365055852331003e-06,
+      "loss": 0.8096,
+      "step": 20744
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.364137506696428e-06,
+      "loss": 1.1211,
+      "step": 20745
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.363219210858621e-06,
+      "loss": 1.1289,
+      "step": 20746
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3623009648274385e-06,
+      "loss": 0.895,
+      "step": 20747
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.361382768612743e-06,
+      "loss": 0.8848,
+      "step": 20748
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3604646222244e-06,
+      "loss": 0.9346,
+      "step": 20749
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.359546525672276e-06,
+      "loss": 0.7217,
+      "step": 20750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.358628478966228e-06,
+      "loss": 1.1797,
+      "step": 20751
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357710482116116e-06,
+      "loss": 0.8652,
+      "step": 20752
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.356792535131805e-06,
+      "loss": 1.3008,
+      "step": 20753
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355874638023155e-06,
+      "loss": 0.999,
+      "step": 20754
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.354956790800019e-06,
+      "loss": 0.9014,
+      "step": 20755
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.354038993472262e-06,
+      "loss": 1.1572,
+      "step": 20756
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.353121246049745e-06,
+      "loss": 0.9619,
+      "step": 20757
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.352203548542323e-06,
+      "loss": 0.8245,
+      "step": 20758
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.351285900959851e-06,
+      "loss": 1.0635,
+      "step": 20759
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.350368303312185e-06,
+      "loss": 1.0771,
+      "step": 20760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.349450755609192e-06,
+      "loss": 0.9893,
+      "step": 20761
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3485332578607205e-06,
+      "loss": 1.0117,
+      "step": 20762
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.347615810076623e-06,
+      "loss": 0.9473,
+      "step": 20763
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.346698412266761e-06,
+      "loss": 0.8867,
+      "step": 20764
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.345781064440983e-06,
+      "loss": 1.0605,
+      "step": 20765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3448637666091475e-06,
+      "loss": 0.8491,
+      "step": 20766
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.343946518781103e-06,
+      "loss": 1.0605,
+      "step": 20767
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3430293209667085e-06,
+      "loss": 1.0098,
+      "step": 20768
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.342112173175813e-06,
+      "loss": 1.1445,
+      "step": 20769
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.34119507541826e-06,
+      "loss": 0.9688,
+      "step": 20770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.340278027703916e-06,
+      "loss": 0.9873,
+      "step": 20771
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.339361030042625e-06,
+      "loss": 0.8779,
+      "step": 20772
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.338444082444237e-06,
+      "loss": 0.8979,
+      "step": 20773
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.337527184918594e-06,
+      "loss": 1.1328,
+      "step": 20774
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.336610337475554e-06,
+      "loss": 1.1055,
+      "step": 20775
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3356935401249664e-06,
+      "loss": 0.9043,
+      "step": 20776
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.334776792876677e-06,
+      "loss": 1.0576,
+      "step": 20777
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.333860095740526e-06,
+      "loss": 0.7637,
+      "step": 20778
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.332943448726371e-06,
+      "loss": 1.126,
+      "step": 20779
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.33202685184405e-06,
+      "loss": 0.9092,
+      "step": 20780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.331110305103413e-06,
+      "loss": 0.8394,
+      "step": 20781
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.330193808514308e-06,
+      "loss": 0.708,
+      "step": 20782
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.329277362086576e-06,
+      "loss": 1.0928,
+      "step": 20783
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.32836096583006e-06,
+      "loss": 0.8555,
+      "step": 20784
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.327444619754602e-06,
+      "loss": 1.0537,
+      "step": 20785
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326528323870049e-06,
+      "loss": 0.7878,
+      "step": 20786
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.325612078186247e-06,
+      "loss": 0.8906,
+      "step": 20787
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3246958827130265e-06,
+      "loss": 1.1221,
+      "step": 20788
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3237797374602415e-06,
+      "loss": 0.8599,
+      "step": 20789
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3228636424377275e-06,
+      "loss": 0.833,
+      "step": 20790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.321947597655321e-06,
+      "loss": 0.7886,
+      "step": 20791
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.321031603122866e-06,
+      "loss": 1.0928,
+      "step": 20792
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.320115658850204e-06,
+      "loss": 1.0391,
+      "step": 20793
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.319199764847171e-06,
+      "loss": 0.9424,
+      "step": 20794
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.318283921123606e-06,
+      "loss": 1.1094,
+      "step": 20795
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3173681276893385e-06,
+      "loss": 1.0625,
+      "step": 20796
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.31645238455422e-06,
+      "loss": 0.9141,
+      "step": 20797
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.31553669172808e-06,
+      "loss": 0.8013,
+      "step": 20798
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3146210492207515e-06,
+      "loss": 1.3438,
+      "step": 20799
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3137054570420775e-06,
+      "loss": 1.2236,
+      "step": 20800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.312789915201884e-06,
+      "loss": 1.0449,
+      "step": 20801
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.311874423710013e-06,
+      "loss": 1.0674,
+      "step": 20802
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.310958982576293e-06,
+      "loss": 1.0405,
+      "step": 20803
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.310043591810563e-06,
+      "loss": 1.1836,
+      "step": 20804
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.309128251422653e-06,
+      "loss": 0.835,
+      "step": 20805
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.30821296142239e-06,
+      "loss": 0.7759,
+      "step": 20806
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.307297721819611e-06,
+      "loss": 1.252,
+      "step": 20807
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.30638253262415e-06,
+      "loss": 1.1572,
+      "step": 20808
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.305467393845835e-06,
+      "loss": 0.8096,
+      "step": 20809
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.304552305494493e-06,
+      "loss": 0.6997,
+      "step": 20810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.303637267579953e-06,
+      "loss": 0.9443,
+      "step": 20811
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3027222801120536e-06,
+      "loss": 0.7437,
+      "step": 20812
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3018073431006155e-06,
+      "loss": 1.1475,
+      "step": 20813
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.300892456555463e-06,
+      "loss": 1.1309,
+      "step": 20814
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.299977620486434e-06,
+      "loss": 0.8418,
+      "step": 20815
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.299062834903348e-06,
+      "loss": 1.085,
+      "step": 20816
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.298148099816029e-06,
+      "loss": 0.9756,
+      "step": 20817
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.297233415234308e-06,
+      "loss": 0.6738,
+      "step": 20818
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.296318781168011e-06,
+      "loss": 1.084,
+      "step": 20819
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.295404197626962e-06,
+      "loss": 1.0723,
+      "step": 20820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.294489664620979e-06,
+      "loss": 0.9414,
+      "step": 20821
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2935751821598914e-06,
+      "loss": 1.0078,
+      "step": 20822
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.292660750253524e-06,
+      "loss": 1.0264,
+      "step": 20823
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2917463689116964e-06,
+      "loss": 0.7832,
+      "step": 20824
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.290832038144227e-06,
+      "loss": 1.1465,
+      "step": 20825
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.289917757960943e-06,
+      "loss": 0.8291,
+      "step": 20826
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.289003528371661e-06,
+      "loss": 0.6641,
+      "step": 20827
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.288089349386208e-06,
+      "loss": 1.0527,
+      "step": 20828
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2871752210143925e-06,
+      "loss": 1.3418,
+      "step": 20829
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286261143266045e-06,
+      "loss": 0.7041,
+      "step": 20830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2853471161509805e-06,
+      "loss": 0.7334,
+      "step": 20831
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.284433139679013e-06,
+      "loss": 1.1768,
+      "step": 20832
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.283519213859961e-06,
+      "loss": 0.9883,
+      "step": 20833
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2826053387036495e-06,
+      "loss": 1.0488,
+      "step": 20834
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.281691514219889e-06,
+      "loss": 1.189,
+      "step": 20835
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.280777740418492e-06,
+      "loss": 0.875,
+      "step": 20836
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.279864017309277e-06,
+      "loss": 1.3438,
+      "step": 20837
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.278950344902064e-06,
+      "loss": 0.9746,
+      "step": 20838
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.278036723206663e-06,
+      "loss": 1.1562,
+      "step": 20839
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.277123152232885e-06,
+      "loss": 0.9814,
+      "step": 20840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.276209631990549e-06,
+      "loss": 0.8398,
+      "step": 20841
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.275296162489461e-06,
+      "loss": 1.002,
+      "step": 20842
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.27438274373944e-06,
+      "loss": 0.7998,
+      "step": 20843
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.273469375750291e-06,
+      "loss": 1.167,
+      "step": 20844
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.272556058531832e-06,
+      "loss": 1.1187,
+      "step": 20845
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2716427920938714e-06,
+      "loss": 1.1094,
+      "step": 20846
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.270729576446214e-06,
+      "loss": 1.2041,
+      "step": 20847
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.269816411598673e-06,
+      "loss": 1.2988,
+      "step": 20848
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.268903297561061e-06,
+      "loss": 0.8164,
+      "step": 20849
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.267990234343182e-06,
+      "loss": 0.8428,
+      "step": 20850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.267077221954842e-06,
+      "loss": 1.2197,
+      "step": 20851
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.266164260405854e-06,
+      "loss": 0.8857,
+      "step": 20852
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.265251349706018e-06,
+      "loss": 0.8267,
+      "step": 20853
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.264338489865146e-06,
+      "loss": 1.0938,
+      "step": 20854
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.26342568089304e-06,
+      "loss": 0.8391,
+      "step": 20855
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.262512922799508e-06,
+      "loss": 1.0659,
+      "step": 20856
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.261600215594354e-06,
+      "loss": 1.0918,
+      "step": 20857
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.260687559287376e-06,
+      "loss": 1.2012,
+      "step": 20858
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2597749538883835e-06,
+      "loss": 0.9683,
+      "step": 20859
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.258862399407181e-06,
+      "loss": 1.0049,
+      "step": 20860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.257949895853569e-06,
+      "loss": 1.0703,
+      "step": 20861
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.257037443237343e-06,
+      "loss": 0.667,
+      "step": 20862
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.256125041568311e-06,
+      "loss": 1.0039,
+      "step": 20863
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.255212690856275e-06,
+      "loss": 1.0244,
+      "step": 20864
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2543003911110325e-06,
+      "loss": 0.9492,
+      "step": 20865
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.253388142342379e-06,
+      "loss": 0.9434,
+      "step": 20866
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.252475944560122e-06,
+      "loss": 0.9053,
+      "step": 20867
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.251563797774051e-06,
+      "loss": 1.1768,
+      "step": 20868
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2506517019939716e-06,
+      "loss": 0.9858,
+      "step": 20869
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.249739657229676e-06,
+      "loss": 1.4854,
+      "step": 20870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.248827663490965e-06,
+      "loss": 1.2314,
+      "step": 20871
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247915720787633e-06,
+      "loss": 1.0947,
+      "step": 20872
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247003829129469e-06,
+      "loss": 0.9561,
+      "step": 20873
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2460919885262825e-06,
+      "loss": 0.7783,
+      "step": 20874
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.24518019898786e-06,
+      "loss": 1.5957,
+      "step": 20875
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.244268460523998e-06,
+      "loss": 0.9863,
+      "step": 20876
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.243356773144484e-06,
+      "loss": 0.7778,
+      "step": 20877
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.242445136859118e-06,
+      "loss": 0.877,
+      "step": 20878
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.241533551677686e-06,
+      "loss": 0.8818,
+      "step": 20879
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.240622017609984e-06,
+      "loss": 0.9121,
+      "step": 20880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.239710534665807e-06,
+      "loss": 1.1787,
+      "step": 20881
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.238799102854941e-06,
+      "loss": 0.8174,
+      "step": 20882
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.237887722187179e-06,
+      "loss": 1.0352,
+      "step": 20883
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.236976392672299e-06,
+      "loss": 0.7803,
+      "step": 20884
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.236065114320109e-06,
+      "loss": 0.5229,
+      "step": 20885
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.235153887140389e-06,
+      "loss": 1.1348,
+      "step": 20886
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.234242711142923e-06,
+      "loss": 1.2344,
+      "step": 20887
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.233331586337505e-06,
+      "loss": 1.1201,
+      "step": 20888
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.232420512733916e-06,
+      "loss": 1.2656,
+      "step": 20889
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.231509490341949e-06,
+      "loss": 0.9648,
+      "step": 20890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.230598519171383e-06,
+      "loss": 1.0078,
+      "step": 20891
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.229687599232012e-06,
+      "loss": 1.0654,
+      "step": 20892
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.228776730533614e-06,
+      "loss": 0.9277,
+      "step": 20893
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.227865913085972e-06,
+      "loss": 1.1592,
+      "step": 20894
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.226955146898871e-06,
+      "loss": 1.2197,
+      "step": 20895
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.226044431982101e-06,
+      "loss": 0.7017,
+      "step": 20896
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2251337683454385e-06,
+      "loss": 1.1924,
+      "step": 20897
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.224223155998663e-06,
+      "loss": 0.9102,
+      "step": 20898
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2233125949515595e-06,
+      "loss": 0.8892,
+      "step": 20899
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.222402085213911e-06,
+      "loss": 0.4302,
+      "step": 20900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.221491626795496e-06,
+      "loss": 1.1689,
+      "step": 20901
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.220581219706091e-06,
+      "loss": 1.0771,
+      "step": 20902
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.219670863955482e-06,
+      "loss": 0.958,
+      "step": 20903
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.21876055955344e-06,
+      "loss": 0.7446,
+      "step": 20904
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.217850306509748e-06,
+      "loss": 0.8252,
+      "step": 20905
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2169401048341815e-06,
+      "loss": 1.1431,
+      "step": 20906
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.216029954536522e-06,
+      "loss": 0.8872,
+      "step": 20907
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.215119855626543e-06,
+      "loss": 1.1357,
+      "step": 20908
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214209808114015e-06,
+      "loss": 0.9272,
+      "step": 20909
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.213299812008719e-06,
+      "loss": 0.8237,
+      "step": 20910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2123898673204335e-06,
+      "loss": 0.8652,
+      "step": 20911
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2114799740589285e-06,
+      "loss": 0.9023,
+      "step": 20912
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2105701322339744e-06,
+      "loss": 1.2354,
+      "step": 20913
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.209660341855351e-06,
+      "loss": 1.2812,
+      "step": 20914
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.208750602932824e-06,
+      "loss": 1.0859,
+      "step": 20915
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.207840915476175e-06,
+      "loss": 0.8406,
+      "step": 20916
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.206931279495163e-06,
+      "loss": 1.1162,
+      "step": 20917
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.206021694999571e-06,
+      "loss": 0.7656,
+      "step": 20918
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.205112161999164e-06,
+      "loss": 1.1436,
+      "step": 20919
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.204202680503709e-06,
+      "loss": 1.1855,
+      "step": 20920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.203293250522978e-06,
+      "loss": 0.9854,
+      "step": 20921
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202383872066744e-06,
+      "loss": 1.1133,
+      "step": 20922
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.201474545144771e-06,
+      "loss": 1.0342,
+      "step": 20923
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.200565269766824e-06,
+      "loss": 0.7935,
+      "step": 20924
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.199656045942674e-06,
+      "loss": 1.2158,
+      "step": 20925
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.198746873682089e-06,
+      "loss": 1.1309,
+      "step": 20926
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1978377529948344e-06,
+      "loss": 0.9805,
+      "step": 20927
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.196928683890667e-06,
+      "loss": 1.1899,
+      "step": 20928
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1960196663793664e-06,
+      "loss": 0.7871,
+      "step": 20929
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1951107004706825e-06,
+      "loss": 0.6104,
+      "step": 20930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.194201786174391e-06,
+      "loss": 0.8159,
+      "step": 20931
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193292923500246e-06,
+      "loss": 0.979,
+      "step": 20932
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.192384112458017e-06,
+      "loss": 0.9062,
+      "step": 20933
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1914753530574644e-06,
+      "loss": 0.3833,
+      "step": 20934
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1905666453083445e-06,
+      "loss": 1.3496,
+      "step": 20935
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.189657989220422e-06,
+      "loss": 1.4502,
+      "step": 20936
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.188749384803462e-06,
+      "loss": 1.0615,
+      "step": 20937
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.187840832067221e-06,
+      "loss": 1.0483,
+      "step": 20938
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.186932331021453e-06,
+      "loss": 0.7742,
+      "step": 20939
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1860238816759235e-06,
+      "loss": 0.6992,
+      "step": 20940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.185115484040386e-06,
+      "loss": 0.8848,
+      "step": 20941
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.184207138124606e-06,
+      "loss": 0.7979,
+      "step": 20942
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.183298843938329e-06,
+      "loss": 0.8682,
+      "step": 20943
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.182390601491323e-06,
+      "loss": 0.8291,
+      "step": 20944
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.181482410793339e-06,
+      "loss": 0.9375,
+      "step": 20945
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.180574271854129e-06,
+      "loss": 0.8457,
+      "step": 20946
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.17966618468345e-06,
+      "loss": 0.8247,
+      "step": 20947
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.178758149291063e-06,
+      "loss": 1.2363,
+      "step": 20948
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.177850165686715e-06,
+      "loss": 1.541,
+      "step": 20949
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.176942233880158e-06,
+      "loss": 0.7324,
+      "step": 20950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.176034353881145e-06,
+      "loss": 1.084,
+      "step": 20951
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.175126525699437e-06,
+      "loss": 0.8804,
+      "step": 20952
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.174218749344777e-06,
+      "loss": 1.2158,
+      "step": 20953
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.173311024826916e-06,
+      "loss": 1.2578,
+      "step": 20954
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.172403352155609e-06,
+      "loss": 1.1416,
+      "step": 20955
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1714957313406015e-06,
+      "loss": 0.9199,
+      "step": 20956
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.170588162391645e-06,
+      "loss": 0.856,
+      "step": 20957
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.169680645318487e-06,
+      "loss": 1.2246,
+      "step": 20958
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.168773180130879e-06,
+      "loss": 1.0488,
+      "step": 20959
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1678657668385675e-06,
+      "loss": 1.0762,
+      "step": 20960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16695840545129e-06,
+      "loss": 1.4512,
+      "step": 20961
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16605109597881e-06,
+      "loss": 1.2734,
+      "step": 20962
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1651438384308635e-06,
+      "loss": 0.6074,
+      "step": 20963
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.164236632817199e-06,
+      "loss": 0.7578,
+      "step": 20964
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.163329479147555e-06,
+      "loss": 1.2119,
+      "step": 20965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.162422377431682e-06,
+      "loss": 0.8516,
+      "step": 20966
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.161515327679325e-06,
+      "loss": 1.084,
+      "step": 20967
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.160608329900225e-06,
+      "loss": 1.1157,
+      "step": 20968
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.159701384104119e-06,
+      "loss": 0.7178,
+      "step": 20969
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.15879449030076e-06,
+      "loss": 0.9902,
+      "step": 20970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.157887648499883e-06,
+      "loss": 1.3066,
+      "step": 20971
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.156980858711223e-06,
+      "loss": 1.251,
+      "step": 20972
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.156074120944534e-06,
+      "loss": 0.8413,
+      "step": 20973
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.155167435209549e-06,
+      "loss": 0.8408,
+      "step": 20974
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.154260801516007e-06,
+      "loss": 1.1211,
+      "step": 20975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.153354219873644e-06,
+      "loss": 0.8213,
+      "step": 20976
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.152447690292202e-06,
+      "loss": 1.0186,
+      "step": 20977
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.15154121278142e-06,
+      "loss": 1.0732,
+      "step": 20978
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1506347873510295e-06,
+      "loss": 0.5913,
+      "step": 20979
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149728414010775e-06,
+      "loss": 0.7354,
+      "step": 20980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.148822092770387e-06,
+      "loss": 0.9771,
+      "step": 20981
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1479158236395994e-06,
+      "loss": 1.1318,
+      "step": 20982
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.147009606628148e-06,
+      "loss": 0.896,
+      "step": 20983
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.146103441745773e-06,
+      "loss": 0.7378,
+      "step": 20984
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1451973290022025e-06,
+      "loss": 1.1465,
+      "step": 20985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.144291268407168e-06,
+      "loss": 1.2461,
+      "step": 20986
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.143385259970404e-06,
+      "loss": 0.8799,
+      "step": 20987
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.142479303701646e-06,
+      "loss": 0.9121,
+      "step": 20988
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.141573399610624e-06,
+      "loss": 1.1099,
+      "step": 20989
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.140667547707065e-06,
+      "loss": 0.9365,
+      "step": 20990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.139761748000705e-06,
+      "loss": 0.9351,
+      "step": 20991
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1388560005012645e-06,
+      "loss": 0.868,
+      "step": 20992
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137950305218484e-06,
+      "loss": 0.7036,
+      "step": 20993
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137044662162085e-06,
+      "loss": 1.4761,
+      "step": 20994
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.136139071341799e-06,
+      "loss": 0.9609,
+      "step": 20995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.135233532767354e-06,
+      "loss": 1.1289,
+      "step": 20996
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.134328046448471e-06,
+      "loss": 0.8223,
+      "step": 20997
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.13342261239488e-06,
+      "loss": 1.0107,
+      "step": 20998
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132517230616312e-06,
+      "loss": 0.8408,
+      "step": 20999
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.131611901122486e-06,
+      "loss": 0.9155,
+      "step": 21000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.130706623923126e-06,
+      "loss": 0.8477,
+      "step": 21001
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.129801399027962e-06,
+      "loss": 0.9255,
+      "step": 21002
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.128896226446711e-06,
+      "loss": 0.874,
+      "step": 21003
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.127991106189102e-06,
+      "loss": 0.7119,
+      "step": 21004
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.127086038264851e-06,
+      "loss": 0.8408,
+      "step": 21005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.126181022683689e-06,
+      "loss": 0.6699,
+      "step": 21006
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12527605945533e-06,
+      "loss": 0.916,
+      "step": 21007
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.124371148589493e-06,
+      "loss": 1.1191,
+      "step": 21008
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.123466290095902e-06,
+      "loss": 0.9004,
+      "step": 21009
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.122561483984282e-06,
+      "loss": 0.9814,
+      "step": 21010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.121656730264346e-06,
+      "loss": 0.9785,
+      "step": 21011
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.120752028945808e-06,
+      "loss": 0.9121,
+      "step": 21012
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.119847380038392e-06,
+      "loss": 1.0811,
+      "step": 21013
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.118942783551819e-06,
+      "loss": 0.9629,
+      "step": 21014
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.118038239495802e-06,
+      "loss": 0.9707,
+      "step": 21015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.117133747880052e-06,
+      "loss": 0.9434,
+      "step": 21016
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.116229308714293e-06,
+      "loss": 0.9385,
+      "step": 21017
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1153249220082335e-06,
+      "loss": 0.8809,
+      "step": 21018
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.114420587771596e-06,
+      "loss": 0.8916,
+      "step": 21019
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1135163060140835e-06,
+      "loss": 0.7832,
+      "step": 21020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.112612076745422e-06,
+      "loss": 0.9307,
+      "step": 21021
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1117078999753174e-06,
+      "loss": 0.8691,
+      "step": 21022
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.110803775713478e-06,
+      "loss": 1.2695,
+      "step": 21023
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.109899703969621e-06,
+      "loss": 1.0283,
+      "step": 21024
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.108995684753462e-06,
+      "loss": 1.0586,
+      "step": 21025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.108091718074706e-06,
+      "loss": 0.8208,
+      "step": 21026
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.107187803943059e-06,
+      "loss": 1.1895,
+      "step": 21027
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.106283942368235e-06,
+      "loss": 0.9219,
+      "step": 21028
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.105380133359948e-06,
+      "loss": 1.2031,
+      "step": 21029
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.104476376927902e-06,
+      "loss": 0.8843,
+      "step": 21030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.103572673081799e-06,
+      "loss": 1.0815,
+      "step": 21031
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1026690218313565e-06,
+      "loss": 1.6445,
+      "step": 21032
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.101765423186276e-06,
+      "loss": 1.3955,
+      "step": 21033
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.10086187715626e-06,
+      "loss": 0.729,
+      "step": 21034
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.09995838375102e-06,
+      "loss": 1.0146,
+      "step": 21035
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.099054942980261e-06,
+      "loss": 0.9932,
+      "step": 21036
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0981515548536855e-06,
+      "loss": 1.1738,
+      "step": 21037
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.097248219380994e-06,
+      "loss": 1.1484,
+      "step": 21038
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.096344936571893e-06,
+      "loss": 0.707,
+      "step": 21039
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0954417064360905e-06,
+      "loss": 0.8364,
+      "step": 21040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.094538528983282e-06,
+      "loss": 0.9492,
+      "step": 21041
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.093635404223168e-06,
+      "loss": 0.7744,
+      "step": 21042
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.092732332165454e-06,
+      "loss": 1.2646,
+      "step": 21043
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.091829312819836e-06,
+      "loss": 0.8467,
+      "step": 21044
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0909263461960215e-06,
+      "loss": 1.04,
+      "step": 21045
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0900234323037e-06,
+      "loss": 0.9551,
+      "step": 21046
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.089120571152579e-06,
+      "loss": 1.115,
+      "step": 21047
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.088217762752355e-06,
+      "loss": 0.7451,
+      "step": 21048
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.087315007112717e-06,
+      "loss": 0.6216,
+      "step": 21049
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.08641230424337e-06,
+      "loss": 1.1035,
+      "step": 21050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.085509654154013e-06,
+      "loss": 1.1289,
+      "step": 21051
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0846070568543375e-06,
+      "loss": 0.9932,
+      "step": 21052
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.083704512354035e-06,
+      "loss": 0.9214,
+      "step": 21053
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.082802020662807e-06,
+      "loss": 1.0342,
+      "step": 21054
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.081899581790348e-06,
+      "loss": 1.0244,
+      "step": 21055
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.080997195746349e-06,
+      "loss": 1.0391,
+      "step": 21056
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0800948625405e-06,
+      "loss": 0.9463,
+      "step": 21057
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.079192582182501e-06,
+      "loss": 1.0781,
+      "step": 21058
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.078290354682035e-06,
+      "loss": 0.9775,
+      "step": 21059
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077388180048799e-06,
+      "loss": 0.9497,
+      "step": 21060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.076486058292486e-06,
+      "loss": 1.2456,
+      "step": 21061
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.075583989422783e-06,
+      "loss": 1.127,
+      "step": 21062
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074681973449383e-06,
+      "loss": 1.251,
+      "step": 21063
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.073780010381966e-06,
+      "loss": 0.7622,
+      "step": 21064
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.072878100230225e-06,
+      "loss": 0.8643,
+      "step": 21065
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.071976243003856e-06,
+      "loss": 1.1348,
+      "step": 21066
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.07107443871254e-06,
+      "loss": 0.7693,
+      "step": 21067
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.07017268736596e-06,
+      "loss": 0.7812,
+      "step": 21068
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.069270988973809e-06,
+      "loss": 1.1143,
+      "step": 21069
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.068369343545767e-06,
+      "loss": 1.0615,
+      "step": 21070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0674677510915225e-06,
+      "loss": 1.4961,
+      "step": 21071
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.066566211620763e-06,
+      "loss": 0.7041,
+      "step": 21072
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.065664725143169e-06,
+      "loss": 1.0693,
+      "step": 21073
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.064763291668425e-06,
+      "loss": 1.0996,
+      "step": 21074
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.063861911206205e-06,
+      "loss": 1.0049,
+      "step": 21075
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.062960583766208e-06,
+      "loss": 1.0674,
+      "step": 21076
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.062059309358105e-06,
+      "loss": 1.0605,
+      "step": 21077
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.061158087991577e-06,
+      "loss": 1.0264,
+      "step": 21078
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.06025691967631e-06,
+      "loss": 1.1572,
+      "step": 21079
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.059355804421977e-06,
+      "loss": 1.0508,
+      "step": 21080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.058454742238266e-06,
+      "loss": 1.1523,
+      "step": 21081
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.057553733134845e-06,
+      "loss": 0.7427,
+      "step": 21082
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.056652777121405e-06,
+      "loss": 1.0015,
+      "step": 21083
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.055751874207616e-06,
+      "loss": 0.6794,
+      "step": 21084
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.054851024403154e-06,
+      "loss": 0.9111,
+      "step": 21085
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.053950227717696e-06,
+      "loss": 0.8955,
+      "step": 21086
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.053049484160925e-06,
+      "loss": 1.2227,
+      "step": 21087
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.052148793742511e-06,
+      "loss": 1.0137,
+      "step": 21088
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.051248156472126e-06,
+      "loss": 0.8984,
+      "step": 21089
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.050347572359452e-06,
+      "loss": 1.0596,
+      "step": 21090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.049447041414155e-06,
+      "loss": 1.0571,
+      "step": 21091
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.048546563645914e-06,
+      "loss": 1.0869,
+      "step": 21092
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.047646139064397e-06,
+      "loss": 1.1816,
+      "step": 21093
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.046745767679281e-06,
+      "loss": 0.8857,
+      "step": 21094
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.045845449500235e-06,
+      "loss": 0.9492,
+      "step": 21095
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.044945184536927e-06,
+      "loss": 0.9658,
+      "step": 21096
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.044044972799028e-06,
+      "loss": 0.8535,
+      "step": 21097
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.043144814296214e-06,
+      "loss": 1.1191,
+      "step": 21098
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.042244709038151e-06,
+      "loss": 0.8452,
+      "step": 21099
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.041344657034502e-06,
+      "loss": 1.3652,
+      "step": 21100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.04044465829494e-06,
+      "loss": 0.6953,
+      "step": 21101
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.039544712829136e-06,
+      "loss": 0.9683,
+      "step": 21102
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.038644820646752e-06,
+      "loss": 1.332,
+      "step": 21103
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.037744981757453e-06,
+      "loss": 1.0332,
+      "step": 21104
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.03684519617091e-06,
+      "loss": 1.2178,
+      "step": 21105
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.035945463896781e-06,
+      "loss": 1.3906,
+      "step": 21106
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.035045784944738e-06,
+      "loss": 1.0098,
+      "step": 21107
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.03414615932444e-06,
+      "loss": 0.8369,
+      "step": 21108
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0332465870455535e-06,
+      "loss": 0.9326,
+      "step": 21109
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.032347068117742e-06,
+      "loss": 0.8574,
+      "step": 21110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.031447602550661e-06,
+      "loss": 0.6724,
+      "step": 21111
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0305481903539785e-06,
+      "loss": 0.6553,
+      "step": 21112
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.029648831537357e-06,
+      "loss": 0.9189,
+      "step": 21113
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.028749526110455e-06,
+      "loss": 0.8936,
+      "step": 21114
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0278502740829275e-06,
+      "loss": 0.9033,
+      "step": 21115
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.026951075464439e-06,
+      "loss": 1.0762,
+      "step": 21116
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.02605193026465e-06,
+      "loss": 1.0654,
+      "step": 21117
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.025152838493219e-06,
+      "loss": 0.7937,
+      "step": 21118
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.024253800159795e-06,
+      "loss": 0.9785,
+      "step": 21119
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.023354815274047e-06,
+      "loss": 1.0918,
+      "step": 21120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.022455883845625e-06,
+      "loss": 0.9805,
+      "step": 21121
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.021557005884183e-06,
+      "loss": 1.0615,
+      "step": 21122
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0206581813993784e-06,
+      "loss": 0.7837,
+      "step": 21123
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019759410400871e-06,
+      "loss": 0.8613,
+      "step": 21124
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.018860692898312e-06,
+      "loss": 1.2324,
+      "step": 21125
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.017962028901349e-06,
+      "loss": 0.8518,
+      "step": 21126
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.017063418419641e-06,
+      "loss": 0.856,
+      "step": 21127
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.016164861462844e-06,
+      "loss": 1.2441,
+      "step": 21128
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.015266358040606e-06,
+      "loss": 1.582,
+      "step": 21129
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.014367908162574e-06,
+      "loss": 0.9053,
+      "step": 21130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.013469511838407e-06,
+      "loss": 1.1655,
+      "step": 21131
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.012571169077747e-06,
+      "loss": 0.7759,
+      "step": 21132
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.011672879890252e-06,
+      "loss": 0.8281,
+      "step": 21133
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.010774644285564e-06,
+      "loss": 0.8169,
+      "step": 21134
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.009876462273337e-06,
+      "loss": 0.7068,
+      "step": 21135
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.008978333863219e-06,
+      "loss": 0.9453,
+      "step": 21136
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.008080259064849e-06,
+      "loss": 1.0156,
+      "step": 21137
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.00718223788788e-06,
+      "loss": 1.188,
+      "step": 21138
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.006284270341963e-06,
+      "loss": 0.9453,
+      "step": 21139
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.005386356436737e-06,
+      "loss": 0.981,
+      "step": 21140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.004488496181846e-06,
+      "loss": 1.0859,
+      "step": 21141
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0035906895869365e-06,
+      "loss": 0.8955,
+      "step": 21142
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0026929366616574e-06,
+      "loss": 0.8228,
+      "step": 21143
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.001795237415647e-06,
+      "loss": 1.1572,
+      "step": 21144
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.000897591858546e-06,
+      "loss": 1.0557,
+      "step": 21145
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.000000000000003e-06,
+      "loss": 1.207,
+      "step": 21146
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.999102461849651e-06,
+      "loss": 1.208,
+      "step": 21147
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9982049774171404e-06,
+      "loss": 1.1318,
+      "step": 21148
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.997307546712104e-06,
+      "loss": 0.5342,
+      "step": 21149
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996410169744188e-06,
+      "loss": 0.8169,
+      "step": 21150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9955128465230285e-06,
+      "loss": 1.1553,
+      "step": 21151
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.994615577058262e-06,
+      "loss": 1.0146,
+      "step": 21152
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9937183613595275e-06,
+      "loss": 1.0337,
+      "step": 21153
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.992821199436469e-06,
+      "loss": 1.0332,
+      "step": 21154
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991924091298716e-06,
+      "loss": 0.9258,
+      "step": 21155
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991027036955906e-06,
+      "loss": 1.1133,
+      "step": 21156
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9901300364176796e-06,
+      "loss": 1.291,
+      "step": 21157
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.989233089693665e-06,
+      "loss": 0.915,
+      "step": 21158
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.988336196793501e-06,
+      "loss": 0.8896,
+      "step": 21159
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.987439357726824e-06,
+      "loss": 1.1338,
+      "step": 21160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986542572503266e-06,
+      "loss": 1.2236,
+      "step": 21161
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.985645841132458e-06,
+      "loss": 1.2695,
+      "step": 21162
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.984749163624025e-06,
+      "loss": 0.8037,
+      "step": 21163
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9838525399876156e-06,
+      "loss": 0.8208,
+      "step": 21164
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9829559702328525e-06,
+      "loss": 0.8408,
+      "step": 21165
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9820594543693654e-06,
+      "loss": 1.0161,
+      "step": 21166
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.981162992406782e-06,
+      "loss": 0.9121,
+      "step": 21167
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.980266584354735e-06,
+      "loss": 1.043,
+      "step": 21168
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.979370230222856e-06,
+      "loss": 1.1924,
+      "step": 21169
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9784739300207675e-06,
+      "loss": 1.1396,
+      "step": 21170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.977577683758104e-06,
+      "loss": 1.0059,
+      "step": 21171
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.976681491444488e-06,
+      "loss": 0.9336,
+      "step": 21172
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.975785353089544e-06,
+      "loss": 0.8174,
+      "step": 21173
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.974889268702902e-06,
+      "loss": 0.8975,
+      "step": 21174
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.973993238294188e-06,
+      "loss": 1.2676,
+      "step": 21175
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.973097261873027e-06,
+      "loss": 1.1201,
+      "step": 21176
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.972201339449036e-06,
+      "loss": 0.5879,
+      "step": 21177
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.971305471031845e-06,
+      "loss": 0.6938,
+      "step": 21178
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.97040965663108e-06,
+      "loss": 1.25,
+      "step": 21179
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9695138962563605e-06,
+      "loss": 1.21,
+      "step": 21180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.968618189917303e-06,
+      "loss": 0.8154,
+      "step": 21181
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.967722537623537e-06,
+      "loss": 0.998,
+      "step": 21182
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.966826939384681e-06,
+      "loss": 1.0078,
+      "step": 21183
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9659313952103494e-06,
+      "loss": 1.0347,
+      "step": 21184
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.965035905110166e-06,
+      "loss": 0.9746,
+      "step": 21185
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.964140469093755e-06,
+      "loss": 0.8579,
+      "step": 21186
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.96324508717073e-06,
+      "loss": 0.8828,
+      "step": 21187
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.962349759350704e-06,
+      "loss": 1.5488,
+      "step": 21188
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.961454485643301e-06,
+      "loss": 1.1875,
+      "step": 21189
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.960559266058139e-06,
+      "loss": 1.4863,
+      "step": 21190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.959664100604831e-06,
+      "loss": 1.0137,
+      "step": 21191
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9587689892929895e-06,
+      "loss": 1.1621,
+      "step": 21192
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957873932132235e-06,
+      "loss": 0.6787,
+      "step": 21193
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.956978929132177e-06,
+      "loss": 1.0088,
+      "step": 21194
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.956083980302434e-06,
+      "loss": 1.1074,
+      "step": 21195
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.955189085652614e-06,
+      "loss": 1.0576,
+      "step": 21196
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.954294245192337e-06,
+      "loss": 0.751,
+      "step": 21197
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9533994589312105e-06,
+      "loss": 1.1094,
+      "step": 21198
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.952504726878842e-06,
+      "loss": 1.1562,
+      "step": 21199
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9516100490448475e-06,
+      "loss": 1.0488,
+      "step": 21200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.950715425438839e-06,
+      "loss": 0.4663,
+      "step": 21201
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.949820856070424e-06,
+      "loss": 1.3477,
+      "step": 21202
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9489263409492076e-06,
+      "loss": 0.79,
+      "step": 21203
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.948031880084802e-06,
+      "loss": 0.7275,
+      "step": 21204
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.947137473486818e-06,
+      "loss": 1.0176,
+      "step": 21205
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.946243121164861e-06,
+      "loss": 0.7441,
+      "step": 21206
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.945348823128533e-06,
+      "loss": 1.3037,
+      "step": 21207
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.944454579387448e-06,
+      "loss": 1.1006,
+      "step": 21208
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.943560389951204e-06,
+      "loss": 0.8896,
+      "step": 21209
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.942666254829414e-06,
+      "loss": 1.0635,
+      "step": 21210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.941772174031674e-06,
+      "loss": 1.2812,
+      "step": 21211
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.940878147567596e-06,
+      "loss": 1.1748,
+      "step": 21212
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.939984175446779e-06,
+      "loss": 0.8486,
+      "step": 21213
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.939090257678824e-06,
+      "loss": 0.9131,
+      "step": 21214
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.938196394273335e-06,
+      "loss": 1.1416,
+      "step": 21215
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.937302585239917e-06,
+      "loss": 0.6074,
+      "step": 21216
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.936408830588169e-06,
+      "loss": 0.9702,
+      "step": 21217
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.935515130327686e-06,
+      "loss": 1.0176,
+      "step": 21218
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.934621484468076e-06,
+      "loss": 0.9062,
+      "step": 21219
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9337278930189305e-06,
+      "loss": 1.0615,
+      "step": 21220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.932834355989856e-06,
+      "loss": 1.083,
+      "step": 21221
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.931940873390443e-06,
+      "loss": 1.2061,
+      "step": 21222
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9310474452302955e-06,
+      "loss": 1.0801,
+      "step": 21223
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.930154071519009e-06,
+      "loss": 1.0508,
+      "step": 21224
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.929260752266173e-06,
+      "loss": 0.5227,
+      "step": 21225
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.928367487481389e-06,
+      "loss": 0.958,
+      "step": 21226
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.927474277174255e-06,
+      "loss": 1.1865,
+      "step": 21227
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.926581121354362e-06,
+      "loss": 0.7642,
+      "step": 21228
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.925688020031301e-06,
+      "loss": 1.3174,
+      "step": 21229
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.924794973214668e-06,
+      "loss": 1.3105,
+      "step": 21230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.92390198091406e-06,
+      "loss": 1.0107,
+      "step": 21231
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.923009043139065e-06,
+      "loss": 0.9121,
+      "step": 21232
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.922116159899272e-06,
+      "loss": 0.7002,
+      "step": 21233
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.921223331204279e-06,
+      "loss": 1.3027,
+      "step": 21234
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.920330557063668e-06,
+      "loss": 1.1123,
+      "step": 21235
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.919437837487038e-06,
+      "loss": 1.3555,
+      "step": 21236
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.918545172483969e-06,
+      "loss": 0.5906,
+      "step": 21237
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.917652562064058e-06,
+      "loss": 1.1211,
+      "step": 21238
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.916760006236891e-06,
+      "loss": 1.0801,
+      "step": 21239
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.915867505012047e-06,
+      "loss": 1.1797,
+      "step": 21240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.914975058399123e-06,
+      "loss": 1.4668,
+      "step": 21241
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.914082666407705e-06,
+      "loss": 0.9678,
+      "step": 21242
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.913190329047376e-06,
+      "loss": 0.8789,
+      "step": 21243
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.912298046327717e-06,
+      "loss": 0.8154,
+      "step": 21244
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.91140581825832e-06,
+      "loss": 0.9243,
+      "step": 21245
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910513644848762e-06,
+      "loss": 1.0273,
+      "step": 21246
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.909621526108635e-06,
+      "loss": 1.0381,
+      "step": 21247
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908729462047512e-06,
+      "loss": 0.9219,
+      "step": 21248
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.907837452674985e-06,
+      "loss": 1.0078,
+      "step": 21249
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90694549800063e-06,
+      "loss": 1.0479,
+      "step": 21250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.906053598034022e-06,
+      "loss": 0.6553,
+      "step": 21251
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.905161752784755e-06,
+      "loss": 1.0332,
+      "step": 21252
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.904269962262402e-06,
+      "loss": 1.2588,
+      "step": 21253
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.903378226476544e-06,
+      "loss": 1.0156,
+      "step": 21254
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.902486545436753e-06,
+      "loss": 1.1807,
+      "step": 21255
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.901594919152614e-06,
+      "loss": 1.1396,
+      "step": 21256
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.900703347633705e-06,
+      "loss": 0.8647,
+      "step": 21257
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.899811830889597e-06,
+      "loss": 1.7148,
+      "step": 21258
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8989203689298745e-06,
+      "loss": 0.9609,
+      "step": 21259
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.898028961764108e-06,
+      "loss": 1.1475,
+      "step": 21260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.897137609401871e-06,
+      "loss": 0.8955,
+      "step": 21261
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.89624631185274e-06,
+      "loss": 0.8154,
+      "step": 21262
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8953550691262944e-06,
+      "loss": 1.1914,
+      "step": 21263
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.894463881232101e-06,
+      "loss": 1.3438,
+      "step": 21264
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.893572748179736e-06,
+      "loss": 0.6113,
+      "step": 21265
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.892681669978762e-06,
+      "loss": 1.2139,
+      "step": 21266
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.891790646638766e-06,
+      "loss": 1.3516,
+      "step": 21267
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.890899678169312e-06,
+      "loss": 0.9082,
+      "step": 21268
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.890008764579967e-06,
+      "loss": 0.9014,
+      "step": 21269
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.889117905880307e-06,
+      "loss": 0.7026,
+      "step": 21270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.888227102079896e-06,
+      "loss": 1.0361,
+      "step": 21271
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.887336353188306e-06,
+      "loss": 1.1802,
+      "step": 21272
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.886445659215103e-06,
+      "loss": 1.0244,
+      "step": 21273
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.885555020169859e-06,
+      "loss": 0.7432,
+      "step": 21274
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.884664436062138e-06,
+      "loss": 0.8862,
+      "step": 21275
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.883773906901501e-06,
+      "loss": 0.7861,
+      "step": 21276
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.882883432697519e-06,
+      "loss": 1.0693,
+      "step": 21277
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.881993013459762e-06,
+      "loss": 0.8145,
+      "step": 21278
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8811026491977885e-06,
+      "loss": 0.8076,
+      "step": 21279
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.88021233992116e-06,
+      "loss": 1.4092,
+      "step": 21280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.879322085639446e-06,
+      "loss": 1.2051,
+      "step": 21281
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8784318863622036e-06,
+      "loss": 1.0479,
+      "step": 21282
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8775417420990015e-06,
+      "loss": 0.959,
+      "step": 21283
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.876651652859394e-06,
+      "loss": 0.8535,
+      "step": 21284
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.87576161865295e-06,
+      "loss": 1.1699,
+      "step": 21285
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874871639489226e-06,
+      "loss": 1.0986,
+      "step": 21286
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.873981715377777e-06,
+      "loss": 1.1689,
+      "step": 21287
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.873091846328167e-06,
+      "loss": 0.9067,
+      "step": 21288
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.87220203234996e-06,
+      "loss": 0.4709,
+      "step": 21289
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.871312273452706e-06,
+      "loss": 0.8457,
+      "step": 21290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.870422569645963e-06,
+      "loss": 0.9004,
+      "step": 21291
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.86953292093929e-06,
+      "loss": 0.8013,
+      "step": 21292
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.868643327342246e-06,
+      "loss": 0.7817,
+      "step": 21293
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.867753788864385e-06,
+      "loss": 0.4243,
+      "step": 21294
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8668643055152564e-06,
+      "loss": 0.8271,
+      "step": 21295
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8659748773044235e-06,
+      "loss": 1.4629,
+      "step": 21296
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8650855042414325e-06,
+      "loss": 1.2822,
+      "step": 21297
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.864196186335845e-06,
+      "loss": 0.8721,
+      "step": 21298
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8633069235972034e-06,
+      "loss": 0.9351,
+      "step": 21299
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.86241771603507e-06,
+      "loss": 0.8936,
+      "step": 21300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.861528563658993e-06,
+      "loss": 1.25,
+      "step": 21301
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.860639466478516e-06,
+      "loss": 0.55,
+      "step": 21302
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.859750424503197e-06,
+      "loss": 1.165,
+      "step": 21303
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.858861437742588e-06,
+      "loss": 0.8916,
+      "step": 21304
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.857972506206235e-06,
+      "loss": 1.0776,
+      "step": 21305
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.85708362990368e-06,
+      "loss": 0.8076,
+      "step": 21306
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.856194808844483e-06,
+      "loss": 1.0857,
+      "step": 21307
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.855306043038182e-06,
+      "loss": 1.126,
+      "step": 21308
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.854417332494329e-06,
+      "loss": 0.7988,
+      "step": 21309
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8535286772224655e-06,
+      "loss": 0.9214,
+      "step": 21310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.852640077232145e-06,
+      "loss": 0.9487,
+      "step": 21311
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.851751532532906e-06,
+      "loss": 0.9399,
+      "step": 21312
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8508630431342915e-06,
+      "loss": 0.9658,
+      "step": 21313
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849974609045849e-06,
+      "loss": 0.9473,
+      "step": 21314
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849086230277124e-06,
+      "loss": 1.04,
+      "step": 21315
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.848197906837656e-06,
+      "loss": 0.8906,
+      "step": 21316
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.847309638736984e-06,
+      "loss": 1.3301,
+      "step": 21317
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.846421425984652e-06,
+      "loss": 0.9971,
+      "step": 21318
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.845533268590205e-06,
+      "loss": 0.9492,
+      "step": 21319
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.844645166563181e-06,
+      "loss": 1.0557,
+      "step": 21320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.843757119913115e-06,
+      "loss": 1.3926,
+      "step": 21321
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842869128649554e-06,
+      "loss": 1.457,
+      "step": 21322
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.841981192782026e-06,
+      "loss": 1.1631,
+      "step": 21323
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.84109331232008e-06,
+      "loss": 0.9082,
+      "step": 21324
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.840205487273244e-06,
+      "loss": 1.1943,
+      "step": 21325
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.839317717651063e-06,
+      "loss": 1.103,
+      "step": 21326
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.838430003463069e-06,
+      "loss": 1.0684,
+      "step": 21327
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.837542344718795e-06,
+      "loss": 0.7861,
+      "step": 21328
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.836654741427778e-06,
+      "loss": 1.0088,
+      "step": 21329
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.835767193599556e-06,
+      "loss": 0.834,
+      "step": 21330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.834879701243659e-06,
+      "loss": 0.6914,
+      "step": 21331
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.833992264369617e-06,
+      "loss": 0.9277,
+      "step": 21332
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.833104882986966e-06,
+      "loss": 0.7939,
+      "step": 21333
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.832217557105243e-06,
+      "loss": 1.1426,
+      "step": 21334
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.831330286733974e-06,
+      "loss": 0.999,
+      "step": 21335
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.830443071882686e-06,
+      "loss": 0.7393,
+      "step": 21336
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.829555912560917e-06,
+      "loss": 0.708,
+      "step": 21337
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.828668808778194e-06,
+      "loss": 0.9199,
+      "step": 21338
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8277817605440404e-06,
+      "loss": 1.3691,
+      "step": 21339
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826894767867988e-06,
+      "loss": 0.9346,
+      "step": 21340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826007830759571e-06,
+      "loss": 0.9111,
+      "step": 21341
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.825120949228309e-06,
+      "loss": 1.2295,
+      "step": 21342
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.824234123283728e-06,
+      "loss": 1.1357,
+      "step": 21343
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.823347352935357e-06,
+      "loss": 0.7896,
+      "step": 21344
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.822460638192724e-06,
+      "loss": 0.8828,
+      "step": 21345
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.821573979065351e-06,
+      "loss": 0.7188,
+      "step": 21346
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8206873755627584e-06,
+      "loss": 0.9912,
+      "step": 21347
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.819800827694476e-06,
+      "loss": 0.8218,
+      "step": 21348
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.818914335470021e-06,
+      "loss": 0.7007,
+      "step": 21349
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.818027898898919e-06,
+      "loss": 0.8726,
+      "step": 21350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.817141517990695e-06,
+      "loss": 1.2988,
+      "step": 21351
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.816255192754867e-06,
+      "loss": 0.6133,
+      "step": 21352
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.815368923200955e-06,
+      "loss": 1.165,
+      "step": 21353
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.814482709338473e-06,
+      "loss": 0.6904,
+      "step": 21354
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8135965511769535e-06,
+      "loss": 1.0273,
+      "step": 21355
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.812710448725908e-06,
+      "loss": 1.0264,
+      "step": 21356
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.811824401994852e-06,
+      "loss": 0.9004,
+      "step": 21357
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.81093841099331e-06,
+      "loss": 1.3633,
+      "step": 21358
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.810052475730792e-06,
+      "loss": 0.7334,
+      "step": 21359
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.809166596216821e-06,
+      "loss": 1.1279,
+      "step": 21360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.808280772460906e-06,
+      "loss": 0.8384,
+      "step": 21361
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.807395004472569e-06,
+      "loss": 0.9941,
+      "step": 21362
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.806509292261322e-06,
+      "loss": 1.0615,
+      "step": 21363
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.805623635836675e-06,
+      "loss": 0.6658,
+      "step": 21364
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.804738035208144e-06,
+      "loss": 1.1943,
+      "step": 21365
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.803852490385246e-06,
+      "loss": 0.9189,
+      "step": 21366
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.80296700137749e-06,
+      "loss": 0.8623,
+      "step": 21367
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.802081568194383e-06,
+      "loss": 1.0166,
+      "step": 21368
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8011961908454454e-06,
+      "loss": 0.7119,
+      "step": 21369
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.800310869340178e-06,
+      "loss": 0.9229,
+      "step": 21370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799425603688099e-06,
+      "loss": 0.9629,
+      "step": 21371
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.798540393898711e-06,
+      "loss": 1.1328,
+      "step": 21372
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.797655239981528e-06,
+      "loss": 1.1172,
+      "step": 21373
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.796770141946056e-06,
+      "loss": 1.5234,
+      "step": 21374
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.795885099801797e-06,
+      "loss": 1.0391,
+      "step": 21375
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.795000113558262e-06,
+      "loss": 1.1123,
+      "step": 21376
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.794115183224964e-06,
+      "loss": 0.9473,
+      "step": 21377
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7932303088114e-06,
+      "loss": 0.9932,
+      "step": 21378
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.792345490327075e-06,
+      "loss": 1.2129,
+      "step": 21379
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7914607277814965e-06,
+      "loss": 0.5947,
+      "step": 21380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.790576021184169e-06,
+      "loss": 1.0127,
+      "step": 21381
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7896913705445955e-06,
+      "loss": 1.4453,
+      "step": 21382
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.788806775872275e-06,
+      "loss": 1.2363,
+      "step": 21383
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.787922237176713e-06,
+      "loss": 0.7686,
+      "step": 21384
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.787037754467408e-06,
+      "loss": 0.853,
+      "step": 21385
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.786153327753865e-06,
+      "loss": 1.167,
+      "step": 21386
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.785268957045579e-06,
+      "loss": 0.8594,
+      "step": 21387
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.784384642352056e-06,
+      "loss": 0.8877,
+      "step": 21388
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.783500383682791e-06,
+      "loss": 1.1055,
+      "step": 21389
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.782616181047278e-06,
+      "loss": 1.1021,
+      "step": 21390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.781732034455021e-06,
+      "loss": 0.9727,
+      "step": 21391
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.780847943915519e-06,
+      "loss": 1.1729,
+      "step": 21392
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.779963909438265e-06,
+      "loss": 1.3984,
+      "step": 21393
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.779079931032752e-06,
+      "loss": 0.8081,
+      "step": 21394
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.778196008708479e-06,
+      "loss": 0.9658,
+      "step": 21395
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.777312142474943e-06,
+      "loss": 0.8926,
+      "step": 21396
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.776428332341636e-06,
+      "loss": 1.0225,
+      "step": 21397
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775544578318047e-06,
+      "loss": 0.8413,
+      "step": 21398
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.774660880413679e-06,
+      "loss": 0.9814,
+      "step": 21399
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7737772386380154e-06,
+      "loss": 1.0371,
+      "step": 21400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772893653000549e-06,
+      "loss": 1.0254,
+      "step": 21401
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772010123510773e-06,
+      "loss": 0.877,
+      "step": 21402
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.771126650178182e-06,
+      "loss": 1.2656,
+      "step": 21403
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.770243233012261e-06,
+      "loss": 1.0596,
+      "step": 21404
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769359872022497e-06,
+      "loss": 0.9951,
+      "step": 21405
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.768476567218381e-06,
+      "loss": 1.0732,
+      "step": 21406
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.767593318609407e-06,
+      "loss": 1.0098,
+      "step": 21407
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7667101262050575e-06,
+      "loss": 1.1953,
+      "step": 21408
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7658269900148155e-06,
+      "loss": 1.3047,
+      "step": 21409
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.764943910048175e-06,
+      "loss": 0.9697,
+      "step": 21410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.764060886314616e-06,
+      "loss": 0.7178,
+      "step": 21411
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.763177918823627e-06,
+      "loss": 1.0098,
+      "step": 21412
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.762295007584689e-06,
+      "loss": 0.9756,
+      "step": 21413
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.761412152607292e-06,
+      "loss": 0.9189,
+      "step": 21414
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.760529353900916e-06,
+      "loss": 0.9463,
+      "step": 21415
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.759646611475038e-06,
+      "loss": 0.8271,
+      "step": 21416
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.758763925339147e-06,
+      "loss": 1.4434,
+      "step": 21417
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.757881295502725e-06,
+      "loss": 1.2061,
+      "step": 21418
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.756998721975252e-06,
+      "loss": 0.8857,
+      "step": 21419
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.756116204766203e-06,
+      "loss": 0.8442,
+      "step": 21420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.755233743885061e-06,
+      "loss": 0.9351,
+      "step": 21421
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.754351339341311e-06,
+      "loss": 1.0,
+      "step": 21422
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.753468991144425e-06,
+      "loss": 1.1943,
+      "step": 21423
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.752586699303878e-06,
+      "loss": 1.0859,
+      "step": 21424
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.751704463829157e-06,
+      "loss": 1.2012,
+      "step": 21425
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.750822284729727e-06,
+      "loss": 1.1113,
+      "step": 21426
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749940162015075e-06,
+      "loss": 0.8867,
+      "step": 21427
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749058095694667e-06,
+      "loss": 0.6692,
+      "step": 21428
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.748176085777988e-06,
+      "loss": 0.6523,
+      "step": 21429
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7472941322745045e-06,
+      "loss": 0.9785,
+      "step": 21430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.746412235193689e-06,
+      "loss": 1.0811,
+      "step": 21431
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.745530394545018e-06,
+      "loss": 0.8223,
+      "step": 21432
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.744648610337967e-06,
+      "loss": 0.9863,
+      "step": 21433
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.743766882582006e-06,
+      "loss": 0.9502,
+      "step": 21434
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7428852112866e-06,
+      "loss": 1.1094,
+      "step": 21435
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.742003596461228e-06,
+      "loss": 0.834,
+      "step": 21436
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.741122038115351e-06,
+      "loss": 1.0459,
+      "step": 21437
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.740240536258449e-06,
+      "loss": 1.0049,
+      "step": 21438
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.739359090899981e-06,
+      "loss": 1.1465,
+      "step": 21439
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.738477702049423e-06,
+      "loss": 1.1484,
+      "step": 21440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.737596369716241e-06,
+      "loss": 0.8447,
+      "step": 21441
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.736715093909891e-06,
+      "loss": 1.4297,
+      "step": 21442
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.735833874639856e-06,
+      "loss": 1.248,
+      "step": 21443
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.734952711915595e-06,
+      "loss": 0.9434,
+      "step": 21444
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.734071605746571e-06,
+      "loss": 0.9824,
+      "step": 21445
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.733190556142246e-06,
+      "loss": 0.8525,
+      "step": 21446
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.732309563112088e-06,
+      "loss": 0.978,
+      "step": 21447
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.731428626665565e-06,
+      "loss": 1.373,
+      "step": 21448
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.730547746812131e-06,
+      "loss": 0.8677,
+      "step": 21449
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.729666923561254e-06,
+      "loss": 0.6777,
+      "step": 21450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.728786156922395e-06,
+      "loss": 1.1885,
+      "step": 21451
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.727905446905008e-06,
+      "loss": 0.8291,
+      "step": 21452
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.727024793518559e-06,
+      "loss": 1.0088,
+      "step": 21453
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.726144196772511e-06,
+      "loss": 1.1709,
+      "step": 21454
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.725263656676319e-06,
+      "loss": 0.8818,
+      "step": 21455
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.724383173239442e-06,
+      "loss": 1.0488,
+      "step": 21456
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.723502746471333e-06,
+      "loss": 1.3438,
+      "step": 21457
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.722622376381455e-06,
+      "loss": 0.9658,
+      "step": 21458
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.721742062979265e-06,
+      "loss": 1.1191,
+      "step": 21459
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.720861806274215e-06,
+      "loss": 1.1943,
+      "step": 21460
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.719981606275766e-06,
+      "loss": 0.8311,
+      "step": 21461
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7191014629933704e-06,
+      "loss": 0.9287,
+      "step": 21462
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.718221376436477e-06,
+      "loss": 0.729,
+      "step": 21463
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.717341346614545e-06,
+      "loss": 0.958,
+      "step": 21464
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.716461373537028e-06,
+      "loss": 1.0444,
+      "step": 21465
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.715581457213377e-06,
+      "loss": 1.1582,
+      "step": 21466
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.714701597653041e-06,
+      "loss": 0.9268,
+      "step": 21467
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.713821794865473e-06,
+      "loss": 1.1152,
+      "step": 21468
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7129420488601275e-06,
+      "loss": 0.9902,
+      "step": 21469
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.71206235964645e-06,
+      "loss": 0.9248,
+      "step": 21470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.711182727233887e-06,
+      "loss": 0.9473,
+      "step": 21471
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7103031516318965e-06,
+      "loss": 0.8154,
+      "step": 21472
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.709423632849916e-06,
+      "loss": 1.1123,
+      "step": 21473
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.708544170897401e-06,
+      "loss": 0.9004,
+      "step": 21474
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7076647657837924e-06,
+      "loss": 1.4521,
+      "step": 21475
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.706785417518544e-06,
+      "loss": 1.2109,
+      "step": 21476
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.705906126111095e-06,
+      "loss": 1.1855,
+      "step": 21477
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.705026891570889e-06,
+      "loss": 0.8574,
+      "step": 21478
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704147713907374e-06,
+      "loss": 1.5195,
+      "step": 21479
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.703268593129998e-06,
+      "loss": 0.8945,
+      "step": 21480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.702389529248198e-06,
+      "loss": 0.8429,
+      "step": 21481
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.701510522271416e-06,
+      "loss": 0.8145,
+      "step": 21482
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.700631572209095e-06,
+      "loss": 1.0635,
+      "step": 21483
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.699752679070683e-06,
+      "loss": 0.7405,
+      "step": 21484
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.698873842865613e-06,
+      "loss": 1.5039,
+      "step": 21485
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.697995063603326e-06,
+      "loss": 0.8506,
+      "step": 21486
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.697116341293265e-06,
+      "loss": 1.1953,
+      "step": 21487
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.696237675944868e-06,
+      "loss": 1.1143,
+      "step": 21488
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.695359067567569e-06,
+      "loss": 1.2197,
+      "step": 21489
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.694480516170807e-06,
+      "loss": 0.877,
+      "step": 21490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.693602021764027e-06,
+      "loss": 0.8926,
+      "step": 21491
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.692723584356658e-06,
+      "loss": 1.0957,
+      "step": 21492
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.691845203958134e-06,
+      "loss": 0.9082,
+      "step": 21493
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.690966880577894e-06,
+      "loss": 1.291,
+      "step": 21494
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.690088614225376e-06,
+      "loss": 1.2061,
+      "step": 21495
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.68921040491001e-06,
+      "loss": 0.957,
+      "step": 21496
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6883322526412244e-06,
+      "loss": 0.8252,
+      "step": 21497
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.687454157428462e-06,
+      "loss": 1.0518,
+      "step": 21498
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.686576119281148e-06,
+      "loss": 1.0029,
+      "step": 21499
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.685698138208719e-06,
+      "loss": 1.1123,
+      "step": 21500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.684820214220599e-06,
+      "loss": 1.3125,
+      "step": 21501
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.683942347326226e-06,
+      "loss": 0.8301,
+      "step": 21502
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.683064537535027e-06,
+      "loss": 0.7847,
+      "step": 21503
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.682186784856426e-06,
+      "loss": 0.9316,
+      "step": 21504
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.681309089299856e-06,
+      "loss": 0.8281,
+      "step": 21505
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.680431450874748e-06,
+      "loss": 1.2988,
+      "step": 21506
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.679553869590527e-06,
+      "loss": 0.9502,
+      "step": 21507
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.678676345456614e-06,
+      "loss": 0.7529,
+      "step": 21508
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.677798878482441e-06,
+      "loss": 1.4014,
+      "step": 21509
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6769214686774345e-06,
+      "loss": 0.8691,
+      "step": 21510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.676044116051019e-06,
+      "loss": 1.1416,
+      "step": 21511
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.675166820612611e-06,
+      "loss": 0.9678,
+      "step": 21512
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.674289582371644e-06,
+      "loss": 1.0898,
+      "step": 21513
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.673412401337533e-06,
+      "loss": 1.0742,
+      "step": 21514
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.672535277519708e-06,
+      "loss": 1.1074,
+      "step": 21515
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.671658210927582e-06,
+      "loss": 1.0576,
+      "step": 21516
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.670781201570587e-06,
+      "loss": 0.7241,
+      "step": 21517
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.669904249458136e-06,
+      "loss": 1.1074,
+      "step": 21518
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.669027354599646e-06,
+      "loss": 1.1826,
+      "step": 21519
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.668150517004541e-06,
+      "loss": 1.4316,
+      "step": 21520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.667273736682245e-06,
+      "loss": 0.9131,
+      "step": 21521
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.666397013642169e-06,
+      "loss": 1.1484,
+      "step": 21522
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.665520347893728e-06,
+      "loss": 1.1338,
+      "step": 21523
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.664643739446347e-06,
+      "loss": 1.168,
+      "step": 21524
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.663767188309434e-06,
+      "loss": 1.3594,
+      "step": 21525
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.662890694492413e-06,
+      "loss": 1.5469,
+      "step": 21526
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.662014258004689e-06,
+      "loss": 0.873,
+      "step": 21527
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.661137878855686e-06,
+      "loss": 1.209,
+      "step": 21528
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.660261557054815e-06,
+      "loss": 1.1099,
+      "step": 21529
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.659385292611479e-06,
+      "loss": 1.0039,
+      "step": 21530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6585090855351065e-06,
+      "loss": 0.9043,
+      "step": 21531
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.657632935835102e-06,
+      "loss": 0.8594,
+      "step": 21532
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.656756843520879e-06,
+      "loss": 1.1406,
+      "step": 21533
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65588080860184e-06,
+      "loss": 0.8955,
+      "step": 21534
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.655004831087403e-06,
+      "loss": 1.1797,
+      "step": 21535
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.654128910986978e-06,
+      "loss": 1.1731,
+      "step": 21536
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6532530483099725e-06,
+      "loss": 1.1416,
+      "step": 21537
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6523772430657885e-06,
+      "loss": 1.1113,
+      "step": 21538
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6515014952638434e-06,
+      "loss": 0.9912,
+      "step": 21539
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.650625804913535e-06,
+      "loss": 1.1631,
+      "step": 21540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.649750172024275e-06,
+      "loss": 0.8623,
+      "step": 21541
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.648874596605471e-06,
+      "loss": 0.6631,
+      "step": 21542
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6479990786665265e-06,
+      "loss": 1.1475,
+      "step": 21543
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6471236182168445e-06,
+      "loss": 1.1089,
+      "step": 21544
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6462482152658216e-06,
+      "loss": 0.9775,
+      "step": 21545
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.645372869822876e-06,
+      "loss": 0.6992,
+      "step": 21546
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.644497581897402e-06,
+      "loss": 0.7644,
+      "step": 21547
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.643622351498801e-06,
+      "loss": 0.7666,
+      "step": 21548
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.642747178636479e-06,
+      "loss": 1.0703,
+      "step": 21549
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.641872063319833e-06,
+      "loss": 0.9414,
+      "step": 21550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.640997005558262e-06,
+      "loss": 1.1104,
+      "step": 21551
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.640122005361166e-06,
+      "loss": 0.6709,
+      "step": 21552
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.639247062737951e-06,
+      "loss": 1.0078,
+      "step": 21553
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638372177698007e-06,
+      "loss": 1.2168,
+      "step": 21554
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.637497350250737e-06,
+      "loss": 0.9546,
+      "step": 21555
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.636622580405528e-06,
+      "loss": 0.9912,
+      "step": 21556
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63574786817179e-06,
+      "loss": 0.6982,
+      "step": 21557
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.634873213558915e-06,
+      "loss": 0.9463,
+      "step": 21558
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.633998616576291e-06,
+      "loss": 0.594,
+      "step": 21559
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.633124077233322e-06,
+      "loss": 0.8325,
+      "step": 21560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.632249595539393e-06,
+      "loss": 1.0791,
+      "step": 21561
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.631375171503907e-06,
+      "loss": 0.9619,
+      "step": 21562
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.630500805136248e-06,
+      "loss": 0.9414,
+      "step": 21563
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6296264964458155e-06,
+      "loss": 1.0156,
+      "step": 21564
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.628752245441998e-06,
+      "loss": 0.8799,
+      "step": 21565
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.627878052134182e-06,
+      "loss": 1.0156,
+      "step": 21566
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.627003916531761e-06,
+      "loss": 1.1611,
+      "step": 21567
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.626129838644129e-06,
+      "loss": 1.3652,
+      "step": 21568
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.625255818480672e-06,
+      "loss": 0.6221,
+      "step": 21569
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.624381856050773e-06,
+      "loss": 1.0029,
+      "step": 21570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.623507951363826e-06,
+      "loss": 1.1162,
+      "step": 21571
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.62263410442922e-06,
+      "loss": 1.1016,
+      "step": 21572
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.621760315256337e-06,
+      "loss": 0.605,
+      "step": 21573
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6208865838545625e-06,
+      "loss": 0.9727,
+      "step": 21574
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.620012910233286e-06,
+      "loss": 0.9263,
+      "step": 21575
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.619139294401886e-06,
+      "loss": 1.2139,
+      "step": 21576
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.618265736369753e-06,
+      "loss": 0.8281,
+      "step": 21577
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.617392236146267e-06,
+      "loss": 0.9785,
+      "step": 21578
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.616518793740812e-06,
+      "loss": 1.0488,
+      "step": 21579
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.615645409162771e-06,
+      "loss": 1.377,
+      "step": 21580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6147720824215204e-06,
+      "loss": 0.7017,
+      "step": 21581
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.613898813526445e-06,
+      "loss": 0.7852,
+      "step": 21582
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.613025602486929e-06,
+      "loss": 0.7939,
+      "step": 21583
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61215244931235e-06,
+      "loss": 0.6284,
+      "step": 21584
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.611279354012079e-06,
+      "loss": 0.6138,
+      "step": 21585
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.610406316595506e-06,
+      "loss": 0.9795,
+      "step": 21586
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.609533337071999e-06,
+      "loss": 1.3057,
+      "step": 21587
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.608660415450945e-06,
+      "loss": 1.3291,
+      "step": 21588
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.607787551741712e-06,
+      "loss": 1.001,
+      "step": 21589
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.606914745953682e-06,
+      "loss": 0.8154,
+      "step": 21590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.606041998096228e-06,
+      "loss": 0.8438,
+      "step": 21591
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.605169308178722e-06,
+      "loss": 0.9961,
+      "step": 21592
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.60429667621054e-06,
+      "loss": 1.1875,
+      "step": 21593
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6034241022010604e-06,
+      "loss": 1.0361,
+      "step": 21594
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6025515861596514e-06,
+      "loss": 0.6892,
+      "step": 21595
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6016791280956826e-06,
+      "loss": 1.1768,
+      "step": 21596
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.600806728018529e-06,
+      "loss": 1.0015,
+      "step": 21597
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.599934385937564e-06,
+      "loss": 1.2246,
+      "step": 21598
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.599062101862157e-06,
+      "loss": 1.2861,
+      "step": 21599
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.598189875801672e-06,
+      "loss": 1.1475,
+      "step": 21600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5973177077654865e-06,
+      "loss": 1.166,
+      "step": 21601
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.59644559776296e-06,
+      "loss": 1.0176,
+      "step": 21602
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.595573545803469e-06,
+      "loss": 0.9326,
+      "step": 21603
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.594701551896375e-06,
+      "loss": 0.9194,
+      "step": 21604
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.593829616051049e-06,
+      "loss": 0.9941,
+      "step": 21605
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.592957738276854e-06,
+      "loss": 0.6758,
+      "step": 21606
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.592085918583154e-06,
+      "loss": 1.001,
+      "step": 21607
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.591214156979314e-06,
+      "loss": 1.0537,
+      "step": 21608
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.590342453474706e-06,
+      "loss": 0.7822,
+      "step": 21609
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.589470808078686e-06,
+      "loss": 1.3223,
+      "step": 21610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.588599220800614e-06,
+      "loss": 0.8179,
+      "step": 21611
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.587727691649861e-06,
+      "loss": 0.7944,
+      "step": 21612
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.586856220635781e-06,
+      "loss": 1.0654,
+      "step": 21613
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5859848077677425e-06,
+      "loss": 0.9556,
+      "step": 21614
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.585113453055098e-06,
+      "loss": 0.9917,
+      "step": 21615
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.584242156507214e-06,
+      "loss": 0.8828,
+      "step": 21616
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.583370918133447e-06,
+      "loss": 1.0259,
+      "step": 21617
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.582499737943152e-06,
+      "loss": 1.2051,
+      "step": 21618
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.581628615945689e-06,
+      "loss": 1.0713,
+      "step": 21619
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.580757552150421e-06,
+      "loss": 0.7852,
+      "step": 21620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5798865465667e-06,
+      "loss": 0.5903,
+      "step": 21621
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.579015599203878e-06,
+      "loss": 0.665,
+      "step": 21622
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.578144710071315e-06,
+      "loss": 0.8086,
+      "step": 21623
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.577273879178369e-06,
+      "loss": 0.8545,
+      "step": 21624
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5764031065343905e-06,
+      "loss": 1.2402,
+      "step": 21625
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.575532392148729e-06,
+      "loss": 0.791,
+      "step": 21626
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5746617360307445e-06,
+      "loss": 0.5801,
+      "step": 21627
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.573791138189784e-06,
+      "loss": 0.8828,
+      "step": 21628
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5729205986352e-06,
+      "loss": 1.0762,
+      "step": 21629
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.572050117376349e-06,
+      "loss": 0.915,
+      "step": 21630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.571179694422576e-06,
+      "loss": 0.6855,
+      "step": 21631
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.570309329783233e-06,
+      "loss": 0.5291,
+      "step": 21632
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.56943902346766e-06,
+      "loss": 1.0469,
+      "step": 21633
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5685687754852225e-06,
+      "loss": 1.2637,
+      "step": 21634
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567698585845258e-06,
+      "loss": 1.2031,
+      "step": 21635
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.566828454557117e-06,
+      "loss": 0.8262,
+      "step": 21636
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.565958381630139e-06,
+      "loss": 0.9141,
+      "step": 21637
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.565088367073675e-06,
+      "loss": 1.126,
+      "step": 21638
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.564218410897076e-06,
+      "loss": 1.2646,
+      "step": 21639
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.563348513109677e-06,
+      "loss": 0.8137,
+      "step": 21640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.562478673720832e-06,
+      "loss": 0.8608,
+      "step": 21641
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5616088927398775e-06,
+      "loss": 0.9521,
+      "step": 21642
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.560739170176158e-06,
+      "loss": 0.6719,
+      "step": 21643
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.559869506039009e-06,
+      "loss": 0.917,
+      "step": 21644
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.558999900337786e-06,
+      "loss": 0.9019,
+      "step": 21645
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.558130353081823e-06,
+      "loss": 0.9141,
+      "step": 21646
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5572608642804576e-06,
+      "loss": 0.8203,
+      "step": 21647
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.556391433943034e-06,
+      "loss": 1.1123,
+      "step": 21648
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.555522062078886e-06,
+      "loss": 0.8906,
+      "step": 21649
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.554652748697361e-06,
+      "loss": 0.9639,
+      "step": 21650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.553783493807786e-06,
+      "loss": 1.0215,
+      "step": 21651
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.55291429741951e-06,
+      "loss": 0.9624,
+      "step": 21652
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.55204515954186e-06,
+      "loss": 0.9941,
+      "step": 21653
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.551176080184173e-06,
+      "loss": 0.9521,
+      "step": 21654
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5503070593557875e-06,
+      "loss": 1.2617,
+      "step": 21655
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549438097066039e-06,
+      "loss": 1.2773,
+      "step": 21656
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5485691933242615e-06,
+      "loss": 0.8066,
+      "step": 21657
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.547700348139782e-06,
+      "loss": 0.9438,
+      "step": 21658
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.546831561521939e-06,
+      "loss": 0.6887,
+      "step": 21659
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.545962833480067e-06,
+      "loss": 1.168,
+      "step": 21660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5450941640234945e-06,
+      "loss": 1.0088,
+      "step": 21661
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.544225553161549e-06,
+      "loss": 0.9863,
+      "step": 21662
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.543357000903568e-06,
+      "loss": 1.0117,
+      "step": 21663
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.542488507258872e-06,
+      "loss": 0.9395,
+      "step": 21664
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5416200722368e-06,
+      "loss": 1.0674,
+      "step": 21665
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5407516958466725e-06,
+      "loss": 0.9951,
+      "step": 21666
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.539883378097824e-06,
+      "loss": 0.9087,
+      "step": 21667
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.539015118999578e-06,
+      "loss": 0.8691,
+      "step": 21668
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.538146918561257e-06,
+      "loss": 1.125,
+      "step": 21669
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5372787767921925e-06,
+      "loss": 1.2041,
+      "step": 21670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.536410693701711e-06,
+      "loss": 1.1846,
+      "step": 21671
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5355426692991345e-06,
+      "loss": 1.1748,
+      "step": 21672
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.534674703593785e-06,
+      "loss": 0.875,
+      "step": 21673
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.533806796594989e-06,
+      "loss": 1.082,
+      "step": 21674
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532938948312065e-06,
+      "loss": 1.1846,
+      "step": 21675
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532071158754343e-06,
+      "loss": 1.1494,
+      "step": 21676
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5312034279311354e-06,
+      "loss": 0.9312,
+      "step": 21677
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.530335755851773e-06,
+      "loss": 0.8184,
+      "step": 21678
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.529468142525568e-06,
+      "loss": 1.0068,
+      "step": 21679
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.528600587961839e-06,
+      "loss": 0.7764,
+      "step": 21680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.527733092169909e-06,
+      "loss": 0.8999,
+      "step": 21681
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.526865655159099e-06,
+      "loss": 0.9229,
+      "step": 21682
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525998276938725e-06,
+      "loss": 0.6455,
+      "step": 21683
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525130957518097e-06,
+      "loss": 0.8984,
+      "step": 21684
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.524263696906537e-06,
+      "loss": 0.9951,
+      "step": 21685
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.523396495113365e-06,
+      "loss": 0.8481,
+      "step": 21686
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.522529352147891e-06,
+      "loss": 0.7817,
+      "step": 21687
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.521662268019427e-06,
+      "loss": 1.2891,
+      "step": 21688
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520795242737293e-06,
+      "loss": 0.7393,
+      "step": 21689
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.519928276310798e-06,
+      "loss": 0.6948,
+      "step": 21690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.519061368749259e-06,
+      "loss": 1.083,
+      "step": 21691
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5181945200619814e-06,
+      "loss": 1.3477,
+      "step": 21692
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.517327730258284e-06,
+      "loss": 0.8506,
+      "step": 21693
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.516460999347474e-06,
+      "loss": 1.4043,
+      "step": 21694
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.515594327338857e-06,
+      "loss": 1.1475,
+      "step": 21695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.514727714241748e-06,
+      "loss": 1.4502,
+      "step": 21696
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.513861160065458e-06,
+      "loss": 1.1406,
+      "step": 21697
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5129946648192926e-06,
+      "loss": 0.9971,
+      "step": 21698
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.512128228512555e-06,
+      "loss": 0.6353,
+      "step": 21699
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.511261851154555e-06,
+      "loss": 1.0098,
+      "step": 21700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.510395532754603e-06,
+      "loss": 1.002,
+      "step": 21701
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.509529273322003e-06,
+      "loss": 1.0605,
+      "step": 21702
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.508663072866054e-06,
+      "loss": 1.1094,
+      "step": 21703
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.507796931396069e-06,
+      "loss": 0.7627,
+      "step": 21704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.506930848921348e-06,
+      "loss": 1.0088,
+      "step": 21705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.50606482545119e-06,
+      "loss": 0.8096,
+      "step": 21706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5051988609949005e-06,
+      "loss": 0.9277,
+      "step": 21707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5043329555617866e-06,
+      "loss": 1.5098,
+      "step": 21708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.503467109161146e-06,
+      "loss": 1.0784,
+      "step": 21709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502601321802273e-06,
+      "loss": 1.334,
+      "step": 21710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.501735593494475e-06,
+      "loss": 1.0415,
+      "step": 21711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5008699242470525e-06,
+      "loss": 0.9165,
+      "step": 21712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5000043140693e-06,
+      "loss": 0.9365,
+      "step": 21713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4991387629705156e-06,
+      "loss": 1.1953,
+      "step": 21714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.498273270959999e-06,
+      "loss": 1.0645,
+      "step": 21715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4974078380470435e-06,
+      "loss": 0.832,
+      "step": 21716
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.49654246424095e-06,
+      "loss": 1.4043,
+      "step": 21717
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.49567714955101e-06,
+      "loss": 0.9561,
+      "step": 21718
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.494811893986522e-06,
+      "loss": 0.9268,
+      "step": 21719
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493946697556779e-06,
+      "loss": 0.8896,
+      "step": 21720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493081560271066e-06,
+      "loss": 0.8833,
+      "step": 21721
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4922164821386926e-06,
+      "loss": 1.252,
+      "step": 21722
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.491351463168941e-06,
+      "loss": 0.8311,
+      "step": 21723
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.490486503371106e-06,
+      "loss": 1.0898,
+      "step": 21724
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.489621602754473e-06,
+      "loss": 0.8223,
+      "step": 21725
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.488756761328336e-06,
+      "loss": 1.2842,
+      "step": 21726
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.487891979101988e-06,
+      "loss": 0.3765,
+      "step": 21727
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.487027256084713e-06,
+      "loss": 1.1689,
+      "step": 21728
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.486162592285807e-06,
+      "loss": 0.7275,
+      "step": 21729
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485297987714551e-06,
+      "loss": 1.0283,
+      "step": 21730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.484433442380232e-06,
+      "loss": 0.8516,
+      "step": 21731
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4835689562921385e-06,
+      "loss": 1.1538,
+      "step": 21732
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4827045294595595e-06,
+      "loss": 1.2168,
+      "step": 21733
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.481840161891779e-06,
+      "loss": 1.0781,
+      "step": 21734
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480975853598079e-06,
+      "loss": 0.8691,
+      "step": 21735
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480111604587743e-06,
+      "loss": 0.9043,
+      "step": 21736
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.479247414870056e-06,
+      "loss": 0.751,
+      "step": 21737
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.478383284454305e-06,
+      "loss": 0.6973,
+      "step": 21738
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4775192133497645e-06,
+      "loss": 1.0293,
+      "step": 21739
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.476655201565724e-06,
+      "loss": 1.0107,
+      "step": 21740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4757912491114605e-06,
+      "loss": 1.1289,
+      "step": 21741
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.474927355996252e-06,
+      "loss": 0.9102,
+      "step": 21742
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.474063522229378e-06,
+      "loss": 1.0742,
+      "step": 21743
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.473199747820124e-06,
+      "loss": 1.2324,
+      "step": 21744
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.472336032777766e-06,
+      "loss": 1.1387,
+      "step": 21745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.471472377111574e-06,
+      "loss": 0.9883,
+      "step": 21746
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.470608780830832e-06,
+      "loss": 1.0098,
+      "step": 21747
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.469745243944818e-06,
+      "loss": 1.2725,
+      "step": 21748
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.468881766462807e-06,
+      "loss": 0.9409,
+      "step": 21749
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.468018348394067e-06,
+      "loss": 0.9238,
+      "step": 21750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.467154989747883e-06,
+      "loss": 0.7881,
+      "step": 21751
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46629169053352e-06,
+      "loss": 0.9321,
+      "step": 21752
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.465428450760259e-06,
+      "loss": 1.3594,
+      "step": 21753
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.464565270437363e-06,
+      "loss": 1.1543,
+      "step": 21754
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.463702149574115e-06,
+      "loss": 1.1074,
+      "step": 21755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.462839088179781e-06,
+      "loss": 1.3359,
+      "step": 21756
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.461976086263628e-06,
+      "loss": 1.1401,
+      "step": 21757
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.461113143834928e-06,
+      "loss": 0.9302,
+      "step": 21758
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.460250260902957e-06,
+      "loss": 1.0527,
+      "step": 21759
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.45938743747698e-06,
+      "loss": 0.8018,
+      "step": 21760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.458524673566259e-06,
+      "loss": 1.001,
+      "step": 21761
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4576619691800675e-06,
+      "loss": 0.8994,
+      "step": 21762
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.456799324327674e-06,
+      "loss": 0.9609,
+      "step": 21763
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.455936739018343e-06,
+      "loss": 1.0547,
+      "step": 21764
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4550742132613345e-06,
+      "loss": 0.8579,
+      "step": 21765
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.454211747065923e-06,
+      "loss": 1.1328,
+      "step": 21766
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453349340441368e-06,
+      "loss": 0.9238,
+      "step": 21767
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.452486993396928e-06,
+      "loss": 1.0059,
+      "step": 21768
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.451624705941873e-06,
+      "loss": 1.2012,
+      "step": 21769
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.450762478085465e-06,
+      "loss": 0.8276,
+      "step": 21770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.449900309836966e-06,
+      "loss": 1.083,
+      "step": 21771
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.449038201205631e-06,
+      "loss": 0.9297,
+      "step": 21772
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.448176152200725e-06,
+      "loss": 0.7559,
+      "step": 21773
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.447314162831511e-06,
+      "loss": 1.0127,
+      "step": 21774
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.446452233107245e-06,
+      "loss": 0.9468,
+      "step": 21775
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.445590363037182e-06,
+      "loss": 0.9998,
+      "step": 21776
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.444728552630586e-06,
+      "loss": 0.9194,
+      "step": 21777
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.443866801896709e-06,
+      "loss": 1.0186,
+      "step": 21778
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.443005110844813e-06,
+      "loss": 0.8936,
+      "step": 21779
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.442143479484148e-06,
+      "loss": 0.9092,
+      "step": 21780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.441281907823976e-06,
+      "loss": 1.0195,
+      "step": 21781
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.440420395873547e-06,
+      "loss": 0.8223,
+      "step": 21782
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.439558943642114e-06,
+      "loss": 0.8516,
+      "step": 21783
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.438697551138932e-06,
+      "loss": 0.811,
+      "step": 21784
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.437836218373257e-06,
+      "loss": 1.1426,
+      "step": 21785
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.436974945354339e-06,
+      "loss": 1.2051,
+      "step": 21786
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.436113732091426e-06,
+      "loss": 0.9346,
+      "step": 21787
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.435252578593772e-06,
+      "loss": 1.001,
+      "step": 21788
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.434391484870631e-06,
+      "loss": 0.8218,
+      "step": 21789
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.433530450931247e-06,
+      "loss": 0.8691,
+      "step": 21790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.432669476784868e-06,
+      "loss": 0.6152,
+      "step": 21791
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.431808562440747e-06,
+      "loss": 1.5195,
+      "step": 21792
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.430947707908131e-06,
+      "loss": 1.04,
+      "step": 21793
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.43008691319626e-06,
+      "loss": 1.3154,
+      "step": 21794
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.429226178314388e-06,
+      "loss": 0.96,
+      "step": 21795
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.42836550327176e-06,
+      "loss": 0.5569,
+      "step": 21796
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.42750488807762e-06,
+      "loss": 0.9521,
+      "step": 21797
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.426644332741209e-06,
+      "loss": 0.9893,
+      "step": 21798
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.425783837271772e-06,
+      "loss": 0.8232,
+      "step": 21799
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424923401678559e-06,
+      "loss": 0.8315,
+      "step": 21800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424063025970807e-06,
+      "loss": 0.9165,
+      "step": 21801
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.423202710157755e-06,
+      "loss": 0.875,
+      "step": 21802
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.422342454248649e-06,
+      "loss": 1.0271,
+      "step": 21803
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.421482258252726e-06,
+      "loss": 1.0859,
+      "step": 21804
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.420622122179231e-06,
+      "loss": 1.3496,
+      "step": 21805
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.419762046037397e-06,
+      "loss": 0.7441,
+      "step": 21806
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4189020298364685e-06,
+      "loss": 1.0,
+      "step": 21807
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.418042073585681e-06,
+      "loss": 0.9746,
+      "step": 21808
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4171821772942684e-06,
+      "loss": 0.8032,
+      "step": 21809
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.416322340971469e-06,
+      "loss": 0.9883,
+      "step": 21810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4154625646265245e-06,
+      "loss": 1.2471,
+      "step": 21811
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.414602848268667e-06,
+      "loss": 0.9189,
+      "step": 21812
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.413743191907127e-06,
+      "loss": 1.2012,
+      "step": 21813
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.412883595551142e-06,
+      "loss": 0.916,
+      "step": 21814
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4120240592099485e-06,
+      "loss": 1.1162,
+      "step": 21815
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.411164582892776e-06,
+      "loss": 0.7363,
+      "step": 21816
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.410305166608855e-06,
+      "loss": 0.9678,
+      "step": 21817
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.409445810367421e-06,
+      "loss": 1.2119,
+      "step": 21818
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4085865141776995e-06,
+      "loss": 1.1943,
+      "step": 21819
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.407727278048926e-06,
+      "loss": 0.7812,
+      "step": 21820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.40686810199033e-06,
+      "loss": 1.0791,
+      "step": 21821
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.40600898601114e-06,
+      "loss": 0.8467,
+      "step": 21822
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.405149930120583e-06,
+      "loss": 1.0664,
+      "step": 21823
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.404290934327883e-06,
+      "loss": 0.877,
+      "step": 21824
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.40343199864227e-06,
+      "loss": 1.0049,
+      "step": 21825
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402573123072975e-06,
+      "loss": 0.8672,
+      "step": 21826
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.401714307629221e-06,
+      "loss": 0.6572,
+      "step": 21827
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.400855552320228e-06,
+      "loss": 1.1016,
+      "step": 21828
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.399996857155227e-06,
+      "loss": 1.2334,
+      "step": 21829
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3991382221434365e-06,
+      "loss": 0.9385,
+      "step": 21830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3982796472940825e-06,
+      "loss": 0.6377,
+      "step": 21831
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3974211326163906e-06,
+      "loss": 0.874,
+      "step": 21832
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.396562678119581e-06,
+      "loss": 1.1631,
+      "step": 21833
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.395704283812872e-06,
+      "loss": 1.0439,
+      "step": 21834
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.394845949705477e-06,
+      "loss": 0.811,
+      "step": 21835
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.393987675806634e-06,
+      "loss": 0.9961,
+      "step": 21836
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3931294621255505e-06,
+      "loss": 0.8037,
+      "step": 21837
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3922713086714455e-06,
+      "loss": 1.1758,
+      "step": 21838
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3914132154535414e-06,
+      "loss": 0.8672,
+      "step": 21839
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.39055518248105e-06,
+      "loss": 0.8887,
+      "step": 21840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.389697209763194e-06,
+      "loss": 0.8691,
+      "step": 21841
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.388839297309182e-06,
+      "loss": 0.6929,
+      "step": 21842
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.387981445128239e-06,
+      "loss": 0.7178,
+      "step": 21843
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.387123653229574e-06,
+      "loss": 0.6953,
+      "step": 21844
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.386265921622397e-06,
+      "loss": 0.8057,
+      "step": 21845
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.385408250315927e-06,
+      "loss": 0.9666,
+      "step": 21846
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3845506393193785e-06,
+      "loss": 0.7251,
+      "step": 21847
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.383693088641962e-06,
+      "loss": 1.4062,
+      "step": 21848
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.382835598292882e-06,
+      "loss": 1.042,
+      "step": 21849
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.381978168281357e-06,
+      "loss": 1.0088,
+      "step": 21850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.381120798616599e-06,
+      "loss": 0.9531,
+      "step": 21851
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.380263489307815e-06,
+      "loss": 0.8594,
+      "step": 21852
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.379406240364208e-06,
+      "loss": 1.1885,
+      "step": 21853
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.378549051794996e-06,
+      "loss": 1.3203,
+      "step": 21854
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.377691923609381e-06,
+      "loss": 0.9341,
+      "step": 21855
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.376834855816568e-06,
+      "loss": 0.8198,
+      "step": 21856
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.375977848425764e-06,
+      "loss": 1.2412,
+      "step": 21857
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.375120901446182e-06,
+      "loss": 0.7529,
+      "step": 21858
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.374264014887022e-06,
+      "loss": 0.9219,
+      "step": 21859
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3734071887574835e-06,
+      "loss": 0.9126,
+      "step": 21860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3725504230667755e-06,
+      "loss": 1.1631,
+      "step": 21861
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.371693717824104e-06,
+      "loss": 0.8467,
+      "step": 21862
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.370837073038668e-06,
+      "loss": 0.6797,
+      "step": 21863
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3699804887196655e-06,
+      "loss": 1.2275,
+      "step": 21864
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.369123964876304e-06,
+      "loss": 0.8398,
+      "step": 21865
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.368267501517778e-06,
+      "loss": 1.1953,
+      "step": 21866
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.367411098653295e-06,
+      "loss": 0.9102,
+      "step": 21867
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3665547562920456e-06,
+      "loss": 1.0771,
+      "step": 21868
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.365698474443235e-06,
+      "loss": 1.2051,
+      "step": 21869
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.36484225311606e-06,
+      "loss": 1.0,
+      "step": 21870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.363986092319712e-06,
+      "loss": 1.2349,
+      "step": 21871
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.363129992063392e-06,
+      "loss": 1.1426,
+      "step": 21872
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.362273952356299e-06,
+      "loss": 1.5352,
+      "step": 21873
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.361417973207625e-06,
+      "loss": 0.6792,
+      "step": 21874
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.360562054626561e-06,
+      "loss": 1.1426,
+      "step": 21875
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.359706196622305e-06,
+      "loss": 1.2266,
+      "step": 21876
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3588503992040545e-06,
+      "loss": 1.2598,
+      "step": 21877
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.357994662380996e-06,
+      "loss": 1.0693,
+      "step": 21878
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.35713898616232e-06,
+      "loss": 1.2764,
+      "step": 21879
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.356283370557226e-06,
+      "loss": 1.2148,
+      "step": 21880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.355427815574895e-06,
+      "loss": 0.7158,
+      "step": 21881
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.354572321224525e-06,
+      "loss": 1.0049,
+      "step": 21882
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.353716887515299e-06,
+      "loss": 1.0518,
+      "step": 21883
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.352861514456414e-06,
+      "loss": 1.2324,
+      "step": 21884
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3520062020570505e-06,
+      "loss": 1.0527,
+      "step": 21885
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.351150950326396e-06,
+      "loss": 1.0898,
+      "step": 21886
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.350295759273641e-06,
+      "loss": 1.1553,
+      "step": 21887
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3494406289079725e-06,
+      "loss": 1.0039,
+      "step": 21888
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.348585559238575e-06,
+      "loss": 0.835,
+      "step": 21889
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347730550274628e-06,
+      "loss": 0.8945,
+      "step": 21890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.346875602025325e-06,
+      "loss": 1.4072,
+      "step": 21891
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.346020714499839e-06,
+      "loss": 0.8535,
+      "step": 21892
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.345165887707363e-06,
+      "loss": 1.3613,
+      "step": 21893
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.344311121657071e-06,
+      "loss": 1.0703,
+      "step": 21894
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.343456416358153e-06,
+      "loss": 0.9658,
+      "step": 21895
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.342601771819784e-06,
+      "loss": 1.1738,
+      "step": 21896
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.341747188051142e-06,
+      "loss": 1.2266,
+      "step": 21897
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.34089266506141e-06,
+      "loss": 0.7314,
+      "step": 21898
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.340038202859772e-06,
+      "loss": 1.0732,
+      "step": 21899
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3391838014554e-06,
+      "loss": 1.2188,
+      "step": 21900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.338329460857471e-06,
+      "loss": 0.7944,
+      "step": 21901
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.337475181075162e-06,
+      "loss": 0.7627,
+      "step": 21902
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.336620962117657e-06,
+      "loss": 0.791,
+      "step": 21903
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.335766803994126e-06,
+      "loss": 1.2441,
+      "step": 21904
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.334912706713741e-06,
+      "loss": 0.7632,
+      "step": 21905
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.334058670285682e-06,
+      "loss": 0.8569,
+      "step": 21906
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.333204694719117e-06,
+      "loss": 0.6553,
+      "step": 21907
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3323507800232265e-06,
+      "loss": 1.0503,
+      "step": 21908
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.331496926207176e-06,
+      "loss": 1.0029,
+      "step": 21909
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.330643133280144e-06,
+      "loss": 0.8066,
+      "step": 21910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.329789401251297e-06,
+      "loss": 0.832,
+      "step": 21911
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.328935730129799e-06,
+      "loss": 0.8379,
+      "step": 21912
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.328082119924836e-06,
+      "loss": 0.9932,
+      "step": 21913
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.327228570645566e-06,
+      "loss": 1.0879,
+      "step": 21914
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3263750823011615e-06,
+      "loss": 0.7886,
+      "step": 21915
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.325521654900785e-06,
+      "loss": 1.0293,
+      "step": 21916
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.324668288453612e-06,
+      "loss": 0.8174,
+      "step": 21917
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.323814982968799e-06,
+      "loss": 0.8267,
+      "step": 21918
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.322961738455519e-06,
+      "loss": 1.2246,
+      "step": 21919
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.322108554922938e-06,
+      "loss": 1.0479,
+      "step": 21920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.321255432380219e-06,
+      "loss": 1.1323,
+      "step": 21921
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.320402370836525e-06,
+      "loss": 0.7188,
+      "step": 21922
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.319549370301013e-06,
+      "loss": 0.7075,
+      "step": 21923
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.318696430782858e-06,
+      "loss": 1.3027,
+      "step": 21924
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3178435522912175e-06,
+      "loss": 0.8179,
+      "step": 21925
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.31699073483525e-06,
+      "loss": 0.9326,
+      "step": 21926
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.316137978424116e-06,
+      "loss": 1.1387,
+      "step": 21927
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.315285283066974e-06,
+      "loss": 1.0869,
+      "step": 21928
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3144326487729925e-06,
+      "loss": 1.4531,
+      "step": 21929
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3135800755513205e-06,
+      "loss": 0.791,
+      "step": 21930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.312727563411122e-06,
+      "loss": 0.8682,
+      "step": 21931
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.311875112361552e-06,
+      "loss": 0.9038,
+      "step": 21932
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.311022722411764e-06,
+      "loss": 0.791,
+      "step": 21933
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3101703935709175e-06,
+      "loss": 1.3145,
+      "step": 21934
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3093181258481705e-06,
+      "loss": 0.9229,
+      "step": 21935
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.308465919252676e-06,
+      "loss": 0.7803,
+      "step": 21936
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.307613773793583e-06,
+      "loss": 0.8994,
+      "step": 21937
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.306761689480049e-06,
+      "loss": 0.8398,
+      "step": 21938
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.305909666321231e-06,
+      "loss": 1.1436,
+      "step": 21939
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.305057704326275e-06,
+      "loss": 0.8232,
+      "step": 21940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.304205803504332e-06,
+      "loss": 0.7515,
+      "step": 21941
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.303353963864559e-06,
+      "loss": 1.0273,
+      "step": 21942
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.302502185416099e-06,
+      "loss": 0.9697,
+      "step": 21943
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.301650468168108e-06,
+      "loss": 0.9053,
+      "step": 21944
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.300798812129727e-06,
+      "loss": 1.0703,
+      "step": 21945
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.299947217310115e-06,
+      "loss": 0.9736,
+      "step": 21946
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.299095683718411e-06,
+      "loss": 0.8838,
+      "step": 21947
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.298244211363761e-06,
+      "loss": 1.3525,
+      "step": 21948
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.297392800255314e-06,
+      "loss": 0.9229,
+      "step": 21949
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.296541450402222e-06,
+      "loss": 1.2158,
+      "step": 21950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.295690161813621e-06,
+      "loss": 1.0107,
+      "step": 21951
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.294838934498657e-06,
+      "loss": 0.9004,
+      "step": 21952
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.293987768466477e-06,
+      "loss": 0.9863,
+      "step": 21953
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.293136663726217e-06,
+      "loss": 0.9536,
+      "step": 21954
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.292285620287029e-06,
+      "loss": 0.8262,
+      "step": 21955
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.291434638158046e-06,
+      "loss": 1.123,
+      "step": 21956
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.290583717348415e-06,
+      "loss": 0.9375,
+      "step": 21957
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.289732857867275e-06,
+      "loss": 1.0225,
+      "step": 21958
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.288882059723759e-06,
+      "loss": 0.7607,
+      "step": 21959
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.288031322927012e-06,
+      "loss": 0.7803,
+      "step": 21960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.287180647486177e-06,
+      "loss": 0.6509,
+      "step": 21961
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.286330033410384e-06,
+      "loss": 1.0293,
+      "step": 21962
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.285479480708768e-06,
+      "loss": 0.96,
+      "step": 21963
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2846289893904725e-06,
+      "loss": 0.8291,
+      "step": 21964
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.283778559464632e-06,
+      "loss": 1.1953,
+      "step": 21965
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2829281909403795e-06,
+      "loss": 0.9395,
+      "step": 21966
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.282077883826846e-06,
+      "loss": 1.2012,
+      "step": 21967
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.281227638133173e-06,
+      "loss": 0.8994,
+      "step": 21968
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.280377453868486e-06,
+      "loss": 1.0439,
+      "step": 21969
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2795273310419235e-06,
+      "loss": 0.8281,
+      "step": 21970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2786772696626125e-06,
+      "loss": 1.0205,
+      "step": 21971
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.277827269739687e-06,
+      "loss": 1.2349,
+      "step": 21972
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.276977331282278e-06,
+      "loss": 0.5037,
+      "step": 21973
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.27612745429951e-06,
+      "loss": 0.9395,
+      "step": 21974
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2752776388005155e-06,
+      "loss": 0.835,
+      "step": 21975
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.274427884794426e-06,
+      "loss": 0.7305,
+      "step": 21976
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.273578192290368e-06,
+      "loss": 0.7344,
+      "step": 21977
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.272728561297461e-06,
+      "loss": 1.0186,
+      "step": 21978
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.271878991824843e-06,
+      "loss": 1.085,
+      "step": 21979
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.271029483881629e-06,
+      "loss": 0.7148,
+      "step": 21980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.270180037476953e-06,
+      "loss": 1.228,
+      "step": 21981
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.269330652619933e-06,
+      "loss": 0.9761,
+      "step": 21982
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.268481329319698e-06,
+      "loss": 0.7622,
+      "step": 21983
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.26763206758537e-06,
+      "loss": 0.7539,
+      "step": 21984
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2667828674260645e-06,
+      "loss": 1.0396,
+      "step": 21985
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.26593372885091e-06,
+      "loss": 1.2285,
+      "step": 21986
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.26508465186903e-06,
+      "loss": 1.3643,
+      "step": 21987
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.264235636489542e-06,
+      "loss": 1.0742,
+      "step": 21988
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2633866827215606e-06,
+      "loss": 0.8101,
+      "step": 21989
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.262537790574211e-06,
+      "loss": 0.8584,
+      "step": 21990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.261688960056612e-06,
+      "loss": 1.0205,
+      "step": 21991
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.260840191177882e-06,
+      "loss": 0.9043,
+      "step": 21992
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2599914839471315e-06,
+      "loss": 1.1514,
+      "step": 21993
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.259142838373484e-06,
+      "loss": 0.9351,
+      "step": 21994
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.25829425446605e-06,
+      "loss": 1.0332,
+      "step": 21995
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.257445732233951e-06,
+      "loss": 1.2129,
+      "step": 21996
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.256597271686295e-06,
+      "loss": 1.1338,
+      "step": 21997
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.255748872832201e-06,
+      "loss": 0.9648,
+      "step": 21998
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.254900535680781e-06,
+      "loss": 0.9609,
+      "step": 21999
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.25405226024114e-06,
+      "loss": 0.9795,
+      "step": 22000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.253204046522402e-06,
+      "loss": 1.1865,
+      "step": 22001
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.252355894533674e-06,
+      "loss": 0.9824,
+      "step": 22002
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.251507804284064e-06,
+      "loss": 1.5625,
+      "step": 22003
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.250659775782681e-06,
+      "loss": 0.957,
+      "step": 22004
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.249811809038635e-06,
+      "loss": 0.9756,
+      "step": 22005
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2489639040610396e-06,
+      "loss": 0.8281,
+      "step": 22006
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.248116060859e-06,
+      "loss": 0.8389,
+      "step": 22007
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247268279441617e-06,
+      "loss": 0.8926,
+      "step": 22008
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.246420559818007e-06,
+      "loss": 0.8809,
+      "step": 22009
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2455729019972715e-06,
+      "loss": 0.7104,
+      "step": 22010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.244725305988509e-06,
+      "loss": 1.0996,
+      "step": 22011
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2438777718008375e-06,
+      "loss": 0.8652,
+      "step": 22012
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.243030299443354e-06,
+      "loss": 1.5566,
+      "step": 22013
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.242182888925161e-06,
+      "loss": 0.8057,
+      "step": 22014
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.241335540255359e-06,
+      "loss": 1.0625,
+      "step": 22015
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.240488253443053e-06,
+      "loss": 1.6133,
+      "step": 22016
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2396410284973475e-06,
+      "loss": 0.9102,
+      "step": 22017
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.238793865427337e-06,
+      "loss": 1.1992,
+      "step": 22018
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.237946764242126e-06,
+      "loss": 0.8994,
+      "step": 22019
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.237099724950814e-06,
+      "loss": 1.1396,
+      "step": 22020
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.236252747562491e-06,
+      "loss": 0.9492,
+      "step": 22021
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2354058320862645e-06,
+      "loss": 0.7549,
+      "step": 22022
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.23455897853123e-06,
+      "loss": 0.8706,
+      "step": 22023
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.233712186906483e-06,
+      "loss": 0.7231,
+      "step": 22024
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.232865457221121e-06,
+      "loss": 1.2783,
+      "step": 22025
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.232018789484228e-06,
+      "loss": 1.1426,
+      "step": 22026
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.231172183704917e-06,
+      "loss": 0.8164,
+      "step": 22027
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2303256398922725e-06,
+      "loss": 0.6191,
+      "step": 22028
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229479158055383e-06,
+      "loss": 0.918,
+      "step": 22029
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.228632738203351e-06,
+      "loss": 0.7734,
+      "step": 22030
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.227786380345261e-06,
+      "loss": 0.8516,
+      "step": 22031
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.22694008449021e-06,
+      "loss": 0.9951,
+      "step": 22032
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.226093850647282e-06,
+      "loss": 0.9346,
+      "step": 22033
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2252476788255735e-06,
+      "loss": 0.8101,
+      "step": 22034
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2244015690341725e-06,
+      "loss": 0.8613,
+      "step": 22035
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.223555521282161e-06,
+      "loss": 0.4829,
+      "step": 22036
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.222709535578632e-06,
+      "loss": 0.7861,
+      "step": 22037
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.221863611932677e-06,
+      "loss": 1.0967,
+      "step": 22038
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.221017750353378e-06,
+      "loss": 0.9678,
+      "step": 22039
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.220171950849817e-06,
+      "loss": 0.9023,
+      "step": 22040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.219326213431088e-06,
+      "loss": 0.8833,
+      "step": 22041
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.218480538106267e-06,
+      "loss": 0.9814,
+      "step": 22042
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.217634924884445e-06,
+      "loss": 1.0273,
+      "step": 22043
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2167893737747e-06,
+      "loss": 1.1201,
+      "step": 22044
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.21594388478612e-06,
+      "loss": 0.8286,
+      "step": 22045
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2150984579277835e-06,
+      "loss": 0.9512,
+      "step": 22046
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.214253093208769e-06,
+      "loss": 0.8569,
+      "step": 22047
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2134077906381596e-06,
+      "loss": 0.8672,
+      "step": 22048
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.21256255022504e-06,
+      "loss": 0.6631,
+      "step": 22049
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.211717371978485e-06,
+      "loss": 1.1182,
+      "step": 22050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210872255907571e-06,
+      "loss": 1.1934,
+      "step": 22051
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210027202021377e-06,
+      "loss": 0.6401,
+      "step": 22052
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.209182210328986e-06,
+      "loss": 0.8887,
+      "step": 22053
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20833728083947e-06,
+      "loss": 0.9326,
+      "step": 22054
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2074924135619024e-06,
+      "loss": 1.2852,
+      "step": 22055
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.206647608505364e-06,
+      "loss": 0.8398,
+      "step": 22056
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.205802865678924e-06,
+      "loss": 0.7627,
+      "step": 22057
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.204958185091662e-06,
+      "loss": 0.9795,
+      "step": 22058
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.204113566752644e-06,
+      "loss": 1.0781,
+      "step": 22059
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.203269010670952e-06,
+      "loss": 0.833,
+      "step": 22060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.202424516855652e-06,
+      "loss": 0.8389,
+      "step": 22061
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.201580085315813e-06,
+      "loss": 0.9502,
+      "step": 22062
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.200735716060508e-06,
+      "loss": 0.9062,
+      "step": 22063
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1998914090988105e-06,
+      "loss": 1.002,
+      "step": 22064
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1990471644397875e-06,
+      "loss": 0.938,
+      "step": 22065
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.198202982092504e-06,
+      "loss": 0.9072,
+      "step": 22066
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.19735886206603e-06,
+      "loss": 0.8813,
+      "step": 22067
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.196514804369437e-06,
+      "loss": 0.9414,
+      "step": 22068
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.195670809011787e-06,
+      "loss": 1.0605,
+      "step": 22069
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.194826876002145e-06,
+      "loss": 1.2344,
+      "step": 22070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.193983005349581e-06,
+      "loss": 1.1709,
+      "step": 22071
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.193139197063156e-06,
+      "loss": 1.4453,
+      "step": 22072
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.192295451151932e-06,
+      "loss": 1.1406,
+      "step": 22073
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1914517676249744e-06,
+      "loss": 0.9819,
+      "step": 22074
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.190608146491351e-06,
+      "loss": 0.7129,
+      "step": 22075
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.189764587760117e-06,
+      "loss": 0.9004,
+      "step": 22076
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.188921091440332e-06,
+      "loss": 1.0303,
+      "step": 22077
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.188077657541061e-06,
+      "loss": 1.0347,
+      "step": 22078
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1872342860713655e-06,
+      "loss": 0.9775,
+      "step": 22079
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1863909770403036e-06,
+      "loss": 0.8687,
+      "step": 22080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.185547730456927e-06,
+      "loss": 0.7461,
+      "step": 22081
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.184704546330302e-06,
+      "loss": 0.7114,
+      "step": 22082
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.18386142466948e-06,
+      "loss": 0.8433,
+      "step": 22083
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.183018365483523e-06,
+      "loss": 0.8843,
+      "step": 22084
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.182175368781481e-06,
+      "loss": 1.208,
+      "step": 22085
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.181332434572416e-06,
+      "loss": 1.0391,
+      "step": 22086
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.180489562865377e-06,
+      "loss": 0.7393,
+      "step": 22087
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.179646753669417e-06,
+      "loss": 0.6416,
+      "step": 22088
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.178804006993592e-06,
+      "loss": 0.7612,
+      "step": 22089
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1779613228469564e-06,
+      "loss": 0.916,
+      "step": 22090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.17711870123856e-06,
+      "loss": 0.9878,
+      "step": 22091
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.17627614217745e-06,
+      "loss": 0.8457,
+      "step": 22092
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.17543364567268e-06,
+      "loss": 0.7891,
+      "step": 22093
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1745912117333054e-06,
+      "loss": 0.8203,
+      "step": 22094
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.173748840368369e-06,
+      "loss": 0.9814,
+      "step": 22095
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.172906531586916e-06,
+      "loss": 1.1416,
+      "step": 22096
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.172064285398004e-06,
+      "loss": 0.9062,
+      "step": 22097
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.17122210181067e-06,
+      "loss": 0.7236,
+      "step": 22098
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1703799808339655e-06,
+      "loss": 0.9795,
+      "step": 22099
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.169537922476939e-06,
+      "loss": 1.083,
+      "step": 22100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1686959267486325e-06,
+      "loss": 1.0869,
+      "step": 22101
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.167853993658091e-06,
+      "loss": 1.3398,
+      "step": 22102
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.167012123214354e-06,
+      "loss": 1.0928,
+      "step": 22103
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.166170315426467e-06,
+      "loss": 0.9961,
+      "step": 22104
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.165328570303477e-06,
+      "loss": 1.1484,
+      "step": 22105
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.164486887854424e-06,
+      "loss": 0.5068,
+      "step": 22106
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1636452680883436e-06,
+      "loss": 0.6865,
+      "step": 22107
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1628037110142825e-06,
+      "loss": 0.999,
+      "step": 22108
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1619622166412745e-06,
+      "loss": 1.0376,
+      "step": 22109
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.161120784978362e-06,
+      "loss": 1.2793,
+      "step": 22110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160279416034587e-06,
+      "loss": 0.7656,
+      "step": 22111
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.159438109818983e-06,
+      "loss": 0.832,
+      "step": 22112
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.158596866340588e-06,
+      "loss": 0.8242,
+      "step": 22113
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.157755685608431e-06,
+      "loss": 1.127,
+      "step": 22114
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.156914567631561e-06,
+      "loss": 0.8442,
+      "step": 22115
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.156073512419006e-06,
+      "loss": 1.248,
+      "step": 22116
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.155232519979799e-06,
+      "loss": 1.165,
+      "step": 22117
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1543915903229785e-06,
+      "loss": 1.2354,
+      "step": 22118
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.15355072345757e-06,
+      "loss": 0.8835,
+      "step": 22119
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1527099193926135e-06,
+      "loss": 1.4814,
+      "step": 22120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1518691781371344e-06,
+      "loss": 0.9473,
+      "step": 22121
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.15102849970017e-06,
+      "loss": 1.0762,
+      "step": 22122
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.150187884090747e-06,
+      "loss": 1.1357,
+      "step": 22123
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.149347331317892e-06,
+      "loss": 0.9062,
+      "step": 22124
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.148506841390637e-06,
+      "loss": 0.8359,
+      "step": 22125
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.147666414318013e-06,
+      "loss": 0.8535,
+      "step": 22126
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1468260501090454e-06,
+      "loss": 0.9238,
+      "step": 22127
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.145985748772756e-06,
+      "loss": 1.1484,
+      "step": 22128
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.145145510318181e-06,
+      "loss": 1.4199,
+      "step": 22129
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.144305334754335e-06,
+      "loss": 0.9121,
+      "step": 22130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.143465222090253e-06,
+      "loss": 1.1748,
+      "step": 22131
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.14262517233495e-06,
+      "loss": 0.915,
+      "step": 22132
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.141785185497459e-06,
+      "loss": 0.9287,
+      "step": 22133
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.140945261586796e-06,
+      "loss": 1.0957,
+      "step": 22134
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.140105400611983e-06,
+      "loss": 0.8008,
+      "step": 22135
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.139265602582042e-06,
+      "loss": 0.8096,
+      "step": 22136
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1384258675060005e-06,
+      "loss": 1.0703,
+      "step": 22137
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.137586195392872e-06,
+      "loss": 1.0303,
+      "step": 22138
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1367465862516754e-06,
+      "loss": 0.8867,
+      "step": 22139
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.13590704009143e-06,
+      "loss": 0.8174,
+      "step": 22140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.135067556921159e-06,
+      "loss": 0.9429,
+      "step": 22141
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.134228136749877e-06,
+      "loss": 0.9277,
+      "step": 22142
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.133388779586596e-06,
+      "loss": 0.9834,
+      "step": 22143
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.13254948544034e-06,
+      "loss": 0.792,
+      "step": 22144
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.131710254320116e-06,
+      "loss": 1.0068,
+      "step": 22145
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1308710862349475e-06,
+      "loss": 1.1089,
+      "step": 22146
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.13003198119384e-06,
+      "loss": 0.7163,
+      "step": 22147
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.129192939205815e-06,
+      "loss": 1.0518,
+      "step": 22148
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.128353960279881e-06,
+      "loss": 0.9565,
+      "step": 22149
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.127515044425047e-06,
+      "loss": 0.7568,
+      "step": 22150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.126676191650326e-06,
+      "loss": 1.0059,
+      "step": 22151
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.125837401964735e-06,
+      "loss": 0.9097,
+      "step": 22152
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.124998675377279e-06,
+      "loss": 0.8501,
+      "step": 22153
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.124160011896965e-06,
+      "loss": 0.9473,
+      "step": 22154
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.123321411532804e-06,
+      "loss": 1.1699,
+      "step": 22155
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.122482874293806e-06,
+      "loss": 0.7969,
+      "step": 22156
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.121644400188976e-06,
+      "loss": 0.9902,
+      "step": 22157
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.120805989227319e-06,
+      "loss": 1.0273,
+      "step": 22158
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.119967641417846e-06,
+      "loss": 1.0381,
+      "step": 22159
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.119129356769559e-06,
+      "loss": 1.1611,
+      "step": 22160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.118291135291459e-06,
+      "loss": 1.1553,
+      "step": 22161
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.117452976992553e-06,
+      "loss": 1.2676,
+      "step": 22162
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.116614881881848e-06,
+      "loss": 0.7793,
+      "step": 22163
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.115776849968345e-06,
+      "loss": 1.0103,
+      "step": 22164
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.114938881261038e-06,
+      "loss": 0.9482,
+      "step": 22165
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.114100975768936e-06,
+      "loss": 1.6328,
+      "step": 22166
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1132631335010395e-06,
+      "loss": 0.8372,
+      "step": 22167
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1124253544663485e-06,
+      "loss": 1.3203,
+      "step": 22168
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.111587638673855e-06,
+      "loss": 1.0537,
+      "step": 22169
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.110749986132566e-06,
+      "loss": 1.1309,
+      "step": 22170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.109912396851473e-06,
+      "loss": 0.9355,
+      "step": 22171
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.109074870839579e-06,
+      "loss": 0.9939,
+      "step": 22172
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.108237408105874e-06,
+      "loss": 1.0098,
+      "step": 22173
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.10740000865936e-06,
+      "loss": 0.9824,
+      "step": 22174
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1065626725090294e-06,
+      "loss": 0.9062,
+      "step": 22175
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.105725399663872e-06,
+      "loss": 1.4336,
+      "step": 22176
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.104888190132886e-06,
+      "loss": 1.3076,
+      "step": 22177
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.104051043925068e-06,
+      "loss": 0.8809,
+      "step": 22178
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.103213961049406e-06,
+      "loss": 0.9512,
+      "step": 22179
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1023769415148885e-06,
+      "loss": 1.1299,
+      "step": 22180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.101539985330511e-06,
+      "loss": 0.6807,
+      "step": 22181
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.100703092505265e-06,
+      "loss": 0.8027,
+      "step": 22182
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.099866263048138e-06,
+      "loss": 0.9639,
+      "step": 22183
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.099029496968116e-06,
+      "loss": 1.0879,
+      "step": 22184
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.098192794274193e-06,
+      "loss": 1.0845,
+      "step": 22185
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.097356154975349e-06,
+      "loss": 0.8486,
+      "step": 22186
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.096519579080581e-06,
+      "loss": 1.3857,
+      "step": 22187
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.095683066598865e-06,
+      "loss": 0.8921,
+      "step": 22188
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.094846617539195e-06,
+      "loss": 0.9648,
+      "step": 22189
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0940102319105524e-06,
+      "loss": 0.9033,
+      "step": 22190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.093173909721917e-06,
+      "loss": 0.6997,
+      "step": 22191
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.092337650982275e-06,
+      "loss": 0.5767,
+      "step": 22192
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.091501455700616e-06,
+      "loss": 0.7085,
+      "step": 22193
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.090665323885915e-06,
+      "loss": 1.293,
+      "step": 22194
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.089829255547152e-06,
+      "loss": 1.0518,
+      "step": 22195
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.088993250693315e-06,
+      "loss": 0.8345,
+      "step": 22196
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0881573093333745e-06,
+      "loss": 1.1445,
+      "step": 22197
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.087321431476321e-06,
+      "loss": 1.0547,
+      "step": 22198
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.086485617131122e-06,
+      "loss": 1.0884,
+      "step": 22199
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0856498663067646e-06,
+      "loss": 1.0,
+      "step": 22200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.084814179012223e-06,
+      "loss": 1.0391,
+      "step": 22201
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.083978555256466e-06,
+      "loss": 1.0732,
+      "step": 22202
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0831429950484835e-06,
+      "loss": 0.9648,
+      "step": 22203
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.082307498397245e-06,
+      "loss": 0.9629,
+      "step": 22204
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.081472065311724e-06,
+      "loss": 0.7368,
+      "step": 22205
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.080636695800889e-06,
+      "loss": 0.9795,
+      "step": 22206
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0798013898737206e-06,
+      "loss": 1.1279,
+      "step": 22207
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.078966147539194e-06,
+      "loss": 0.8955,
+      "step": 22208
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.078130968806272e-06,
+      "loss": 0.5815,
+      "step": 22209
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.077295853683933e-06,
+      "loss": 1.1504,
+      "step": 22210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.076460802181147e-06,
+      "loss": 1.1416,
+      "step": 22211
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.075625814306877e-06,
+      "loss": 0.7905,
+      "step": 22212
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.074790890070098e-06,
+      "loss": 0.9883,
+      "step": 22213
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.07395602947978e-06,
+      "loss": 0.939,
+      "step": 22214
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.073121232544889e-06,
+      "loss": 1.3677,
+      "step": 22215
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.072286499274387e-06,
+      "loss": 0.886,
+      "step": 22216
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.071451829677244e-06,
+      "loss": 1.3535,
+      "step": 22217
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.070617223762431e-06,
+      "loss": 0.8872,
+      "step": 22218
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.069782681538908e-06,
+      "loss": 1.0972,
+      "step": 22219
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.068948203015637e-06,
+      "loss": 0.8125,
+      "step": 22220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.068113788201588e-06,
+      "loss": 1.0088,
+      "step": 22221
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.067279437105719e-06,
+      "loss": 0.9668,
+      "step": 22222
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.066445149736991e-06,
+      "loss": 1.2627,
+      "step": 22223
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0656109261043686e-06,
+      "loss": 1.1895,
+      "step": 22224
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.064776766216816e-06,
+      "loss": 0.9375,
+      "step": 22225
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.063942670083288e-06,
+      "loss": 1.2285,
+      "step": 22226
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.063108637712745e-06,
+      "loss": 0.9326,
+      "step": 22227
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.062274669114146e-06,
+      "loss": 0.875,
+      "step": 22228
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.061440764296454e-06,
+      "loss": 1.168,
+      "step": 22229
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.060606923268622e-06,
+      "loss": 1.0674,
+      "step": 22230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.059773146039604e-06,
+      "loss": 0.8696,
+      "step": 22231
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.058939432618364e-06,
+      "loss": 0.8921,
+      "step": 22232
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.058105783013848e-06,
+      "loss": 0.8291,
+      "step": 22233
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.05727219723502e-06,
+      "loss": 1.0312,
+      "step": 22234
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.056438675290827e-06,
+      "loss": 1.3193,
+      "step": 22235
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.055605217190228e-06,
+      "loss": 0.8618,
+      "step": 22236
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0547718229421736e-06,
+      "loss": 0.9004,
+      "step": 22237
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.053938492555611e-06,
+      "loss": 0.6846,
+      "step": 22238
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.053105226039497e-06,
+      "loss": 0.6362,
+      "step": 22239
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.052272023402782e-06,
+      "loss": 1.2539,
+      "step": 22240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0514388846544175e-06,
+      "loss": 0.7974,
+      "step": 22241
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0506058098033456e-06,
+      "loss": 1.209,
+      "step": 22242
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.049772798858519e-06,
+      "loss": 0.9277,
+      "step": 22243
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0489398518288905e-06,
+      "loss": 0.8481,
+      "step": 22244
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0481069687234035e-06,
+      "loss": 1.1143,
+      "step": 22245
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.047274149550998e-06,
+      "loss": 0.7061,
+      "step": 22246
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0464413943206316e-06,
+      "loss": 0.9971,
+      "step": 22247
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045608703041239e-06,
+      "loss": 1.1172,
+      "step": 22248
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.044776075721773e-06,
+      "loss": 0.7705,
+      "step": 22249
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.043943512371171e-06,
+      "loss": 1.0723,
+      "step": 22250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.043111012998382e-06,
+      "loss": 1.3311,
+      "step": 22251
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.042278577612344e-06,
+      "loss": 0.7773,
+      "step": 22252
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.041446206221997e-06,
+      "loss": 0.4712,
+      "step": 22253
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.040613898836287e-06,
+      "loss": 0.873,
+      "step": 22254
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.039781655464154e-06,
+      "loss": 1.2891,
+      "step": 22255
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.038949476114536e-06,
+      "loss": 0.8647,
+      "step": 22256
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0381173607963675e-06,
+      "loss": 1.0752,
+      "step": 22257
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.037285309518596e-06,
+      "loss": 0.8369,
+      "step": 22258
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.03645332229015e-06,
+      "loss": 1.1367,
+      "step": 22259
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.035621399119976e-06,
+      "loss": 0.752,
+      "step": 22260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.034789540016999e-06,
+      "loss": 1.1602,
+      "step": 22261
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.033957744990165e-06,
+      "loss": 1.0527,
+      "step": 22262
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0331260140484055e-06,
+      "loss": 1.2852,
+      "step": 22263
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.032294347200648e-06,
+      "loss": 0.9453,
+      "step": 22264
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0314627444558314e-06,
+      "loss": 0.7153,
+      "step": 22265
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.030631205822894e-06,
+      "loss": 0.9338,
+      "step": 22266
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.02979973131076e-06,
+      "loss": 0.9521,
+      "step": 22267
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.028968320928363e-06,
+      "loss": 1.0166,
+      "step": 22268
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0281369746846314e-06,
+      "loss": 1.1592,
+      "step": 22269
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.027305692588501e-06,
+      "loss": 1.0107,
+      "step": 22270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.026474474648899e-06,
+      "loss": 1.3428,
+      "step": 22271
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.025643320874749e-06,
+      "loss": 1.2256,
+      "step": 22272
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.024812231274986e-06,
+      "loss": 0.915,
+      "step": 22273
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.02398120585853e-06,
+      "loss": 0.9399,
+      "step": 22274
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.023150244634317e-06,
+      "loss": 0.8052,
+      "step": 22275
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0223193476112634e-06,
+      "loss": 0.96,
+      "step": 22276
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0214885147983015e-06,
+      "loss": 1.1504,
+      "step": 22277
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.020657746204353e-06,
+      "loss": 1.0264,
+      "step": 22278
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.019827041838339e-06,
+      "loss": 0.7505,
+      "step": 22279
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.018996401709184e-06,
+      "loss": 0.9727,
+      "step": 22280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.018165825825816e-06,
+      "loss": 0.7402,
+      "step": 22281
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.017335314197152e-06,
+      "loss": 1.4023,
+      "step": 22282
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.016504866832109e-06,
+      "loss": 1.2002,
+      "step": 22283
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0156744837396175e-06,
+      "loss": 0.8682,
+      "step": 22284
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.014844164928586e-06,
+      "loss": 0.9443,
+      "step": 22285
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0140139104079445e-06,
+      "loss": 1.1289,
+      "step": 22286
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.013183720186601e-06,
+      "loss": 1.2383,
+      "step": 22287
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0123535942734814e-06,
+      "loss": 0.9038,
+      "step": 22288
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.011523532677499e-06,
+      "loss": 0.9043,
+      "step": 22289
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.010693535407564e-06,
+      "loss": 1.1787,
+      "step": 22290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.009863602472603e-06,
+      "loss": 1.3242,
+      "step": 22291
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.009033733881528e-06,
+      "loss": 0.9717,
+      "step": 22292
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0082039296432506e-06,
+      "loss": 1.4219,
+      "step": 22293
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.007374189766681e-06,
+      "loss": 1.2158,
+      "step": 22294
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.006544514260735e-06,
+      "loss": 1.0498,
+      "step": 22295
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.00571490313433e-06,
+      "loss": 1.042,
+      "step": 22296
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.004885356396372e-06,
+      "loss": 0.8037,
+      "step": 22297
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.004055874055769e-06,
+      "loss": 0.7969,
+      "step": 22298
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.003226456121438e-06,
+      "loss": 0.9829,
+      "step": 22299
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.002397102602281e-06,
+      "loss": 1.252,
+      "step": 22300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.00156781350721e-06,
+      "loss": 0.7207,
+      "step": 22301
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.000738588845138e-06,
+      "loss": 0.5742,
+      "step": 22302
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.999909428624966e-06,
+      "loss": 0.9727,
+      "step": 22303
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.999080332855604e-06,
+      "loss": 0.8638,
+      "step": 22304
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.998251301545946e-06,
+      "loss": 0.9019,
+      "step": 22305
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.997422334704917e-06,
+      "loss": 0.9404,
+      "step": 22306
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.99659343234141e-06,
+      "loss": 1.1914,
+      "step": 22307
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.995764594464327e-06,
+      "loss": 0.8379,
+      "step": 22308
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.994935821082578e-06,
+      "loss": 1.3906,
+      "step": 22309
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.994107112205058e-06,
+      "loss": 1.0352,
+      "step": 22310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.993278467840675e-06,
+      "loss": 0.8931,
+      "step": 22311
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.992449887998326e-06,
+      "loss": 0.9736,
+      "step": 22312
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.991621372686915e-06,
+      "loss": 1.2988,
+      "step": 22313
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.990792921915339e-06,
+      "loss": 0.6475,
+      "step": 22314
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.989964535692494e-06,
+      "loss": 0.7549,
+      "step": 22315
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.989136214027283e-06,
+      "loss": 1.0818,
+      "step": 22316
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.988307956928604e-06,
+      "loss": 0.958,
+      "step": 22317
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.987479764405352e-06,
+      "loss": 1.3125,
+      "step": 22318
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98665163646642e-06,
+      "loss": 1.1201,
+      "step": 22319
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98582357312071e-06,
+      "loss": 0.9487,
+      "step": 22320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984995574377109e-06,
+      "loss": 1.0059,
+      "step": 22321
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984167640244518e-06,
+      "loss": 0.8145,
+      "step": 22322
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.983339770731825e-06,
+      "loss": 0.8887,
+      "step": 22323
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98251196584793e-06,
+      "loss": 1.2129,
+      "step": 22324
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98168422560172e-06,
+      "loss": 0.8447,
+      "step": 22325
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980856550002082e-06,
+      "loss": 1.002,
+      "step": 22326
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980028939057912e-06,
+      "loss": 0.7056,
+      "step": 22327
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.979201392778101e-06,
+      "loss": 1.1191,
+      "step": 22328
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.978373911171538e-06,
+      "loss": 0.7988,
+      "step": 22329
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.977546494247106e-06,
+      "loss": 0.7759,
+      "step": 22330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.976719142013696e-06,
+      "loss": 0.8828,
+      "step": 22331
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9758918544802e-06,
+      "loss": 0.915,
+      "step": 22332
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9750646316555e-06,
+      "loss": 1.3818,
+      "step": 22333
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.974237473548478e-06,
+      "loss": 0.7217,
+      "step": 22334
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.973410380168026e-06,
+      "loss": 1.4258,
+      "step": 22335
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.972583351523023e-06,
+      "loss": 1.0391,
+      "step": 22336
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9717563876223576e-06,
+      "loss": 0.8535,
+      "step": 22337
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970929488474906e-06,
+      "loss": 1.04,
+      "step": 22338
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970102654089559e-06,
+      "loss": 0.6611,
+      "step": 22339
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9692758844751936e-06,
+      "loss": 1.3672,
+      "step": 22340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.968449179640686e-06,
+      "loss": 1.2666,
+      "step": 22341
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.967622539594923e-06,
+      "loss": 0.625,
+      "step": 22342
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.966795964346785e-06,
+      "loss": 0.7363,
+      "step": 22343
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.965969453905146e-06,
+      "loss": 0.5872,
+      "step": 22344
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9651430082788844e-06,
+      "loss": 0.875,
+      "step": 22345
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.964316627476883e-06,
+      "loss": 0.8418,
+      "step": 22346
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.96349031150801e-06,
+      "loss": 0.9688,
+      "step": 22347
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.96266406038115e-06,
+      "loss": 0.957,
+      "step": 22348
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.961837874105171e-06,
+      "loss": 1.1074,
+      "step": 22349
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.961011752688955e-06,
+      "loss": 0.7852,
+      "step": 22350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.960185696141371e-06,
+      "loss": 0.9834,
+      "step": 22351
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.959359704471289e-06,
+      "loss": 0.9771,
+      "step": 22352
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.958533777687586e-06,
+      "loss": 1.0293,
+      "step": 22353
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.957707915799137e-06,
+      "loss": 0.8535,
+      "step": 22354
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.95688211881481e-06,
+      "loss": 1.082,
+      "step": 22355
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9560563867434706e-06,
+      "loss": 1.1323,
+      "step": 22356
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.955230719593993e-06,
+      "loss": 0.7705,
+      "step": 22357
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9544051173752504e-06,
+      "loss": 0.9062,
+      "step": 22358
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.953579580096105e-06,
+      "loss": 1.0996,
+      "step": 22359
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.952754107765424e-06,
+      "loss": 0.9724,
+      "step": 22360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.95192870039208e-06,
+      "loss": 0.9839,
+      "step": 22361
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.951103357984931e-06,
+      "loss": 0.8145,
+      "step": 22362
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.950278080552852e-06,
+      "loss": 1.0645,
+      "step": 22363
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9494528681047e-06,
+      "loss": 0.751,
+      "step": 22364
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.948627720649344e-06,
+      "loss": 1.0986,
+      "step": 22365
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.947802638195647e-06,
+      "loss": 1.0659,
+      "step": 22366
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946977620752465e-06,
+      "loss": 1.1104,
+      "step": 22367
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946152668328668e-06,
+      "loss": 1.021,
+      "step": 22368
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.945327780933116e-06,
+      "loss": 0.7422,
+      "step": 22369
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9445029585746695e-06,
+      "loss": 0.8584,
+      "step": 22370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.943678201262185e-06,
+      "loss": 0.6465,
+      "step": 22371
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.942853509004522e-06,
+      "loss": 0.8677,
+      "step": 22372
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.942028881810546e-06,
+      "loss": 1.1221,
+      "step": 22373
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9412043196891085e-06,
+      "loss": 1.1338,
+      "step": 22374
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9403798226490655e-06,
+      "loss": 0.9819,
+      "step": 22375
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9395553906992804e-06,
+      "loss": 1.2617,
+      "step": 22376
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.938731023848603e-06,
+      "loss": 1.3105,
+      "step": 22377
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.937906722105887e-06,
+      "loss": 0.8025,
+      "step": 22378
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9370824854799906e-06,
+      "loss": 0.8076,
+      "step": 22379
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9362583139797696e-06,
+      "loss": 1.0625,
+      "step": 22380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.935434207614074e-06,
+      "loss": 0.9551,
+      "step": 22381
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.934610166391752e-06,
+      "loss": 1.4121,
+      "step": 22382
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.93378619032166e-06,
+      "loss": 1.0068,
+      "step": 22383
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.932962279412651e-06,
+      "loss": 1.1895,
+      "step": 22384
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9321384336735724e-06,
+      "loss": 0.8789,
+      "step": 22385
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.93131465311327e-06,
+      "loss": 0.8545,
+      "step": 22386
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9304909377405985e-06,
+      "loss": 0.979,
+      "step": 22387
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.929667287564401e-06,
+      "loss": 0.9102,
+      "step": 22388
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.928843702593527e-06,
+      "loss": 0.917,
+      "step": 22389
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.928020182836827e-06,
+      "loss": 0.8057,
+      "step": 22390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.927196728303143e-06,
+      "loss": 0.834,
+      "step": 22391
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.926373339001321e-06,
+      "loss": 0.7271,
+      "step": 22392
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.925550014940198e-06,
+      "loss": 1.1494,
+      "step": 22393
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.924726756128632e-06,
+      "loss": 1.0713,
+      "step": 22394
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.92390356257546e-06,
+      "loss": 1.0537,
+      "step": 22395
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923080434289522e-06,
+      "loss": 0.7168,
+      "step": 22396
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.922257371279658e-06,
+      "loss": 0.9907,
+      "step": 22397
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.921434373554714e-06,
+      "loss": 1.252,
+      "step": 22398
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.920611441123531e-06,
+      "loss": 0.9893,
+      "step": 22399
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.919788573994944e-06,
+      "loss": 0.8567,
+      "step": 22400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9189657721777965e-06,
+      "loss": 0.9316,
+      "step": 22401
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.918143035680925e-06,
+      "loss": 1.2607,
+      "step": 22402
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.917320364513163e-06,
+      "loss": 0.8389,
+      "step": 22403
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.916497758683352e-06,
+      "loss": 1.2158,
+      "step": 22404
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.91567521820033e-06,
+      "loss": 0.8359,
+      "step": 22405
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9148527430729285e-06,
+      "loss": 1.3789,
+      "step": 22406
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.914030333309982e-06,
+      "loss": 1.1309,
+      "step": 22407
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.913207988920328e-06,
+      "loss": 1.2969,
+      "step": 22408
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.912385709912794e-06,
+      "loss": 1.0078,
+      "step": 22409
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.911563496296219e-06,
+      "loss": 1.2148,
+      "step": 22410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.91074134807943e-06,
+      "loss": 1.1533,
+      "step": 22411
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9099192652712635e-06,
+      "loss": 0.9434,
+      "step": 22412
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.909097247880547e-06,
+      "loss": 1.1538,
+      "step": 22413
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.908275295916106e-06,
+      "loss": 1.0342,
+      "step": 22414
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.907453409386774e-06,
+      "loss": 1.124,
+      "step": 22415
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.906631588301384e-06,
+      "loss": 1.0996,
+      "step": 22416
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.905809832668758e-06,
+      "loss": 0.874,
+      "step": 22417
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.90498814249772e-06,
+      "loss": 1.1309,
+      "step": 22418
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9041665177971e-06,
+      "loss": 0.9473,
+      "step": 22419
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9033449585757285e-06,
+      "loss": 0.7482,
+      "step": 22420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.902523464842425e-06,
+      "loss": 0.6606,
+      "step": 22421
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.901702036606011e-06,
+      "loss": 0.8418,
+      "step": 22422
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.900880673875318e-06,
+      "loss": 0.708,
+      "step": 22423
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9000593766591595e-06,
+      "loss": 1.2021,
+      "step": 22424
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8992381449663665e-06,
+      "loss": 0.9917,
+      "step": 22425
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.898416978805751e-06,
+      "loss": 0.9014,
+      "step": 22426
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.897595878186144e-06,
+      "loss": 0.624,
+      "step": 22427
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8967748431163595e-06,
+      "loss": 0.7653,
+      "step": 22428
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.895953873605215e-06,
+      "loss": 1.0557,
+      "step": 22429
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.89513296966153e-06,
+      "loss": 1.1973,
+      "step": 22430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.894312131294129e-06,
+      "loss": 0.8311,
+      "step": 22431
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.893491358511824e-06,
+      "loss": 0.7632,
+      "step": 22432
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.892670651323427e-06,
+      "loss": 1.0029,
+      "step": 22433
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.891850009737759e-06,
+      "loss": 0.8652,
+      "step": 22434
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8910294337636375e-06,
+      "loss": 1.3359,
+      "step": 22435
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.890208923409874e-06,
+      "loss": 0.7373,
+      "step": 22436
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.889388478685278e-06,
+      "loss": 1.0488,
+      "step": 22437
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.888568099598669e-06,
+      "loss": 0.7021,
+      "step": 22438
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.887747786158857e-06,
+      "loss": 0.8066,
+      "step": 22439
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886927538374651e-06,
+      "loss": 1.0508,
+      "step": 22440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886107356254861e-06,
+      "loss": 0.7891,
+      "step": 22441
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.885287239808306e-06,
+      "loss": 0.9053,
+      "step": 22442
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.884467189043787e-06,
+      "loss": 1.1885,
+      "step": 22443
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8836472039701125e-06,
+      "loss": 1.0605,
+      "step": 22444
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8828272845960924e-06,
+      "loss": 1.0498,
+      "step": 22445
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.882007430930539e-06,
+      "loss": 1.2725,
+      "step": 22446
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.881187642982253e-06,
+      "loss": 1.1455,
+      "step": 22447
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.880367920760038e-06,
+      "loss": 0.9961,
+      "step": 22448
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.879548264272708e-06,
+      "loss": 1.0605,
+      "step": 22449
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.878728673529058e-06,
+      "loss": 1.0342,
+      "step": 22450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.877909148537899e-06,
+      "loss": 1.0928,
+      "step": 22451
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.877089689308028e-06,
+      "loss": 0.8081,
+      "step": 22452
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8762702958482535e-06,
+      "loss": 0.9932,
+      "step": 22453
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.875450968167375e-06,
+      "loss": 0.7695,
+      "step": 22454
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.874631706274188e-06,
+      "loss": 0.835,
+      "step": 22455
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.873812510177497e-06,
+      "loss": 0.7446,
+      "step": 22456
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8729933798861056e-06,
+      "loss": 1.0713,
+      "step": 22457
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.872174315408808e-06,
+      "loss": 0.8389,
+      "step": 22458
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8713553167544e-06,
+      "loss": 0.6123,
+      "step": 22459
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.870536383931681e-06,
+      "loss": 0.8438,
+      "step": 22460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.869717516949453e-06,
+      "loss": 1.1421,
+      "step": 22461
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8688987158165084e-06,
+      "loss": 0.5972,
+      "step": 22462
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.868079980541636e-06,
+      "loss": 0.8496,
+      "step": 22463
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.867261311133641e-06,
+      "loss": 1.4668,
+      "step": 22464
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.866442707601309e-06,
+      "loss": 1.2305,
+      "step": 22465
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8656241699534396e-06,
+      "loss": 1.0781,
+      "step": 22466
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.864805698198817e-06,
+      "loss": 0.9385,
+      "step": 22467
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.863987292346243e-06,
+      "loss": 0.7378,
+      "step": 22468
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.863168952404504e-06,
+      "loss": 1.0645,
+      "step": 22469
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.862350678382385e-06,
+      "loss": 1.042,
+      "step": 22470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.861532470288681e-06,
+      "loss": 1.0312,
+      "step": 22471
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.860714328132184e-06,
+      "loss": 0.8784,
+      "step": 22472
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.859896251921679e-06,
+      "loss": 1.2119,
+      "step": 22473
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8590782416659485e-06,
+      "loss": 1.4961,
+      "step": 22474
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.858260297373788e-06,
+      "loss": 1.208,
+      "step": 22475
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.857442419053976e-06,
+      "loss": 1.6299,
+      "step": 22476
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.856624606715304e-06,
+      "loss": 0.855,
+      "step": 22477
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.855806860366552e-06,
+      "loss": 1.0806,
+      "step": 22478
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.854989180016508e-06,
+      "loss": 1.2148,
+      "step": 22479
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.854171565673954e-06,
+      "loss": 1.0449,
+      "step": 22480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853354017347664e-06,
+      "loss": 1.3496,
+      "step": 22481
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.852536535046435e-06,
+      "loss": 0.979,
+      "step": 22482
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851719118779039e-06,
+      "loss": 0.897,
+      "step": 22483
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.85090176855426e-06,
+      "loss": 0.8525,
+      "step": 22484
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.850084484380871e-06,
+      "loss": 0.6279,
+      "step": 22485
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.849267266267654e-06,
+      "loss": 0.9971,
+      "step": 22486
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8484501142233945e-06,
+      "loss": 1.2754,
+      "step": 22487
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8476330282568595e-06,
+      "loss": 1.1211,
+      "step": 22488
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.846816008376835e-06,
+      "loss": 1.1094,
+      "step": 22489
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.845999054592092e-06,
+      "loss": 0.9932,
+      "step": 22490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.845182166911404e-06,
+      "loss": 0.855,
+      "step": 22491
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.844365345343548e-06,
+      "loss": 0.8906,
+      "step": 22492
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.843548589897302e-06,
+      "loss": 1.1162,
+      "step": 22493
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.842731900581435e-06,
+      "loss": 0.8862,
+      "step": 22494
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8419152774047215e-06,
+      "loss": 1.0938,
+      "step": 22495
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8410987203759284e-06,
+      "loss": 0.7842,
+      "step": 22496
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.840282229503831e-06,
+      "loss": 1.1475,
+      "step": 22497
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.839465804797203e-06,
+      "loss": 1.126,
+      "step": 22498
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.838649446264807e-06,
+      "loss": 1.1221,
+      "step": 22499
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8378331539154175e-06,
+      "loss": 1.2266,
+      "step": 22500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.837016927757803e-06,
+      "loss": 0.8657,
+      "step": 22501
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.836200767800725e-06,
+      "loss": 1.0508,
+      "step": 22502
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.835384674052953e-06,
+      "loss": 1.1533,
+      "step": 22503
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.83456864652326e-06,
+      "loss": 1.0176,
+      "step": 22504
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.833752685220405e-06,
+      "loss": 0.8926,
+      "step": 22505
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.83293679015315e-06,
+      "loss": 0.9609,
+      "step": 22506
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8321209613302635e-06,
+      "loss": 0.8677,
+      "step": 22507
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.831305198760512e-06,
+      "loss": 1.0547,
+      "step": 22508
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.830489502452654e-06,
+      "loss": 0.8535,
+      "step": 22509
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.829673872415448e-06,
+      "loss": 0.917,
+      "step": 22510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8288583086576625e-06,
+      "loss": 0.7314,
+      "step": 22511
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.828042811188051e-06,
+      "loss": 0.7354,
+      "step": 22512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.827227380015381e-06,
+      "loss": 0.998,
+      "step": 22513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.826412015148403e-06,
+      "loss": 0.7661,
+      "step": 22514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.825596716595883e-06,
+      "loss": 0.7913,
+      "step": 22515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.824781484366575e-06,
+      "loss": 1.0264,
+      "step": 22516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8239663184692335e-06,
+      "loss": 1.083,
+      "step": 22517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.823151218912618e-06,
+      "loss": 1.1309,
+      "step": 22518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.822336185705485e-06,
+      "loss": 0.6855,
+      "step": 22519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8215212188565894e-06,
+      "loss": 1.0908,
+      "step": 22520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.82070631837468e-06,
+      "loss": 0.7007,
+      "step": 22521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.819891484268513e-06,
+      "loss": 1.2139,
+      "step": 22522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8190767165468456e-06,
+      "loss": 0.8701,
+      "step": 22523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.818262015218427e-06,
+      "loss": 1.4043,
+      "step": 22524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.817447380292003e-06,
+      "loss": 0.8521,
+      "step": 22525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.816632811776332e-06,
+      "loss": 1.1523,
+      "step": 22526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.81581830968016e-06,
+      "loss": 0.7646,
+      "step": 22527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.815003874012233e-06,
+      "loss": 0.7554,
+      "step": 22528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.814189504781304e-06,
+      "loss": 0.8008,
+      "step": 22529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8133752019961213e-06,
+      "loss": 0.8418,
+      "step": 22530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8125609656654307e-06,
+      "loss": 1.0957,
+      "step": 22531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.811746795797975e-06,
+      "loss": 1.3496,
+      "step": 22532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8109326924025014e-06,
+      "loss": 1.1914,
+      "step": 22533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.810118655487759e-06,
+      "loss": 1.4766,
+      "step": 22534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8093046850624895e-06,
+      "loss": 0.5723,
+      "step": 22535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.808490781135432e-06,
+      "loss": 0.7402,
+      "step": 22536
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.807676943715335e-06,
+      "loss": 0.6177,
+      "step": 22537
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8068631728109364e-06,
+      "loss": 1.0439,
+      "step": 22538
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.80604946843098e-06,
+      "loss": 0.9766,
+      "step": 22539
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8052358305842043e-06,
+      "loss": 0.7505,
+      "step": 22540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8044222592793522e-06,
+      "loss": 0.8169,
+      "step": 22541
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.803608754525161e-06,
+      "loss": 0.8369,
+      "step": 22542
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8027953163303655e-06,
+      "loss": 0.7295,
+      "step": 22543
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8019819447037077e-06,
+      "loss": 1.2012,
+      "step": 22544
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.801168639653926e-06,
+      "loss": 1.4375,
+      "step": 22545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8003554011897536e-06,
+      "loss": 0.7598,
+      "step": 22546
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7995422293199247e-06,
+      "loss": 1.0791,
+      "step": 22547
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.798729124053174e-06,
+      "loss": 0.5156,
+      "step": 22548
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7979160853982434e-06,
+      "loss": 0.8892,
+      "step": 22549
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7971031133638593e-06,
+      "loss": 0.8125,
+      "step": 22550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7962902079587516e-06,
+      "loss": 1.0361,
+      "step": 22551
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.79547736919166e-06,
+      "loss": 1.0811,
+      "step": 22552
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.794664597071308e-06,
+      "loss": 1.2734,
+      "step": 22553
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7938518916064335e-06,
+      "loss": 0.9443,
+      "step": 22554
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7930392528057604e-06,
+      "loss": 0.7769,
+      "step": 22555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.792226680678023e-06,
+      "loss": 1.6992,
+      "step": 22556
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7914141752319457e-06,
+      "loss": 0.7021,
+      "step": 22557
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.790601736476255e-06,
+      "loss": 0.6943,
+      "step": 22558
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.789789364419679e-06,
+      "loss": 0.9902,
+      "step": 22559
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7889770590709496e-06,
+      "loss": 0.96,
+      "step": 22560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7881648204387866e-06,
+      "loss": 1.2031,
+      "step": 22561
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7873526485319135e-06,
+      "loss": 1.1445,
+      "step": 22562
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7865405433590597e-06,
+      "loss": 1.1377,
+      "step": 22563
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7857285049289415e-06,
+      "loss": 1.1685,
+      "step": 22564
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.784916533250289e-06,
+      "loss": 1.1768,
+      "step": 22565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7841046283318173e-06,
+      "loss": 0.9268,
+      "step": 22566
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7832927901822545e-06,
+      "loss": 0.8652,
+      "step": 22567
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7824810188103176e-06,
+      "loss": 0.8896,
+      "step": 22568
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.781669314224723e-06,
+      "loss": 1.3994,
+      "step": 22569
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.780857676434193e-06,
+      "loss": 0.8779,
+      "step": 22570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.780046105447449e-06,
+      "loss": 0.7173,
+      "step": 22571
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.779234601273206e-06,
+      "loss": 1.1914,
+      "step": 22572
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7784231639201763e-06,
+      "loss": 1.0469,
+      "step": 22573
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7776117933970814e-06,
+      "loss": 1.1523,
+      "step": 22574
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7768004897126376e-06,
+      "loss": 0.8047,
+      "step": 22575
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7759892528755584e-06,
+      "loss": 1.3887,
+      "step": 22576
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.775178082894553e-06,
+      "loss": 0.584,
+      "step": 22577
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7743669797783424e-06,
+      "loss": 0.8809,
+      "step": 22578
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7735559435356316e-06,
+      "loss": 0.9033,
+      "step": 22579
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7727449741751365e-06,
+      "loss": 1.498,
+      "step": 22580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.771934071705572e-06,
+      "loss": 0.6025,
+      "step": 22581
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.771123236135644e-06,
+      "loss": 1.0459,
+      "step": 22582
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7703124674740633e-06,
+      "loss": 1.021,
+      "step": 22583
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.769501765729531e-06,
+      "loss": 0.7939,
+      "step": 22584
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.76869113091077e-06,
+      "loss": 0.9229,
+      "step": 22585
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7678805630264802e-06,
+      "loss": 1.0977,
+      "step": 22586
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.767070062085365e-06,
+      "loss": 0.9297,
+      "step": 22587
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.766259628096138e-06,
+      "loss": 0.9395,
+      "step": 22588
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7654492610675007e-06,
+      "loss": 0.8047,
+      "step": 22589
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7646389610081558e-06,
+      "loss": 1.1709,
+      "step": 22590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.763828727926808e-06,
+      "loss": 1.3711,
+      "step": 22591
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.763018561832166e-06,
+      "loss": 1.4512,
+      "step": 22592
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.762208462732928e-06,
+      "loss": 0.7139,
+      "step": 22593
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.761398430637796e-06,
+      "loss": 0.9912,
+      "step": 22594
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7605884655554647e-06,
+      "loss": 0.9639,
+      "step": 22595
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7597785674946476e-06,
+      "loss": 0.8799,
+      "step": 22596
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.758968736464037e-06,
+      "loss": 1.1035,
+      "step": 22597
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.75815897247233e-06,
+      "loss": 1.0508,
+      "step": 22598
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.75734927552823e-06,
+      "loss": 0.9648,
+      "step": 22599
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.756539645640428e-06,
+      "loss": 0.8848,
+      "step": 22600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.755730082817628e-06,
+      "loss": 0.9961,
+      "step": 22601
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.75492058706852e-06,
+      "loss": 1.3535,
+      "step": 22602
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7541111584018043e-06,
+      "loss": 0.9639,
+      "step": 22603
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.753301796826173e-06,
+      "loss": 1.1094,
+      "step": 22604
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7524925023503165e-06,
+      "loss": 0.8052,
+      "step": 22605
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7516832749829314e-06,
+      "loss": 1.1865,
+      "step": 22606
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7508741147327134e-06,
+      "loss": 0.6135,
+      "step": 22607
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7500650216083513e-06,
+      "loss": 0.7334,
+      "step": 22608
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7492559956185327e-06,
+      "loss": 1.0342,
+      "step": 22609
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7484470367719493e-06,
+      "loss": 1.2871,
+      "step": 22610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.747638145077297e-06,
+      "loss": 0.9004,
+      "step": 22611
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7468293205432593e-06,
+      "loss": 1.2559,
+      "step": 22612
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.746020563178522e-06,
+      "loss": 1.3203,
+      "step": 22613
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7452118729917786e-06,
+      "loss": 1.1836,
+      "step": 22614
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.744403249991708e-06,
+      "loss": 0.873,
+      "step": 22615
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.743594694187006e-06,
+      "loss": 1.1143,
+      "step": 22616
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7427862055863474e-06,
+      "loss": 1.208,
+      "step": 22617
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.741977784198426e-06,
+      "loss": 1.459,
+      "step": 22618
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.74116943003192e-06,
+      "loss": 1.4287,
+      "step": 22619
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7403611430955123e-06,
+      "loss": 0.8123,
+      "step": 22620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7395529233978854e-06,
+      "loss": 1.2178,
+      "step": 22621
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7387447709477255e-06,
+      "loss": 0.7529,
+      "step": 22622
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7379366857537105e-06,
+      "loss": 1.4531,
+      "step": 22623
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.737128667824517e-06,
+      "loss": 0.8975,
+      "step": 22624
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.736320717168831e-06,
+      "loss": 1.2266,
+      "step": 22625
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.735512833795324e-06,
+      "loss": 0.8936,
+      "step": 22626
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.734705017712682e-06,
+      "loss": 0.6753,
+      "step": 22627
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.733897268929574e-06,
+      "loss": 0.9648,
+      "step": 22628
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7330895874546847e-06,
+      "loss": 1.4756,
+      "step": 22629
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.732281973296685e-06,
+      "loss": 1.0664,
+      "step": 22630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7314744264642477e-06,
+      "loss": 0.9033,
+      "step": 22631
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7306669469660507e-06,
+      "loss": 0.8213,
+      "step": 22632
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7298595348107715e-06,
+      "loss": 1.0264,
+      "step": 22633
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7290521900070774e-06,
+      "loss": 1.0107,
+      "step": 22634
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.72824491256364e-06,
+      "loss": 1.1621,
+      "step": 22635
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7274377024891317e-06,
+      "loss": 0.8252,
+      "step": 22636
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7266305597922282e-06,
+      "loss": 1.2222,
+      "step": 22637
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7258234844815967e-06,
+      "loss": 0.8271,
+      "step": 22638
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7250164765659014e-06,
+      "loss": 1.1543,
+      "step": 22639
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7242095360538176e-06,
+      "loss": 1.0771,
+      "step": 22640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7234026629540076e-06,
+      "loss": 1.1152,
+      "step": 22641
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7225958572751462e-06,
+      "loss": 0.7568,
+      "step": 22642
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7217891190258905e-06,
+      "loss": 0.8735,
+      "step": 22643
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.720982448214915e-06,
+      "loss": 0.8896,
+      "step": 22644
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7201758448508798e-06,
+      "loss": 0.7861,
+      "step": 22645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7193693089424465e-06,
+      "loss": 0.9766,
+      "step": 22646
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7185628404982834e-06,
+      "loss": 0.8145,
+      "step": 22647
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.717756439527054e-06,
+      "loss": 0.9092,
+      "step": 22648
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.716950106037418e-06,
+      "loss": 0.9893,
+      "step": 22649
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7161438400380336e-06,
+      "loss": 0.959,
+      "step": 22650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7153376415375685e-06,
+      "loss": 0.8232,
+      "step": 22651
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7145315105446756e-06,
+      "loss": 1.3682,
+      "step": 22652
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7137254470680208e-06,
+      "loss": 0.6914,
+      "step": 22653
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7129194511162557e-06,
+      "loss": 0.9863,
+      "step": 22654
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7121135226980455e-06,
+      "loss": 0.6731,
+      "step": 22655
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.711307661822042e-06,
+      "loss": 1.2041,
+      "step": 22656
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7105018684968995e-06,
+      "loss": 0.6826,
+      "step": 22657
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.709696142731277e-06,
+      "loss": 0.7349,
+      "step": 22658
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7088904845338315e-06,
+      "loss": 0.8145,
+      "step": 22659
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7080848939132153e-06,
+      "loss": 0.7891,
+      "step": 22660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7072793708780775e-06,
+      "loss": 0.9897,
+      "step": 22661
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7064739154370744e-06,
+      "loss": 0.9189,
+      "step": 22662
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.70566852759886e-06,
+      "loss": 1.0752,
+      "step": 22663
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7048632073720846e-06,
+      "loss": 0.9102,
+      "step": 22664
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7040579547653933e-06,
+      "loss": 0.9761,
+      "step": 22665
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.703252769787443e-06,
+      "loss": 0.8809,
+      "step": 22666
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7024476524468768e-06,
+      "loss": 1.0117,
+      "step": 22667
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.701642602752349e-06,
+      "loss": 0.51,
+      "step": 22668
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7008376207124994e-06,
+      "loss": 0.8799,
+      "step": 22669
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.700032706335984e-06,
+      "loss": 0.7007,
+      "step": 22670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6992278596314436e-06,
+      "loss": 1.3203,
+      "step": 22671
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6984230806075173e-06,
+      "loss": 0.8804,
+      "step": 22672
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.697618369272864e-06,
+      "loss": 0.7666,
+      "step": 22673
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6968137256361192e-06,
+      "loss": 1.0615,
+      "step": 22674
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.696009149705928e-06,
+      "loss": 0.7998,
+      "step": 22675
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.695204641490928e-06,
+      "loss": 0.9307,
+      "step": 22676
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.694400200999765e-06,
+      "loss": 0.9058,
+      "step": 22677
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.693595828241083e-06,
+      "loss": 1.1396,
+      "step": 22678
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6927915232235166e-06,
+      "loss": 0.8003,
+      "step": 22679
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6919872859557114e-06,
+      "loss": 0.5972,
+      "step": 22680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.691183116446303e-06,
+      "loss": 0.9399,
+      "step": 22681
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6903790147039286e-06,
+      "loss": 0.8076,
+      "step": 22682
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6895749807372194e-06,
+      "loss": 0.772,
+      "step": 22683
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6887710145548263e-06,
+      "loss": 1.1797,
+      "step": 22684
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6879671161653775e-06,
+      "loss": 0.6338,
+      "step": 22685
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6871632855775043e-06,
+      "loss": 0.8281,
+      "step": 22686
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6863595227998493e-06,
+      "loss": 0.7451,
+      "step": 22687
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6855558278410396e-06,
+      "loss": 0.9912,
+      "step": 22688
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6847522007097138e-06,
+      "loss": 0.9448,
+      "step": 22689
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6839486414144974e-06,
+      "loss": 0.98,
+      "step": 22690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6831451499640304e-06,
+      "loss": 1.1836,
+      "step": 22691
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.682341726366938e-06,
+      "loss": 1.416,
+      "step": 22692
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6815383706318488e-06,
+      "loss": 0.7451,
+      "step": 22693
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.680735082767394e-06,
+      "loss": 0.605,
+      "step": 22694
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6799318627822056e-06,
+      "loss": 1.1191,
+      "step": 22695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6791287106849094e-06,
+      "loss": 0.9912,
+      "step": 22696
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6783256264841293e-06,
+      "loss": 1.0781,
+      "step": 22697
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6775226101884932e-06,
+      "loss": 1.0322,
+      "step": 22698
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6767196618066324e-06,
+      "loss": 0.7759,
+      "step": 22699
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6759167813471675e-06,
+      "loss": 0.9951,
+      "step": 22700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6751139688187188e-06,
+      "loss": 1.3594,
+      "step": 22701
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.674311224229917e-06,
+      "loss": 1.0566,
+      "step": 22702
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6735085475893785e-06,
+      "loss": 1.0771,
+      "step": 22703
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.672705938905733e-06,
+      "loss": 0.897,
+      "step": 22704
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.671903398187593e-06,
+      "loss": 1.0049,
+      "step": 22705
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.671100925443586e-06,
+      "loss": 0.9355,
+      "step": 22706
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6702985206823303e-06,
+      "loss": 1.0254,
+      "step": 22707
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.669496183912441e-06,
+      "loss": 1.2842,
+      "step": 22708
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6686939151425393e-06,
+      "loss": 1.0645,
+      "step": 22709
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6678917143812464e-06,
+      "loss": 1.0361,
+      "step": 22710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6670895816371765e-06,
+      "loss": 0.6758,
+      "step": 22711
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6662875169189418e-06,
+      "loss": 0.7234,
+      "step": 22712
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6654855202351638e-06,
+      "loss": 1.3066,
+      "step": 22713
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6646835915944525e-06,
+      "loss": 1.124,
+      "step": 22714
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6638817310054266e-06,
+      "loss": 0.9648,
+      "step": 22715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6630799384766945e-06,
+      "loss": 1.0674,
+      "step": 22716
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6622782140168746e-06,
+      "loss": 0.8604,
+      "step": 22717
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6614765576345755e-06,
+      "loss": 1.0508,
+      "step": 22718
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6606749693384046e-06,
+      "loss": 1.1465,
+      "step": 22719
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.659873449136976e-06,
+      "loss": 0.9771,
+      "step": 22720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6590719970389023e-06,
+      "loss": 1.1719,
+      "step": 22721
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.658270613052789e-06,
+      "loss": 0.8252,
+      "step": 22722
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6574692971872418e-06,
+      "loss": 0.8284,
+      "step": 22723
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.656668049450871e-06,
+      "loss": 0.8857,
+      "step": 22724
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.655866869852287e-06,
+      "loss": 1.0771,
+      "step": 22725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6550657584000914e-06,
+      "loss": 0.8887,
+      "step": 22726
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6542647151028877e-06,
+      "loss": 1.3174,
+      "step": 22727
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6534637399692853e-06,
+      "loss": 0.8071,
+      "step": 22728
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6526628330078816e-06,
+      "loss": 1.2324,
+      "step": 22729
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6518619942272883e-06,
+      "loss": 0.7388,
+      "step": 22730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6510612236360986e-06,
+      "loss": 1.2852,
+      "step": 22731
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6502605212429234e-06,
+      "loss": 1.1807,
+      "step": 22732
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6494598870563567e-06,
+      "loss": 0.7207,
+      "step": 22733
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6486593210849985e-06,
+      "loss": 1.2812,
+      "step": 22734
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.64785882333745e-06,
+      "loss": 0.8506,
+      "step": 22735
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6470583938223127e-06,
+      "loss": 0.9331,
+      "step": 22736
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.646258032548183e-06,
+      "loss": 0.8857,
+      "step": 22737
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6454577395236533e-06,
+      "loss": 0.9062,
+      "step": 22738
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.644657514757324e-06,
+      "loss": 1.0557,
+      "step": 22739
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.643857358257793e-06,
+      "loss": 0.9375,
+      "step": 22740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.643057270033654e-06,
+      "loss": 0.7656,
+      "step": 22741
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.642257250093496e-06,
+      "loss": 1.1846,
+      "step": 22742
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6414572984459206e-06,
+      "loss": 0.8628,
+      "step": 22743
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6406574150995166e-06,
+      "loss": 0.666,
+      "step": 22744
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6398576000628727e-06,
+      "loss": 1.1406,
+      "step": 22745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.639057853344583e-06,
+      "loss": 0.8237,
+      "step": 22746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.638258174953243e-06,
+      "loss": 0.9414,
+      "step": 22747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.637458564897438e-06,
+      "loss": 0.9233,
+      "step": 22748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6366590231857547e-06,
+      "loss": 1.1758,
+      "step": 22749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6358595498267833e-06,
+      "loss": 0.7871,
+      "step": 22750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.635060144829117e-06,
+      "loss": 0.8916,
+      "step": 22751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6342608082013366e-06,
+      "loss": 1.0332,
+      "step": 22752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6334615399520277e-06,
+      "loss": 1.0566,
+      "step": 22753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6326623400897797e-06,
+      "loss": 0.8633,
+      "step": 22754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.631863208623173e-06,
+      "loss": 1.2812,
+      "step": 22755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6310641455607978e-06,
+      "loss": 0.749,
+      "step": 22756
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6302651509112298e-06,
+      "loss": 1.0703,
+      "step": 22757
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6294662246830592e-06,
+      "loss": 1.1777,
+      "step": 22758
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.628667366884863e-06,
+      "loss": 1.1777,
+      "step": 22759
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6278685775252166e-06,
+      "loss": 0.8799,
+      "step": 22760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.627069856612714e-06,
+      "loss": 0.7422,
+      "step": 22761
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6262712041559265e-06,
+      "loss": 0.6904,
+      "step": 22762
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.625472620163435e-06,
+      "loss": 1.0234,
+      "step": 22763
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6246741046438127e-06,
+      "loss": 0.8242,
+      "step": 22764
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6238756576056424e-06,
+      "loss": 0.8008,
+      "step": 22765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6230772790575008e-06,
+      "loss": 0.9546,
+      "step": 22766
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.622278969007963e-06,
+      "loss": 1.1753,
+      "step": 22767
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6214807274655992e-06,
+      "loss": 1.0068,
+      "step": 22768
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.620682554438992e-06,
+      "loss": 1.293,
+      "step": 22769
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.619884449936708e-06,
+      "loss": 0.5806,
+      "step": 22770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.619086413967323e-06,
+      "loss": 0.9941,
+      "step": 22771
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.618288446539413e-06,
+      "loss": 0.9453,
+      "step": 22772
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.617490547661546e-06,
+      "loss": 1.127,
+      "step": 22773
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.616692717342293e-06,
+      "loss": 1.0518,
+      "step": 22774
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.615894955590219e-06,
+      "loss": 0.8574,
+      "step": 22775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.615097262413899e-06,
+      "loss": 0.9316,
+      "step": 22776
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.614299637821903e-06,
+      "loss": 0.9507,
+      "step": 22777
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.613502081822794e-06,
+      "loss": 1.0361,
+      "step": 22778
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6127045944251437e-06,
+      "loss": 0.626,
+      "step": 22779
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6119071756375168e-06,
+      "loss": 0.769,
+      "step": 22780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6111098254684753e-06,
+      "loss": 1.5547,
+      "step": 22781
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6103125439265853e-06,
+      "loss": 1.2646,
+      "step": 22782
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6095153310204166e-06,
+      "loss": 0.7539,
+      "step": 22783
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6087181867585286e-06,
+      "loss": 1.1963,
+      "step": 22784
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6079211111494838e-06,
+      "loss": 0.7046,
+      "step": 22785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.607124104201837e-06,
+      "loss": 0.9697,
+      "step": 22786
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.606327165924164e-06,
+      "loss": 0.9023,
+      "step": 22787
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6055302963250174e-06,
+      "loss": 1.1099,
+      "step": 22788
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.604733495412953e-06,
+      "loss": 1.2822,
+      "step": 22789
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6039367631965383e-06,
+      "loss": 1.1436,
+      "step": 22790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6031400996843236e-06,
+      "loss": 1.3857,
+      "step": 22791
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6023435048848734e-06,
+      "loss": 0.8721,
+      "step": 22792
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.601546978806737e-06,
+      "loss": 0.7695,
+      "step": 22793
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6007505214584772e-06,
+      "loss": 0.7773,
+      "step": 22794
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5999541328486477e-06,
+      "loss": 0.8159,
+      "step": 22795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5991578129857973e-06,
+      "loss": 0.8721,
+      "step": 22796
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.598361561878485e-06,
+      "loss": 1.1543,
+      "step": 22797
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.597565379535266e-06,
+      "loss": 1.0449,
+      "step": 22798
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.596769265964689e-06,
+      "loss": 0.9551,
+      "step": 22799
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.595973221175304e-06,
+      "loss": 1.3145,
+      "step": 22800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.595177245175664e-06,
+      "loss": 1.4512,
+      "step": 22801
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.594381337974322e-06,
+      "loss": 0.6772,
+      "step": 22802
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5935854995798248e-06,
+      "loss": 1.0222,
+      "step": 22803
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5927897300007165e-06,
+      "loss": 1.1875,
+      "step": 22804
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5919940292455535e-06,
+      "loss": 0.9443,
+      "step": 22805
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.591198397322878e-06,
+      "loss": 1.1924,
+      "step": 22806
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5904028342412336e-06,
+      "loss": 1.0479,
+      "step": 22807
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.589607340009169e-06,
+      "loss": 0.8857,
+      "step": 22808
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.588811914635234e-06,
+      "loss": 1.0586,
+      "step": 22809
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5880165581279667e-06,
+      "loss": 0.8535,
+      "step": 22810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.58722127049591e-06,
+      "loss": 0.8691,
+      "step": 22811
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.586426051747608e-06,
+      "loss": 0.6627,
+      "step": 22812
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.585630901891607e-06,
+      "loss": 0.9033,
+      "step": 22813
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5848358209364444e-06,
+      "loss": 0.875,
+      "step": 22814
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5840408088906565e-06,
+      "loss": 1.1719,
+      "step": 22815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.583245865762791e-06,
+      "loss": 0.7549,
+      "step": 22816
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.58245099156138e-06,
+      "loss": 1.3223,
+      "step": 22817
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5816561862949695e-06,
+      "loss": 0.916,
+      "step": 22818
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5808614499720873e-06,
+      "loss": 1.0693,
+      "step": 22819
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.58006678260128e-06,
+      "loss": 0.9204,
+      "step": 22820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5792721841910784e-06,
+      "loss": 0.958,
+      "step": 22821
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5784776547500144e-06,
+      "loss": 1.0703,
+      "step": 22822
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5776831942866274e-06,
+      "loss": 0.8838,
+      "step": 22823
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5768888028094537e-06,
+      "loss": 0.8003,
+      "step": 22824
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5760944803270227e-06,
+      "loss": 0.5117,
+      "step": 22825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.575300226847863e-06,
+      "loss": 0.7559,
+      "step": 22826
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5745060423805113e-06,
+      "loss": 1.252,
+      "step": 22827
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5737119269335e-06,
+      "loss": 1.0117,
+      "step": 22828
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5729178805153562e-06,
+      "loss": 0.6626,
+      "step": 22829
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5721239031346067e-06,
+      "loss": 1.123,
+      "step": 22830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.571329994799787e-06,
+      "loss": 1.0264,
+      "step": 22831
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5705361555194196e-06,
+      "loss": 0.6382,
+      "step": 22832
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5697423853020306e-06,
+      "loss": 0.9092,
+      "step": 22833
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5689486841561482e-06,
+      "loss": 0.9668,
+      "step": 22834
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.568155052090303e-06,
+      "loss": 1.1074,
+      "step": 22835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.567361489113015e-06,
+      "loss": 0.9775,
+      "step": 22836
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5665679952328057e-06,
+      "loss": 1.2354,
+      "step": 22837
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5657745704582013e-06,
+      "loss": 0.752,
+      "step": 22838
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.564981214797728e-06,
+      "loss": 1.3037,
+      "step": 22839
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5641879282599057e-06,
+      "loss": 1.6807,
+      "step": 22840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5633947108532507e-06,
+      "loss": 0.8975,
+      "step": 22841
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5626015625862907e-06,
+      "loss": 0.5342,
+      "step": 22842
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.561808483467539e-06,
+      "loss": 1.0176,
+      "step": 22843
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5610154735055203e-06,
+      "loss": 1.1353,
+      "step": 22844
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5602225327087483e-06,
+      "loss": 0.7451,
+      "step": 22845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5594296610857435e-06,
+      "loss": 0.9312,
+      "step": 22846
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5586368586450226e-06,
+      "loss": 1.126,
+      "step": 22847
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.557844125395098e-06,
+      "loss": 0.7891,
+      "step": 22848
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5570514613444864e-06,
+      "loss": 1.082,
+      "step": 22849
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5562588665017063e-06,
+      "loss": 0.9775,
+      "step": 22850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5554663408752688e-06,
+      "loss": 0.9561,
+      "step": 22851
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.554673884473684e-06,
+      "loss": 0.668,
+      "step": 22852
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.553881497305466e-06,
+      "loss": 0.9033,
+      "step": 22853
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5530891793791313e-06,
+      "loss": 1.0078,
+      "step": 22854
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.552296930703185e-06,
+      "loss": 1.2236,
+      "step": 22855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.551504751286137e-06,
+      "loss": 0.9346,
+      "step": 22856
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5507126411365012e-06,
+      "loss": 1.123,
+      "step": 22857
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5499206002627796e-06,
+      "loss": 0.9092,
+      "step": 22858
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5491286286734836e-06,
+      "loss": 1.1914,
+      "step": 22859
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.548336726377124e-06,
+      "loss": 0.9395,
+      "step": 22860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5475448933822033e-06,
+      "loss": 0.8848,
+      "step": 22861
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.546753129697227e-06,
+      "loss": 1.1436,
+      "step": 22862
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.545961435330697e-06,
+      "loss": 0.9219,
+      "step": 22863
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5451698102911213e-06,
+      "loss": 0.9883,
+      "step": 22864
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5443782545870053e-06,
+      "loss": 1.2031,
+      "step": 22865
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5435867682268487e-06,
+      "loss": 0.9351,
+      "step": 22866
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5427953512191503e-06,
+      "loss": 1.4141,
+      "step": 22867
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.542004003572418e-06,
+      "loss": 0.8267,
+      "step": 22868
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5412127252951444e-06,
+      "loss": 1.3311,
+      "step": 22869
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5404215163958343e-06,
+      "loss": 1.0518,
+      "step": 22870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5396303768829887e-06,
+      "loss": 0.9277,
+      "step": 22871
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.538839306765103e-06,
+      "loss": 0.8447,
+      "step": 22872
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5380483060506733e-06,
+      "loss": 0.8633,
+      "step": 22873
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5372573747481907e-06,
+      "loss": 0.8096,
+      "step": 22874
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5364665128661645e-06,
+      "loss": 0.9502,
+      "step": 22875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.535675720413083e-06,
+      "loss": 1.0225,
+      "step": 22876
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5348849973974375e-06,
+      "loss": 0.7026,
+      "step": 22877
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5340943438277275e-06,
+      "loss": 1.2686,
+      "step": 22878
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5333037597124397e-06,
+      "loss": 0.8809,
+      "step": 22879
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.532513245060073e-06,
+      "loss": 1.0771,
+      "step": 22880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5317227998791124e-06,
+      "loss": 0.7573,
+      "step": 22881
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.530932424178055e-06,
+      "loss": 0.999,
+      "step": 22882
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5301421179653873e-06,
+      "loss": 0.7314,
+      "step": 22883
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.529351881249595e-06,
+      "loss": 1.0303,
+      "step": 22884
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5285617140391713e-06,
+      "loss": 1.207,
+      "step": 22885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5277716163426044e-06,
+      "loss": 0.7339,
+      "step": 22886
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5269815881683812e-06,
+      "loss": 1.2441,
+      "step": 22887
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.526191629524982e-06,
+      "loss": 1.3828,
+      "step": 22888
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5254017404208963e-06,
+      "loss": 0.79,
+      "step": 22889
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5246119208646124e-06,
+      "loss": 1.1289,
+      "step": 22890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5238221708646116e-06,
+      "loss": 0.7915,
+      "step": 22891
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5230324904293723e-06,
+      "loss": 1.0215,
+      "step": 22892
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5222428795673847e-06,
+      "loss": 0.8398,
+      "step": 22893
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5214533382871272e-06,
+      "loss": 0.9355,
+      "step": 22894
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5206638665970774e-06,
+      "loss": 1.1807,
+      "step": 22895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.519874464505717e-06,
+      "loss": 0.6265,
+      "step": 22896
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5190851320215313e-06,
+      "loss": 0.9678,
+      "step": 22897
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5182958691529945e-06,
+      "loss": 1.2373,
+      "step": 22898
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5175066759085817e-06,
+      "loss": 0.8262,
+      "step": 22899
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5167175522967735e-06,
+      "loss": 0.8228,
+      "step": 22900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5159284983260487e-06,
+      "loss": 0.6558,
+      "step": 22901
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5151395140048815e-06,
+      "loss": 1.0376,
+      "step": 22902
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5143505993417416e-06,
+      "loss": 1.1709,
+      "step": 22903
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.513561754345112e-06,
+      "loss": 0.8755,
+      "step": 22904
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5127729790234564e-06,
+      "loss": 1.0137,
+      "step": 22905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5119842733852573e-06,
+      "loss": 1.1953,
+      "step": 22906
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5111956374389787e-06,
+      "loss": 0.7202,
+      "step": 22907
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.510407071193099e-06,
+      "loss": 1.0674,
+      "step": 22908
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.509618574656085e-06,
+      "loss": 0.7905,
+      "step": 22909
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.508830147836403e-06,
+      "loss": 0.8809,
+      "step": 22910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.508041790742526e-06,
+      "loss": 0.9971,
+      "step": 22911
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.507253503382925e-06,
+      "loss": 0.6567,
+      "step": 22912
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.506465285766063e-06,
+      "loss": 0.6196,
+      "step": 22913
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.505677137900405e-06,
+      "loss": 1.0054,
+      "step": 22914
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5048890597944207e-06,
+      "loss": 1.1616,
+      "step": 22915
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5041010514565776e-06,
+      "loss": 1.1685,
+      "step": 22916
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.503313112895338e-06,
+      "loss": 1.0674,
+      "step": 22917
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.50252524411916e-06,
+      "loss": 0.959,
+      "step": 22918
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.501737445136516e-06,
+      "loss": 0.9563,
+      "step": 22919
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5009497159558604e-06,
+      "loss": 0.9966,
+      "step": 22920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.500162056585661e-06,
+      "loss": 0.8037,
+      "step": 22921
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4993744670343723e-06,
+      "loss": 1.4453,
+      "step": 22922
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4985869473104605e-06,
+      "loss": 1.2656,
+      "step": 22923
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4977994974223827e-06,
+      "loss": 1.0537,
+      "step": 22924
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.497012117378593e-06,
+      "loss": 1.3096,
+      "step": 22925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4962248071875525e-06,
+      "loss": 0.8564,
+      "step": 22926
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4954375668577222e-06,
+      "loss": 0.7729,
+      "step": 22927
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4946503963975543e-06,
+      "loss": 1.1611,
+      "step": 22928
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.493863295815502e-06,
+      "loss": 0.9346,
+      "step": 22929
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4930762651200257e-06,
+      "loss": 0.7688,
+      "step": 22930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4922893043195726e-06,
+      "loss": 0.6963,
+      "step": 22931
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.491502413422604e-06,
+      "loss": 1.0635,
+      "step": 22932
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4907155924375646e-06,
+      "loss": 0.9116,
+      "step": 22933
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4899288413729128e-06,
+      "loss": 1.1934,
+      "step": 22934
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4891421602370958e-06,
+      "loss": 0.9141,
+      "step": 22935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4883555490385624e-06,
+      "loss": 0.8743,
+      "step": 22936
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.487569007785764e-06,
+      "loss": 0.5818,
+      "step": 22937
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4867825364871523e-06,
+      "loss": 1.2471,
+      "step": 22938
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4859961351511727e-06,
+      "loss": 0.9585,
+      "step": 22939
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.485209803786269e-06,
+      "loss": 1.2061,
+      "step": 22940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.484423542400891e-06,
+      "loss": 1.0869,
+      "step": 22941
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.483637351003488e-06,
+      "loss": 1.0898,
+      "step": 22942
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.482851229602502e-06,
+      "loss": 0.8154,
+      "step": 22943
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4820651782063717e-06,
+      "loss": 0.7671,
+      "step": 22944
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4812791968235506e-06,
+      "loss": 0.9287,
+      "step": 22945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4804932854624716e-06,
+      "loss": 0.9248,
+      "step": 22946
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.479707444131586e-06,
+      "loss": 1.1914,
+      "step": 22947
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.478921672839326e-06,
+      "loss": 0.8623,
+      "step": 22948
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.478135971594141e-06,
+      "loss": 0.8105,
+      "step": 22949
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4773503404044652e-06,
+      "loss": 1.124,
+      "step": 22950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.476564779278733e-06,
+      "loss": 0.9727,
+      "step": 22951
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.475779288225394e-06,
+      "loss": 0.5864,
+      "step": 22952
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4749938672528805e-06,
+      "loss": 0.9941,
+      "step": 22953
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4742085163696283e-06,
+      "loss": 0.9141,
+      "step": 22954
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.47342323558407e-06,
+      "loss": 0.9023,
+      "step": 22955
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.472638024904649e-06,
+      "loss": 0.8545,
+      "step": 22956
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.47185288433979e-06,
+      "loss": 0.792,
+      "step": 22957
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4710678138979326e-06,
+      "loss": 0.9199,
+      "step": 22958
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4702828135875133e-06,
+      "loss": 0.8477,
+      "step": 22959
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.469497883416959e-06,
+      "loss": 0.9131,
+      "step": 22960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4687130233947032e-06,
+      "loss": 1.0723,
+      "step": 22961
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4679282335291674e-06,
+      "loss": 0.8345,
+      "step": 22962
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.467143513828798e-06,
+      "loss": 1.1729,
+      "step": 22963
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4663588643020153e-06,
+      "loss": 0.8169,
+      "step": 22964
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4655742849572482e-06,
+      "loss": 1.0356,
+      "step": 22965
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4647897758029204e-06,
+      "loss": 0.5347,
+      "step": 22966
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4640053368474637e-06,
+      "loss": 0.9814,
+      "step": 22967
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4632209680993065e-06,
+      "loss": 0.9307,
+      "step": 22968
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4624366695668675e-06,
+      "loss": 0.998,
+      "step": 22969
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4616524412585797e-06,
+      "loss": 1.0566,
+      "step": 22970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.460868283182861e-06,
+      "loss": 0.9702,
+      "step": 22971
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4600841953481323e-06,
+      "loss": 1.4014,
+      "step": 22972
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.45930017776282e-06,
+      "loss": 1.0386,
+      "step": 22973
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4585162304353493e-06,
+      "loss": 0.9707,
+      "step": 22974
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4577323533741356e-06,
+      "loss": 0.9497,
+      "step": 22975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4569485465875986e-06,
+      "loss": 1.0479,
+      "step": 22976
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4561648100841583e-06,
+      "loss": 1.0459,
+      "step": 22977
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4553811438722384e-06,
+      "loss": 1.0586,
+      "step": 22978
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.454597547960252e-06,
+      "loss": 0.7861,
+      "step": 22979
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4538140223566154e-06,
+      "loss": 0.9189,
+      "step": 22980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4530305670697483e-06,
+      "loss": 1.2227,
+      "step": 22981
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4522471821080628e-06,
+      "loss": 0.7603,
+      "step": 22982
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4514638674799782e-06,
+      "loss": 1.3154,
+      "step": 22983
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4506806231939027e-06,
+      "loss": 0.8945,
+      "step": 22984
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.449897449258256e-06,
+      "loss": 1.1875,
+      "step": 22985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.449114345681448e-06,
+      "loss": 1.1963,
+      "step": 22986
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4483313124718865e-06,
+      "loss": 1.3496,
+      "step": 22987
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.447548349637987e-06,
+      "loss": 0.7241,
+      "step": 22988
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4467654571881613e-06,
+      "loss": 0.71,
+      "step": 22989
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.445982635130817e-06,
+      "loss": 0.9883,
+      "step": 22990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.445199883474358e-06,
+      "loss": 1.2432,
+      "step": 22991
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4444172022272017e-06,
+      "loss": 0.6572,
+      "step": 22992
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4436345913977465e-06,
+      "loss": 0.7354,
+      "step": 22993
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4428520509944065e-06,
+      "loss": 0.8225,
+      "step": 22994
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.44206958102558e-06,
+      "loss": 0.9067,
+      "step": 22995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.441287181499681e-06,
+      "loss": 0.8711,
+      "step": 22996
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.440504852425106e-06,
+      "loss": 0.6855,
+      "step": 22997
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.43972259381026e-06,
+      "loss": 1.3311,
+      "step": 22998
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4389404056635454e-06,
+      "loss": 1.0273,
+      "step": 22999
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4381582879933693e-06,
+      "loss": 0.6118,
+      "step": 23000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.43737624080813e-06,
+      "loss": 0.8799,
+      "step": 23001
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.436594264116224e-06,
+      "loss": 0.7568,
+      "step": 23002
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.435812357926054e-06,
+      "loss": 0.8965,
+      "step": 23003
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.435030522246022e-06,
+      "loss": 1.105,
+      "step": 23004
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.434248757084524e-06,
+      "loss": 1.2529,
+      "step": 23005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.433467062449953e-06,
+      "loss": 0.9375,
+      "step": 23006
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4326854383507137e-06,
+      "loss": 0.9385,
+      "step": 23007
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4319038847951948e-06,
+      "loss": 1.5254,
+      "step": 23008
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4311224017917966e-06,
+      "loss": 0.6914,
+      "step": 23009
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4303409893489093e-06,
+      "loss": 1.1943,
+      "step": 23010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4295596474749305e-06,
+      "loss": 0.7646,
+      "step": 23011
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.428778376178252e-06,
+      "loss": 0.8887,
+      "step": 23012
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4279971754672624e-06,
+      "loss": 0.627,
+      "step": 23013
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4272160453503566e-06,
+      "loss": 1.0674,
+      "step": 23014
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4264349858359267e-06,
+      "loss": 1.373,
+      "step": 23015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4256539969323597e-06,
+      "loss": 1.0674,
+      "step": 23016
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4248730786480433e-06,
+      "loss": 1.5352,
+      "step": 23017
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4240922309913715e-06,
+      "loss": 0.8301,
+      "step": 23018
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4233114539707246e-06,
+      "loss": 0.8979,
+      "step": 23019
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4225307475944957e-06,
+      "loss": 1.0117,
+      "step": 23020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4217501118710662e-06,
+      "loss": 1.0957,
+      "step": 23021
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4209695468088256e-06,
+      "loss": 0.8042,
+      "step": 23022
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4201890524161574e-06,
+      "loss": 0.7471,
+      "step": 23023
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4194086287014405e-06,
+      "loss": 1.0605,
+      "step": 23024
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4186282756730616e-06,
+      "loss": 0.9746,
+      "step": 23025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.417847993339407e-06,
+      "loss": 0.8398,
+      "step": 23026
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4170677817088537e-06,
+      "loss": 0.8672,
+      "step": 23027
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4162876407897806e-06,
+      "loss": 0.9644,
+      "step": 23028
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4155075705905695e-06,
+      "loss": 1.3027,
+      "step": 23029
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.414727571119604e-06,
+      "loss": 0.7063,
+      "step": 23030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4139476423852592e-06,
+      "loss": 1.2139,
+      "step": 23031
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.413167784395909e-06,
+      "loss": 0.8359,
+      "step": 23032
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4123879971599373e-06,
+      "loss": 1.168,
+      "step": 23033
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.411608280685713e-06,
+      "loss": 0.8838,
+      "step": 23034
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4108286349816188e-06,
+      "loss": 0.9619,
+      "step": 23035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4100490600560232e-06,
+      "loss": 1.0762,
+      "step": 23036
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.409269555917305e-06,
+      "loss": 0.7959,
+      "step": 23037
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4084901225738365e-06,
+      "loss": 0.9785,
+      "step": 23038
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4077107600339856e-06,
+      "loss": 1.0098,
+      "step": 23039
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.406931468306126e-06,
+      "loss": 1.2891,
+      "step": 23040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4061522473986342e-06,
+      "loss": 0.5245,
+      "step": 23041
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.405373097319875e-06,
+      "loss": 0.917,
+      "step": 23042
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4045940180782155e-06,
+      "loss": 1.0625,
+      "step": 23043
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4038150096820276e-06,
+      "loss": 1.3672,
+      "step": 23044
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4030360721396828e-06,
+      "loss": 1.0596,
+      "step": 23045
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.402257205459544e-06,
+      "loss": 1.1221,
+      "step": 23046
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.401478409649973e-06,
+      "loss": 1.0811,
+      "step": 23047
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4006996847193453e-06,
+      "loss": 0.9824,
+      "step": 23048
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.39992103067602e-06,
+      "loss": 0.8438,
+      "step": 23049
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3991424475283543e-06,
+      "loss": 1.1289,
+      "step": 23050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3983639352847264e-06,
+      "loss": 0.7305,
+      "step": 23051
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3975854939534903e-06,
+      "loss": 0.8203,
+      "step": 23052
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.396807123543009e-06,
+      "loss": 0.6641,
+      "step": 23053
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.396028824061639e-06,
+      "loss": 0.8232,
+      "step": 23054
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3952505955177463e-06,
+      "loss": 1.0479,
+      "step": 23055
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3944724379196903e-06,
+      "loss": 0.6514,
+      "step": 23056
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3936943512758292e-06,
+      "loss": 1.1309,
+      "step": 23057
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.392916335594516e-06,
+      "loss": 1.2373,
+      "step": 23058
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.392138390884114e-06,
+      "loss": 0.7568,
+      "step": 23059
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3913605171529763e-06,
+      "loss": 0.8496,
+      "step": 23060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3905827144094574e-06,
+      "loss": 1.1621,
+      "step": 23061
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3898049826619183e-06,
+      "loss": 1.4277,
+      "step": 23062
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.389027321918709e-06,
+      "loss": 1.1772,
+      "step": 23063
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3882497321881823e-06,
+      "loss": 0.9863,
+      "step": 23064
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3874722134786863e-06,
+      "loss": 1.1606,
+      "step": 23065
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3866947657985827e-06,
+      "loss": 0.833,
+      "step": 23066
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.385917389156219e-06,
+      "loss": 0.791,
+      "step": 23067
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.385140083559941e-06,
+      "loss": 0.9736,
+      "step": 23068
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3843628490181044e-06,
+      "loss": 0.6807,
+      "step": 23069
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3835856855390515e-06,
+      "loss": 0.9102,
+      "step": 23070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3828085931311382e-06,
+      "loss": 1.2036,
+      "step": 23071
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.382031571802703e-06,
+      "loss": 1.2705,
+      "step": 23072
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.381254621562101e-06,
+      "loss": 1.0288,
+      "step": 23073
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3804777424176737e-06,
+      "loss": 1.1123,
+      "step": 23074
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.379700934377762e-06,
+      "loss": 0.9121,
+      "step": 23075
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3789241974507147e-06,
+      "loss": 0.8735,
+      "step": 23076
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.378147531644879e-06,
+      "loss": 0.751,
+      "step": 23077
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3773709369685924e-06,
+      "loss": 0.7686,
+      "step": 23078
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3765944134301944e-06,
+      "loss": 1.3525,
+      "step": 23079
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.375817961038033e-06,
+      "loss": 1.418,
+      "step": 23080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.375041579800441e-06,
+      "loss": 0.6875,
+      "step": 23081
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.374265269725766e-06,
+      "loss": 0.8135,
+      "step": 23082
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3734890308223388e-06,
+      "loss": 0.9189,
+      "step": 23083
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.372712863098505e-06,
+      "loss": 1.291,
+      "step": 23084
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.371936766562599e-06,
+      "loss": 0.707,
+      "step": 23085
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.371160741222953e-06,
+      "loss": 1.0625,
+      "step": 23086
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.370384787087907e-06,
+      "loss": 0.9453,
+      "step": 23087
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3696089041657987e-06,
+      "loss": 0.7998,
+      "step": 23088
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3688330924649593e-06,
+      "loss": 1.1953,
+      "step": 23089
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3680573519937187e-06,
+      "loss": 0.8145,
+      "step": 23090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.367281682760414e-06,
+      "loss": 1.0225,
+      "step": 23091
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3665060847733787e-06,
+      "loss": 1.4102,
+      "step": 23092
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3657305580409417e-06,
+      "loss": 0.9727,
+      "step": 23093
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3649551025714312e-06,
+      "loss": 1.0029,
+      "step": 23094
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3641797183731805e-06,
+      "loss": 1.5381,
+      "step": 23095
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3634044054545145e-06,
+      "loss": 0.8164,
+      "step": 23096
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3626291638237673e-06,
+      "loss": 0.603,
+      "step": 23097
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.36185399348926e-06,
+      "loss": 0.4927,
+      "step": 23098
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.361078894459324e-06,
+      "loss": 1.0439,
+      "step": 23099
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.360303866742285e-06,
+      "loss": 0.8931,
+      "step": 23100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3595289103464613e-06,
+      "loss": 0.9648,
+      "step": 23101
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.358754025280182e-06,
+      "loss": 0.6987,
+      "step": 23102
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.357979211551775e-06,
+      "loss": 1.459,
+      "step": 23103
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.357204469169558e-06,
+      "loss": 0.9893,
+      "step": 23104
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3564297981418514e-06,
+      "loss": 0.6077,
+      "step": 23105
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.355655198476978e-06,
+      "loss": 0.8828,
+      "step": 23106
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.354880670183264e-06,
+      "loss": 0.8154,
+      "step": 23107
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.354106213269023e-06,
+      "loss": 0.9668,
+      "step": 23108
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.353331827742572e-06,
+      "loss": 0.7827,
+      "step": 23109
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3525575136122344e-06,
+      "loss": 0.9209,
+      "step": 23110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3517832708863276e-06,
+      "loss": 0.6533,
+      "step": 23111
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3510090995731616e-06,
+      "loss": 0.9912,
+      "step": 23112
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.350234999681057e-06,
+      "loss": 1.4893,
+      "step": 23113
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3494609712183323e-06,
+      "loss": 1.4609,
+      "step": 23114
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.348687014193298e-06,
+      "loss": 0.8394,
+      "step": 23115
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3479131286142653e-06,
+      "loss": 1.2324,
+      "step": 23116
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3471393144895492e-06,
+      "loss": 0.8613,
+      "step": 23117
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3463655718274656e-06,
+      "loss": 1.0723,
+      "step": 23118
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3455919006363225e-06,
+      "loss": 0.6821,
+      "step": 23119
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.344818300924426e-06,
+      "loss": 0.8032,
+      "step": 23120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3440447727000946e-06,
+      "loss": 0.9023,
+      "step": 23121
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.343271315971629e-06,
+      "loss": 1.0254,
+      "step": 23122
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3424979307473436e-06,
+      "loss": 1.1973,
+      "step": 23123
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.341724617035541e-06,
+      "loss": 0.894,
+      "step": 23124
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3409513748445332e-06,
+      "loss": 0.8672,
+      "step": 23125
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3401782041826236e-06,
+      "loss": 0.9717,
+      "step": 23126
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3394051050581123e-06,
+      "loss": 1.1392,
+      "step": 23127
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.338632077479309e-06,
+      "loss": 1.064,
+      "step": 23128
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3378591214545197e-06,
+      "loss": 1.0117,
+      "step": 23129
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3370862369920444e-06,
+      "loss": 0.9438,
+      "step": 23130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3363134241001804e-06,
+      "loss": 1.1074,
+      "step": 23131
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3355406827872338e-06,
+      "loss": 0.5938,
+      "step": 23132
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.334768013061508e-06,
+      "loss": 1.0146,
+      "step": 23133
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.333995414931299e-06,
+      "loss": 0.918,
+      "step": 23134
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3332228884049035e-06,
+      "loss": 0.8994,
+      "step": 23135
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3324504334906247e-06,
+      "loss": 0.8345,
+      "step": 23136
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3316780501967547e-06,
+      "loss": 1.0186,
+      "step": 23137
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3309057385315966e-06,
+      "loss": 0.9551,
+      "step": 23138
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3301334985034394e-06,
+      "loss": 0.8499,
+      "step": 23139
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3293613301205853e-06,
+      "loss": 1.2544,
+      "step": 23140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3285892333913237e-06,
+      "loss": 0.835,
+      "step": 23141
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3278172083239468e-06,
+      "loss": 0.7363,
+      "step": 23142
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3270452549267507e-06,
+      "loss": 0.75,
+      "step": 23143
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3262733732080297e-06,
+      "loss": 0.668,
+      "step": 23144
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3255015631760723e-06,
+      "loss": 1.168,
+      "step": 23145
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3247298248391656e-06,
+      "loss": 0.9316,
+      "step": 23146
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3239581582056057e-06,
+      "loss": 1.3857,
+      "step": 23147
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3231865632836746e-06,
+      "loss": 0.9766,
+      "step": 23148
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.322415040081666e-06,
+      "loss": 1.043,
+      "step": 23149
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.321643588607868e-06,
+      "loss": 1.1157,
+      "step": 23150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3208722088705646e-06,
+      "loss": 0.96,
+      "step": 23151
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.320100900878043e-06,
+      "loss": 1.3027,
+      "step": 23152
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3193296646385796e-06,
+      "loss": 0.8022,
+      "step": 23153
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3185585001604735e-06,
+      "loss": 1.3037,
+      "step": 23154
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.317787407452002e-06,
+      "loss": 0.832,
+      "step": 23155
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3170163865214467e-06,
+      "loss": 0.79,
+      "step": 23156
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.316245437377087e-06,
+      "loss": 0.9688,
+      "step": 23157
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3154745600272066e-06,
+      "loss": 0.9521,
+      "step": 23158
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3147037544800907e-06,
+      "loss": 1.3125,
+      "step": 23159
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3139330207440114e-06,
+      "loss": 1.1719,
+      "step": 23160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3131623588272542e-06,
+      "loss": 1.0615,
+      "step": 23161
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.312391768738095e-06,
+      "loss": 0.8467,
+      "step": 23162
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3116212504848065e-06,
+      "loss": 0.9448,
+      "step": 23163
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3108508040756695e-06,
+      "loss": 1.1323,
+      "step": 23164
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3100804295189626e-06,
+      "loss": 1.2383,
+      "step": 23165
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.309310126822959e-06,
+      "loss": 1.2559,
+      "step": 23166
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.308539895995927e-06,
+      "loss": 1.1562,
+      "step": 23167
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3077697370461503e-06,
+      "loss": 1.1318,
+      "step": 23168
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3069996499818913e-06,
+      "loss": 0.9512,
+      "step": 23169
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.306229634811432e-06,
+      "loss": 0.8105,
+      "step": 23170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.305459691543035e-06,
+      "loss": 1.1777,
+      "step": 23171
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3046898201849774e-06,
+      "loss": 1.0186,
+      "step": 23172
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3039200207455267e-06,
+      "loss": 1.3223,
+      "step": 23173
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.303150293232947e-06,
+      "loss": 0.7295,
+      "step": 23174
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.302380637655511e-06,
+      "loss": 1.25,
+      "step": 23175
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3016110540214896e-06,
+      "loss": 0.7251,
+      "step": 23176
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3008415423391447e-06,
+      "loss": 0.9434,
+      "step": 23177
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3000721026167393e-06,
+      "loss": 0.8057,
+      "step": 23178
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.299302734862543e-06,
+      "loss": 0.8308,
+      "step": 23179
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.298533439084821e-06,
+      "loss": 1.0713,
+      "step": 23180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2977642152918354e-06,
+      "loss": 0.9189,
+      "step": 23181
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2969950634918457e-06,
+      "loss": 0.8994,
+      "step": 23182
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2962259836931197e-06,
+      "loss": 0.9707,
+      "step": 23183
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.295456975903911e-06,
+      "loss": 0.8506,
+      "step": 23184
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.294688040132489e-06,
+      "loss": 0.6006,
+      "step": 23185
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2939191763871047e-06,
+      "loss": 0.7705,
+      "step": 23186
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2931503846760247e-06,
+      "loss": 0.9883,
+      "step": 23187
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2923816650075038e-06,
+      "loss": 1.3281,
+      "step": 23188
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.291613017389794e-06,
+      "loss": 0.7925,
+      "step": 23189
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.290844441831158e-06,
+      "loss": 0.8208,
+      "step": 23190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.290075938339854e-06,
+      "loss": 0.957,
+      "step": 23191
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.289307506924132e-06,
+      "loss": 0.6172,
+      "step": 23192
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2885391475922448e-06,
+      "loss": 0.625,
+      "step": 23193
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.287770860352448e-06,
+      "loss": 1.1888,
+      "step": 23194
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.287002645212999e-06,
+      "loss": 0.9238,
+      "step": 23195
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.286234502182145e-06,
+      "loss": 0.7124,
+      "step": 23196
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.285466431268135e-06,
+      "loss": 1.1406,
+      "step": 23197
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.284698432479224e-06,
+      "loss": 0.6606,
+      "step": 23198
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2839305058236604e-06,
+      "loss": 0.957,
+      "step": 23199
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2831626513096883e-06,
+      "loss": 1.3154,
+      "step": 23200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.28239486894556e-06,
+      "loss": 1.2456,
+      "step": 23201
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2816271587395255e-06,
+      "loss": 1.1768,
+      "step": 23202
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2808595206998274e-06,
+      "loss": 0.897,
+      "step": 23203
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2800919548347086e-06,
+      "loss": 0.874,
+      "step": 23204
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.279324461152418e-06,
+      "loss": 0.9155,
+      "step": 23205
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.278557039661202e-06,
+      "loss": 1.1494,
+      "step": 23206
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.277789690369302e-06,
+      "loss": 0.8535,
+      "step": 23207
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.277022413284956e-06,
+      "loss": 1.1719,
+      "step": 23208
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2762552084164124e-06,
+      "loss": 0.6787,
+      "step": 23209
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2754880757719065e-06,
+      "loss": 1.3779,
+      "step": 23210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.274721015359685e-06,
+      "loss": 1.0088,
+      "step": 23211
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.27395402718798e-06,
+      "loss": 0.8848,
+      "step": 23212
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2731871112650372e-06,
+      "loss": 0.9951,
+      "step": 23213
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2724202675990926e-06,
+      "loss": 1.1865,
+      "step": 23214
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.271653496198377e-06,
+      "loss": 1.0557,
+      "step": 23215
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.270886797071133e-06,
+      "loss": 0.9297,
+      "step": 23216
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.270120170225598e-06,
+      "loss": 1.0479,
+      "step": 23217
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2693536156700047e-06,
+      "loss": 0.9004,
+      "step": 23218
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.268587133412582e-06,
+      "loss": 1.1045,
+      "step": 23219
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.267820723461568e-06,
+      "loss": 1.0527,
+      "step": 23220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2670543858251978e-06,
+      "loss": 1.2969,
+      "step": 23221
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2662881205117003e-06,
+      "loss": 0.8254,
+      "step": 23222
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2655219275293036e-06,
+      "loss": 0.9746,
+      "step": 23223
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.264755806886243e-06,
+      "loss": 0.9175,
+      "step": 23224
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.263989758590742e-06,
+      "loss": 0.8457,
+      "step": 23225
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.263223782651036e-06,
+      "loss": 0.9717,
+      "step": 23226
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.262457879075347e-06,
+      "loss": 0.9746,
+      "step": 23227
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.261692047871907e-06,
+      "loss": 1.1289,
+      "step": 23228
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2609262890489402e-06,
+      "loss": 1.2383,
+      "step": 23229
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2601606026146683e-06,
+      "loss": 1.0039,
+      "step": 23230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.259394988577319e-06,
+      "loss": 1.0469,
+      "step": 23231
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2586294469451197e-06,
+      "loss": 1.332,
+      "step": 23232
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.257863977726291e-06,
+      "loss": 1.125,
+      "step": 23233
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2570985809290525e-06,
+      "loss": 1.1006,
+      "step": 23234
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2563332565616303e-06,
+      "loss": 0.9326,
+      "step": 23235
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.255568004632239e-06,
+      "loss": 0.6606,
+      "step": 23236
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.254802825149107e-06,
+      "loss": 0.877,
+      "step": 23237
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2540377181204463e-06,
+      "loss": 0.8838,
+      "step": 23238
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2532726835544813e-06,
+      "loss": 1.0732,
+      "step": 23239
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.252507721459428e-06,
+      "loss": 0.8921,
+      "step": 23240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2517428318434942e-06,
+      "loss": 1.1719,
+      "step": 23241
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.250978014714912e-06,
+      "loss": 1.1006,
+      "step": 23242
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2502132700818876e-06,
+      "loss": 0.7959,
+      "step": 23243
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.249448597952637e-06,
+      "loss": 0.7373,
+      "step": 23244
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2486839983353714e-06,
+      "loss": 0.9551,
+      "step": 23245
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2479194712383054e-06,
+      "loss": 0.998,
+      "step": 23246
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2471550166696565e-06,
+      "loss": 0.7631,
+      "step": 23247
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2463906346376274e-06,
+      "loss": 0.8984,
+      "step": 23248
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2456263251504373e-06,
+      "loss": 0.8921,
+      "step": 23249
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.244862088216293e-06,
+      "loss": 1.2471,
+      "step": 23250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2440979238433977e-06,
+      "loss": 0.8232,
+      "step": 23251
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2433338320399667e-06,
+      "loss": 1.0146,
+      "step": 23252
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.242569812814208e-06,
+      "loss": 1.0137,
+      "step": 23253
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.241805866174327e-06,
+      "loss": 0.9639,
+      "step": 23254
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2410419921285286e-06,
+      "loss": 0.6816,
+      "step": 23255
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2402781906850112e-06,
+      "loss": 0.7524,
+      "step": 23256
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2395144618519948e-06,
+      "loss": 0.998,
+      "step": 23257
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2387508056376726e-06,
+      "loss": 1.209,
+      "step": 23258
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2379872220502483e-06,
+      "loss": 1.1904,
+      "step": 23259
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.237223711097928e-06,
+      "loss": 0.876,
+      "step": 23260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2364602727889105e-06,
+      "loss": 0.8525,
+      "step": 23261
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2356969071313927e-06,
+      "loss": 0.999,
+      "step": 23262
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2349336141335787e-06,
+      "loss": 0.7979,
+      "step": 23263
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2341703938036707e-06,
+      "loss": 0.7378,
+      "step": 23264
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2334072461498622e-06,
+      "loss": 0.5364,
+      "step": 23265
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.232644171180349e-06,
+      "loss": 0.8594,
+      "step": 23266
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2318811689033304e-06,
+      "loss": 1.1699,
+      "step": 23267
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2311182393270048e-06,
+      "loss": 0.8286,
+      "step": 23268
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2303553824595655e-06,
+      "loss": 1.1455,
+      "step": 23269
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2295925983092026e-06,
+      "loss": 0.9243,
+      "step": 23270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.228829886884116e-06,
+      "loss": 1.0859,
+      "step": 23271
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2280672481924936e-06,
+      "loss": 0.958,
+      "step": 23272
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.227304682242531e-06,
+      "loss": 0.833,
+      "step": 23273
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2265421890424164e-06,
+      "loss": 0.8076,
+      "step": 23274
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2257797686003443e-06,
+      "loss": 0.8799,
+      "step": 23275
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.225017420924501e-06,
+      "loss": 0.9795,
+      "step": 23276
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2242551460230744e-06,
+      "loss": 1.2441,
+      "step": 23277
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2234929439042526e-06,
+      "loss": 0.9053,
+      "step": 23278
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2227308145762283e-06,
+      "loss": 0.7734,
+      "step": 23279
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2219687580471846e-06,
+      "loss": 1.2959,
+      "step": 23280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.221206774325304e-06,
+      "loss": 0.8691,
+      "step": 23281
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2204448634187734e-06,
+      "loss": 0.8682,
+      "step": 23282
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2196830253357825e-06,
+      "loss": 0.7188,
+      "step": 23283
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2189212600845087e-06,
+      "loss": 0.9062,
+      "step": 23284
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2181595676731336e-06,
+      "loss": 1.123,
+      "step": 23285
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.217397948109845e-06,
+      "loss": 0.9375,
+      "step": 23286
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.216636401402817e-06,
+      "loss": 0.895,
+      "step": 23287
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2158749275602363e-06,
+      "loss": 0.5422,
+      "step": 23288
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2151135265902766e-06,
+      "loss": 1.083,
+      "step": 23289
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2143521985011218e-06,
+      "loss": 0.6685,
+      "step": 23290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.213590943300948e-06,
+      "loss": 1.1455,
+      "step": 23291
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2128297609979276e-06,
+      "loss": 0.7603,
+      "step": 23292
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.212068651600242e-06,
+      "loss": 0.676,
+      "step": 23293
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2113076151160683e-06,
+      "loss": 1.0693,
+      "step": 23294
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.21054665155358e-06,
+      "loss": 0.8115,
+      "step": 23295
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.209785760920945e-06,
+      "loss": 0.9883,
+      "step": 23296
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2090249432263455e-06,
+      "loss": 1.0684,
+      "step": 23297
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2082641984779474e-06,
+      "loss": 0.9922,
+      "step": 23298
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2075035266839262e-06,
+      "loss": 0.9531,
+      "step": 23299
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.20674292785245e-06,
+      "loss": 0.8936,
+      "step": 23300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2059824019916917e-06,
+      "loss": 1.0742,
+      "step": 23301
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.20522194910982e-06,
+      "loss": 0.9121,
+      "step": 23302
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2044615692149992e-06,
+      "loss": 1.084,
+      "step": 23303
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.203701262315401e-06,
+      "loss": 1.1152,
+      "step": 23304
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2029410284191933e-06,
+      "loss": 1.0581,
+      "step": 23305
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.202180867534541e-06,
+      "loss": 0.7773,
+      "step": 23306
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.201420779669607e-06,
+      "loss": 1.2061,
+      "step": 23307
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2006607648325572e-06,
+      "loss": 0.7588,
+      "step": 23308
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1999008230315598e-06,
+      "loss": 1.0215,
+      "step": 23309
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1991409542747742e-06,
+      "loss": 1.2891,
+      "step": 23310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1983811585703596e-06,
+      "loss": 0.8818,
+      "step": 23311
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1976214359264834e-06,
+      "loss": 0.9736,
+      "step": 23312
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1968617863512996e-06,
+      "loss": 0.9697,
+      "step": 23313
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.196102209852976e-06,
+      "loss": 0.7769,
+      "step": 23314
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1953427064396637e-06,
+      "loss": 1.1387,
+      "step": 23315
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1945832761195284e-06,
+      "loss": 0.729,
+      "step": 23316
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.193823918900724e-06,
+      "loss": 1.3066,
+      "step": 23317
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.193064634791404e-06,
+      "loss": 1.4922,
+      "step": 23318
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1923054237997285e-06,
+      "loss": 0.7559,
+      "step": 23319
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.191546285933854e-06,
+      "loss": 0.9775,
+      "step": 23320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1907872212019332e-06,
+      "loss": 0.9873,
+      "step": 23321
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1900282296121166e-06,
+      "loss": 0.6772,
+      "step": 23322
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1892693111725625e-06,
+      "loss": 1.2402,
+      "step": 23323
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.188510465891417e-06,
+      "loss": 0.96,
+      "step": 23324
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.187751693776837e-06,
+      "loss": 0.9756,
+      "step": 23325
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.186992994836967e-06,
+      "loss": 1.2715,
+      "step": 23326
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1862343690799634e-06,
+      "loss": 0.7773,
+      "step": 23327
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.185475816513973e-06,
+      "loss": 0.9233,
+      "step": 23328
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.184717337147135e-06,
+      "loss": 1.0127,
+      "step": 23329
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.183958930987612e-06,
+      "loss": 1.1973,
+      "step": 23330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.183200598043541e-06,
+      "loss": 0.9932,
+      "step": 23331
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1824423383230705e-06,
+      "loss": 1.458,
+      "step": 23332
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.181684151834342e-06,
+      "loss": 1.3037,
+      "step": 23333
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1809260385855023e-06,
+      "loss": 0.9648,
+      "step": 23334
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.180167998584697e-06,
+      "loss": 0.8193,
+      "step": 23335
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1794100318400666e-06,
+      "loss": 0.8853,
+      "step": 23336
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.17865213835975e-06,
+      "loss": 0.999,
+      "step": 23337
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.177894318151894e-06,
+      "loss": 0.8916,
+      "step": 23338
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1771365712246317e-06,
+      "loss": 0.7627,
+      "step": 23339
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1763788975861076e-06,
+      "loss": 0.8574,
+      "step": 23340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1756212972444624e-06,
+      "loss": 1.0322,
+      "step": 23341
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.17486377020783e-06,
+      "loss": 1.2168,
+      "step": 23342
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.17410631648435e-06,
+      "loss": 1.1426,
+      "step": 23343
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.173348936082149e-06,
+      "loss": 1.041,
+      "step": 23344
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1725916290093783e-06,
+      "loss": 0.7642,
+      "step": 23345
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.171834395274165e-06,
+      "loss": 0.7031,
+      "step": 23346
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1710772348846385e-06,
+      "loss": 0.7485,
+      "step": 23347
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1703201478489408e-06,
+      "loss": 0.9004,
+      "step": 23348
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1695631341751965e-06,
+      "loss": 0.7109,
+      "step": 23349
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.168806193871543e-06,
+      "loss": 0.9448,
+      "step": 23350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1680493269461066e-06,
+      "loss": 0.9844,
+      "step": 23351
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1672925334070214e-06,
+      "loss": 1.0146,
+      "step": 23352
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1665358132624158e-06,
+      "loss": 0.8218,
+      "step": 23353
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.165779166520415e-06,
+      "loss": 1.0117,
+      "step": 23354
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.165022593189143e-06,
+      "loss": 0.8457,
+      "step": 23355
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.16426609327674e-06,
+      "loss": 0.9849,
+      "step": 23356
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1635096667913223e-06,
+      "loss": 0.7812,
+      "step": 23357
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.162753313741015e-06,
+      "loss": 0.9434,
+      "step": 23358
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1619970341339477e-06,
+      "loss": 1.1904,
+      "step": 23359
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.161240827978237e-06,
+      "loss": 1.3184,
+      "step": 23360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1604846952820147e-06,
+      "loss": 0.6172,
+      "step": 23361
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.159728636053393e-06,
+      "loss": 1.0137,
+      "step": 23362
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1589726503005026e-06,
+      "loss": 0.9346,
+      "step": 23363
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1582167380314597e-06,
+      "loss": 1.1543,
+      "step": 23364
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1574608992543796e-06,
+      "loss": 1.0234,
+      "step": 23365
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1567051339773867e-06,
+      "loss": 0.833,
+      "step": 23366
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1559494422086023e-06,
+      "loss": 0.8828,
+      "step": 23367
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1551938239561384e-06,
+      "loss": 1.043,
+      "step": 23368
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1544382792281092e-06,
+      "loss": 1.1846,
+      "step": 23369
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1536828080326344e-06,
+      "loss": 0.7168,
+      "step": 23370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.152927410377833e-06,
+      "loss": 1.1816,
+      "step": 23371
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.152172086271813e-06,
+      "loss": 1.1895,
+      "step": 23372
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1514168357226883e-06,
+      "loss": 1.001,
+      "step": 23373
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1506616587385753e-06,
+      "loss": 0.9287,
+      "step": 23374
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1499065553275797e-06,
+      "loss": 0.9722,
+      "step": 23375
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.14915152549782e-06,
+      "loss": 0.9756,
+      "step": 23376
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1483965692573993e-06,
+      "loss": 0.7402,
+      "step": 23377
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1476416866144343e-06,
+      "loss": 0.7988,
+      "step": 23378
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1468868775770302e-06,
+      "loss": 1.001,
+      "step": 23379
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1461321421532907e-06,
+      "loss": 1.2949,
+      "step": 23380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.145377480351327e-06,
+      "loss": 1.0801,
+      "step": 23381
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.144622892179248e-06,
+      "loss": 0.9614,
+      "step": 23382
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.143868377645157e-06,
+      "loss": 1.0605,
+      "step": 23383
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.143113936757154e-06,
+      "loss": 0.8623,
+      "step": 23384
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.142359569523351e-06,
+      "loss": 1.3164,
+      "step": 23385
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1416052759518447e-06,
+      "loss": 0.6777,
+      "step": 23386
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.140851056050742e-06,
+      "loss": 1.0898,
+      "step": 23387
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.14009690982814e-06,
+      "loss": 1.1299,
+      "step": 23388
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1393428372921442e-06,
+      "loss": 0.9004,
+      "step": 23389
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1385888384508535e-06,
+      "loss": 1.0811,
+      "step": 23390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.137834913312362e-06,
+      "loss": 0.7285,
+      "step": 23391
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1370810618847715e-06,
+      "loss": 0.7651,
+      "step": 23392
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.136327284176184e-06,
+      "loss": 1.1758,
+      "step": 23393
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1355735801946918e-06,
+      "loss": 1.291,
+      "step": 23394
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.134819949948388e-06,
+      "loss": 0.71,
+      "step": 23395
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1340663934453718e-06,
+      "loss": 0.8628,
+      "step": 23396
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13331291069374e-06,
+      "loss": 0.9639,
+      "step": 23397
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.132559501701584e-06,
+      "loss": 1.127,
+      "step": 23398
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1318061664769926e-06,
+      "loss": 1.002,
+      "step": 23399
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1310529050280636e-06,
+      "loss": 0.9766,
+      "step": 23400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1302997173628834e-06,
+      "loss": 0.8164,
+      "step": 23401
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1295466034895482e-06,
+      "loss": 0.9521,
+      "step": 23402
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1287935634161414e-06,
+      "loss": 1.0762,
+      "step": 23403
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1280405971507576e-06,
+      "loss": 1.0801,
+      "step": 23404
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1272877047014826e-06,
+      "loss": 1.208,
+      "step": 23405
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1265348860764e-06,
+      "loss": 1.1772,
+      "step": 23406
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1257821412836e-06,
+      "loss": 0.8955,
+      "step": 23407
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1250294703311714e-06,
+      "loss": 1.5508,
+      "step": 23408
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1242768732271956e-06,
+      "loss": 0.9795,
+      "step": 23409
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.123524349979753e-06,
+      "loss": 0.999,
+      "step": 23410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1227719005969305e-06,
+      "loss": 1.0,
+      "step": 23411
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1220195250868144e-06,
+      "loss": 1.0732,
+      "step": 23412
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1212672234574836e-06,
+      "loss": 1.3594,
+      "step": 23413
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1205149957170146e-06,
+      "loss": 0.9561,
+      "step": 23414
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.119762841873495e-06,
+      "loss": 0.793,
+      "step": 23415
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1190107619349995e-06,
+      "loss": 1.0254,
+      "step": 23416
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1182587559096057e-06,
+      "loss": 0.8516,
+      "step": 23417
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.117506823805393e-06,
+      "loss": 0.9404,
+      "step": 23418
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1167549656304407e-06,
+      "loss": 1.4326,
+      "step": 23419
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.116003181392825e-06,
+      "loss": 0.8268,
+      "step": 23420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1152514711006143e-06,
+      "loss": 0.6765,
+      "step": 23421
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1144998347618892e-06,
+      "loss": 1.0322,
+      "step": 23422
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1137482723847255e-06,
+      "loss": 0.9463,
+      "step": 23423
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1129967839771935e-06,
+      "loss": 1.4648,
+      "step": 23424
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.112245369547362e-06,
+      "loss": 1.2461,
+      "step": 23425
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1114940291033092e-06,
+      "loss": 1.0938,
+      "step": 23426
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.110742762653097e-06,
+      "loss": 1.084,
+      "step": 23427
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1099915702048058e-06,
+      "loss": 0.9414,
+      "step": 23428
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.109240451766494e-06,
+      "loss": 1.1299,
+      "step": 23429
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.10848940734624e-06,
+      "loss": 1.0547,
+      "step": 23430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1077384369521057e-06,
+      "loss": 0.8262,
+      "step": 23431
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.106987540592151e-06,
+      "loss": 1.1318,
+      "step": 23432
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.106236718274457e-06,
+      "loss": 1.2051,
+      "step": 23433
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1054859700070794e-06,
+      "loss": 0.9453,
+      "step": 23434
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.104735295798085e-06,
+      "loss": 0.9541,
+      "step": 23435
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.103984695655533e-06,
+      "loss": 1.0234,
+      "step": 23436
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.103234169587488e-06,
+      "loss": 0.6001,
+      "step": 23437
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1024837176020173e-06,
+      "loss": 1.0928,
+      "step": 23438
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.101733339707175e-06,
+      "loss": 0.8916,
+      "step": 23439
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.100983035911027e-06,
+      "loss": 0.7041,
+      "step": 23440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1002328062216293e-06,
+      "loss": 0.9189,
+      "step": 23441
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.099482650647039e-06,
+      "loss": 0.811,
+      "step": 23442
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0987325691953164e-06,
+      "loss": 1.002,
+      "step": 23443
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.097982561874521e-06,
+      "loss": 1.3701,
+      "step": 23444
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.097232628692707e-06,
+      "loss": 0.8823,
+      "step": 23445
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0964827696579256e-06,
+      "loss": 0.9248,
+      "step": 23446
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.095732984778239e-06,
+      "loss": 0.8389,
+      "step": 23447
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.094983274061696e-06,
+      "loss": 1.0498,
+      "step": 23448
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.094233637516353e-06,
+      "loss": 0.6074,
+      "step": 23449
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.093484075150257e-06,
+      "loss": 1.1152,
+      "step": 23450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0927345869714663e-06,
+      "loss": 0.7646,
+      "step": 23451
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0919851729880303e-06,
+      "loss": 0.957,
+      "step": 23452
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0912358332079915e-06,
+      "loss": 0.8662,
+      "step": 23453
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0904865676394057e-06,
+      "loss": 0.9365,
+      "step": 23454
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.089737376290324e-06,
+      "loss": 1.1816,
+      "step": 23455
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0889882591687894e-06,
+      "loss": 1.3135,
+      "step": 23456
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0882392162828466e-06,
+      "loss": 1.1318,
+      "step": 23457
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0874902476405445e-06,
+      "loss": 1.1895,
+      "step": 23458
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.08674135324993e-06,
+      "loss": 1.2373,
+      "step": 23459
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0859925331190468e-06,
+      "loss": 0.5957,
+      "step": 23460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.085243787255934e-06,
+      "loss": 0.6696,
+      "step": 23461
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.08449511566864e-06,
+      "loss": 0.8428,
+      "step": 23462
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0837465183652014e-06,
+      "loss": 1.0635,
+      "step": 23463
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.082997995353665e-06,
+      "loss": 0.9092,
+      "step": 23464
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0822495466420656e-06,
+      "loss": 0.7949,
+      "step": 23465
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.081501172238449e-06,
+      "loss": 0.7559,
+      "step": 23466
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0807528721508506e-06,
+      "loss": 0.9736,
+      "step": 23467
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0800046463873044e-06,
+      "loss": 1.1162,
+      "step": 23468
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0792564949558524e-06,
+      "loss": 1.0332,
+      "step": 23469
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.078508417864533e-06,
+      "loss": 1.0127,
+      "step": 23470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.077760415121379e-06,
+      "loss": 0.8018,
+      "step": 23471
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.077012486734421e-06,
+      "loss": 0.7539,
+      "step": 23472
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0762646327116975e-06,
+      "loss": 0.9443,
+      "step": 23473
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0755168530612444e-06,
+      "loss": 1.0088,
+      "step": 23474
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0747691477910913e-06,
+      "loss": 1.0752,
+      "step": 23475
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0740215169092656e-06,
+      "loss": 0.7827,
+      "step": 23476
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0732739604238047e-06,
+      "loss": 0.9385,
+      "step": 23477
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0725264783427366e-06,
+      "loss": 1.0879,
+      "step": 23478
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.071779070674086e-06,
+      "loss": 1.3574,
+      "step": 23479
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0710317374258845e-06,
+      "loss": 0.8584,
+      "step": 23480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0702844786061638e-06,
+      "loss": 1.1572,
+      "step": 23481
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0695372942229462e-06,
+      "loss": 1.043,
+      "step": 23482
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.068790184284256e-06,
+      "loss": 0.8887,
+      "step": 23483
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.06804314879812e-06,
+      "loss": 1.0645,
+      "step": 23484
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.067296187772567e-06,
+      "loss": 1.2695,
+      "step": 23485
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.066549301215618e-06,
+      "loss": 0.7021,
+      "step": 23486
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.06580248913529e-06,
+      "loss": 1.2134,
+      "step": 23487
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0650557515396127e-06,
+      "loss": 0.9004,
+      "step": 23488
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0643090884366023e-06,
+      "loss": 1.1572,
+      "step": 23489
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.063562499834284e-06,
+      "loss": 0.7939,
+      "step": 23490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0628159857406714e-06,
+      "loss": 1.2383,
+      "step": 23491
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0620695461637905e-06,
+      "loss": 1.1416,
+      "step": 23492
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.061323181111654e-06,
+      "loss": 1.0332,
+      "step": 23493
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0605768905922783e-06,
+      "loss": 0.7233,
+      "step": 23494
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.059830674613681e-06,
+      "loss": 1.4883,
+      "step": 23495
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0590845331838813e-06,
+      "loss": 0.8672,
+      "step": 23496
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.058338466310893e-06,
+      "loss": 1.1919,
+      "step": 23497
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.057592474002723e-06,
+      "loss": 0.8154,
+      "step": 23498
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.056846556267391e-06,
+      "loss": 0.8945,
+      "step": 23499
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0561007131129105e-06,
+      "loss": 1.0107,
+      "step": 23500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0553549445472917e-06,
+      "loss": 0.6484,
+      "step": 23501
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0546092505785407e-06,
+      "loss": 0.6685,
+      "step": 23502
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.053863631214674e-06,
+      "loss": 0.4502,
+      "step": 23503
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.053118086463699e-06,
+      "loss": 0.9502,
+      "step": 23504
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0523726163336187e-06,
+      "loss": 1.04,
+      "step": 23505
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.051627220832446e-06,
+      "loss": 1.0754,
+      "step": 23506
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0508818999681887e-06,
+      "loss": 0.8135,
+      "step": 23507
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0501366537488518e-06,
+      "loss": 1.5625,
+      "step": 23508
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0493914821824367e-06,
+      "loss": 1.1118,
+      "step": 23509
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.048646385276949e-06,
+      "loss": 1.1855,
+      "step": 23510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0479013630403976e-06,
+      "loss": 1.1465,
+      "step": 23511
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0471564154807808e-06,
+      "loss": 1.3789,
+      "step": 23512
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.046411542606099e-06,
+      "loss": 1.3984,
+      "step": 23513
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.045666744424358e-06,
+      "loss": 0.8618,
+      "step": 23514
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0449220209435525e-06,
+      "loss": 0.793,
+      "step": 23515
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.044177372171688e-06,
+      "loss": 0.7625,
+      "step": 23516
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0434327981167577e-06,
+      "loss": 1.0781,
+      "step": 23517
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0426882987867655e-06,
+      "loss": 1.0879,
+      "step": 23518
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0419438741897055e-06,
+      "loss": 0.9561,
+      "step": 23519
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.041199524333568e-06,
+      "loss": 1.2041,
+      "step": 23520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0404552492263595e-06,
+      "loss": 0.9316,
+      "step": 23521
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0397110488760707e-06,
+      "loss": 0.9985,
+      "step": 23522
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0389669232906937e-06,
+      "loss": 0.9482,
+      "step": 23523
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.038222872478219e-06,
+      "loss": 0.769,
+      "step": 23524
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.037478896446644e-06,
+      "loss": 0.9844,
+      "step": 23525
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0367349952039605e-06,
+      "loss": 1.1914,
+      "step": 23526
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0359911687581567e-06,
+      "loss": 0.9741,
+      "step": 23527
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.035247417117221e-06,
+      "loss": 0.7451,
+      "step": 23528
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0345037402891473e-06,
+      "loss": 1.1572,
+      "step": 23529
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.033760138281918e-06,
+      "loss": 1.1631,
+      "step": 23530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.033016611103524e-06,
+      "loss": 0.9419,
+      "step": 23531
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.032273158761955e-06,
+      "loss": 0.8662,
+      "step": 23532
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0315297812651946e-06,
+      "loss": 1.0684,
+      "step": 23533
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.030786478621226e-06,
+      "loss": 0.916,
+      "step": 23534
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0300432508380304e-06,
+      "loss": 0.7803,
+      "step": 23535
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0293000979235964e-06,
+      "loss": 0.7134,
+      "step": 23536
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0285570198859094e-06,
+      "loss": 0.8906,
+      "step": 23537
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.027814016732944e-06,
+      "loss": 0.917,
+      "step": 23538
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.027071088472687e-06,
+      "loss": 0.9775,
+      "step": 23539
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.026328235113117e-06,
+      "loss": 1.1128,
+      "step": 23540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0255854566622113e-06,
+      "loss": 0.7471,
+      "step": 23541
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.024842753127948e-06,
+      "loss": 0.9014,
+      "step": 23542
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.024100124518312e-06,
+      "loss": 1.0859,
+      "step": 23543
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.023357570841274e-06,
+      "loss": 0.7319,
+      "step": 23544
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.022615092104809e-06,
+      "loss": 0.9971,
+      "step": 23545
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.021872688316896e-06,
+      "loss": 1.1602,
+      "step": 23546
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.021130359485511e-06,
+      "loss": 0.8291,
+      "step": 23547
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.020388105618627e-06,
+      "loss": 1.1416,
+      "step": 23548
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0196459267242117e-06,
+      "loss": 0.6641,
+      "step": 23549
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.018903822810244e-06,
+      "loss": 1.0537,
+      "step": 23550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0181617938846895e-06,
+      "loss": 1.3398,
+      "step": 23551
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0174198399555253e-06,
+      "loss": 1.3926,
+      "step": 23552
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.016677961030715e-06,
+      "loss": 1.0498,
+      "step": 23553
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0159361571182335e-06,
+      "loss": 1.4639,
+      "step": 23554
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.015194428226046e-06,
+      "loss": 0.9033,
+      "step": 23555
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0144527743621156e-06,
+      "loss": 0.8369,
+      "step": 23556
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0137111955344144e-06,
+      "loss": 0.7529,
+      "step": 23557
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0129696917509087e-06,
+      "loss": 0.7012,
+      "step": 23558
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.012228263019562e-06,
+      "loss": 0.9707,
+      "step": 23559
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0114869093483357e-06,
+      "loss": 0.6123,
+      "step": 23560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.010745630745194e-06,
+      "loss": 0.8301,
+      "step": 23561
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.010004427218106e-06,
+      "loss": 0.9082,
+      "step": 23562
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.009263298775027e-06,
+      "loss": 0.9336,
+      "step": 23563
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0085222454239173e-06,
+      "loss": 0.9922,
+      "step": 23564
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0077812671727415e-06,
+      "loss": 0.6306,
+      "step": 23565
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0070403640294565e-06,
+      "loss": 1.2354,
+      "step": 23566
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0062995360020176e-06,
+      "loss": 0.8228,
+      "step": 23567
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0055587830983856e-06,
+      "loss": 1.168,
+      "step": 23568
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.00481810532652e-06,
+      "loss": 1.042,
+      "step": 23569
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0040775026943757e-06,
+      "loss": 0.7153,
+      "step": 23570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.003336975209903e-06,
+      "loss": 0.9248,
+      "step": 23571
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0025965228810593e-06,
+      "loss": 0.7378,
+      "step": 23572
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0018561457158035e-06,
+      "loss": 1.1426,
+      "step": 23573
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0011158437220832e-06,
+      "loss": 0.6694,
+      "step": 23574
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0003756169078492e-06,
+      "loss": 0.9238,
+      "step": 23575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.999635465281057e-06,
+      "loss": 0.9514,
+      "step": 23576
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9988953888496507e-06,
+      "loss": 0.792,
+      "step": 23577
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9981553876215883e-06,
+      "loss": 0.7319,
+      "step": 23578
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9974154616048124e-06,
+      "loss": 0.7251,
+      "step": 23579
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9966756108072748e-06,
+      "loss": 1.082,
+      "step": 23580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9959358352369207e-06,
+      "loss": 1.0127,
+      "step": 23581
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9951961349016944e-06,
+      "loss": 1.1436,
+      "step": 23582
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9944565098095437e-06,
+      "loss": 1.209,
+      "step": 23583
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9937169599684158e-06,
+      "loss": 0.9238,
+      "step": 23584
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9929774853862525e-06,
+      "loss": 0.9175,
+      "step": 23585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9922380860709943e-06,
+      "loss": 0.9854,
+      "step": 23586
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.991498762030587e-06,
+      "loss": 0.9619,
+      "step": 23587
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9907595132729727e-06,
+      "loss": 1.1895,
+      "step": 23588
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9900203398060924e-06,
+      "loss": 1.085,
+      "step": 23589
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9892812416378793e-06,
+      "loss": 0.9902,
+      "step": 23590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9885422187762823e-06,
+      "loss": 0.834,
+      "step": 23591
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9878032712292305e-06,
+      "loss": 0.7241,
+      "step": 23592
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.98706439900467e-06,
+      "loss": 0.7588,
+      "step": 23593
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.98632560211053e-06,
+      "loss": 0.8496,
+      "step": 23594
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.985586880554753e-06,
+      "loss": 0.9209,
+      "step": 23595
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.984848234345271e-06,
+      "loss": 1.1089,
+      "step": 23596
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9841096634900144e-06,
+      "loss": 1.0166,
+      "step": 23597
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.983371167996921e-06,
+      "loss": 0.9678,
+      "step": 23598
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9826327478739257e-06,
+      "loss": 0.9287,
+      "step": 23599
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9818944031289586e-06,
+      "loss": 1.2344,
+      "step": 23600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9811561337699457e-06,
+      "loss": 1.3125,
+      "step": 23601
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.980417939804824e-06,
+      "loss": 0.9092,
+      "step": 23602
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9796798212415166e-06,
+      "loss": 1.043,
+      "step": 23603
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9789417780879594e-06,
+      "loss": 0.8203,
+      "step": 23604
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.978203810352073e-06,
+      "loss": 1.2705,
+      "step": 23605
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.977465918041791e-06,
+      "loss": 0.9146,
+      "step": 23606
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9767281011650363e-06,
+      "loss": 0.7437,
+      "step": 23607
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.97599035972973e-06,
+      "loss": 0.9912,
+      "step": 23608
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9752526937438033e-06,
+      "loss": 1.1553,
+      "step": 23609
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9745151032151785e-06,
+      "loss": 0.9924,
+      "step": 23610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9737775881517784e-06,
+      "loss": 0.8428,
+      "step": 23611
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.973040148561521e-06,
+      "loss": 0.8237,
+      "step": 23612
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9723027844523324e-06,
+      "loss": 1.4844,
+      "step": 23613
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9715654958321327e-06,
+      "loss": 0.6157,
+      "step": 23614
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.970828282708842e-06,
+      "loss": 0.676,
+      "step": 23615
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9700911450903735e-06,
+      "loss": 1.3984,
+      "step": 23616
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9693540829846533e-06,
+      "loss": 1.4307,
+      "step": 23617
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.968617096399592e-06,
+      "loss": 0.7622,
+      "step": 23618
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9678801853431085e-06,
+      "loss": 0.8936,
+      "step": 23619
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9671433498231216e-06,
+      "loss": 1.2676,
+      "step": 23620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9664065898475435e-06,
+      "loss": 1.0693,
+      "step": 23621
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9656699054242875e-06,
+      "loss": 1.0146,
+      "step": 23622
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.964933296561262e-06,
+      "loss": 1.084,
+      "step": 23623
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9641967632663903e-06,
+      "loss": 0.6733,
+      "step": 23624
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9634603055475784e-06,
+      "loss": 1.0752,
+      "step": 23625
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.962723923412738e-06,
+      "loss": 0.8965,
+      "step": 23626
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.961987616869774e-06,
+      "loss": 0.8047,
+      "step": 23627
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9612513859266033e-06,
+      "loss": 1.0156,
+      "step": 23628
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9605152305911268e-06,
+      "loss": 0.6973,
+      "step": 23629
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.959779150871256e-06,
+      "loss": 1.1846,
+      "step": 23630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9590431467749013e-06,
+      "loss": 1.1475,
+      "step": 23631
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9583072183099635e-06,
+      "loss": 0.7461,
+      "step": 23632
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9575713654843496e-06,
+      "loss": 0.6577,
+      "step": 23633
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9568355883059573e-06,
+      "loss": 0.9189,
+      "step": 23634
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.956099886782702e-06,
+      "loss": 0.7222,
+      "step": 23635
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9553642609224787e-06,
+      "loss": 1.0762,
+      "step": 23636
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9546287107331894e-06,
+      "loss": 0.7754,
+      "step": 23637
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.953893236222739e-06,
+      "loss": 0.6814,
+      "step": 23638
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.953157837399022e-06,
+      "loss": 0.8071,
+      "step": 23639
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9524225142699435e-06,
+      "loss": 0.7095,
+      "step": 23640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9516872668433973e-06,
+      "loss": 1.0908,
+      "step": 23641
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9509520951272864e-06,
+      "loss": 1.0674,
+      "step": 23642
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.950216999129504e-06,
+      "loss": 1.4785,
+      "step": 23643
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9494819788579444e-06,
+      "loss": 0.8267,
+      "step": 23644
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.948747034320505e-06,
+      "loss": 1.0586,
+      "step": 23645
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.948012165525085e-06,
+      "loss": 0.8066,
+      "step": 23646
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9472773724795733e-06,
+      "loss": 1.1104,
+      "step": 23647
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9465426551918596e-06,
+      "loss": 1.1309,
+      "step": 23648
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.945808013669841e-06,
+      "loss": 0.7852,
+      "step": 23649
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9450734479214095e-06,
+      "loss": 0.7573,
+      "step": 23650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9443389579544546e-06,
+      "loss": 0.9209,
+      "step": 23651
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.943604543776861e-06,
+      "loss": 0.7969,
+      "step": 23652
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.942870205396524e-06,
+      "loss": 0.918,
+      "step": 23653
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.942135942821327e-06,
+      "loss": 0.5581,
+      "step": 23654
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9414017560591622e-06,
+      "loss": 1.2812,
+      "step": 23655
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9406676451179094e-06,
+      "loss": 0.9126,
+      "step": 23656
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.939933610005461e-06,
+      "loss": 1.5498,
+      "step": 23657
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.939199650729698e-06,
+      "loss": 0.8828,
+      "step": 23658
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.938465767298502e-06,
+      "loss": 0.9951,
+      "step": 23659
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9377319597197585e-06,
+      "loss": 1.125,
+      "step": 23660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.936998228001353e-06,
+      "loss": 1.1016,
+      "step": 23661
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9362645721511653e-06,
+      "loss": 0.8281,
+      "step": 23662
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9355309921770713e-06,
+      "loss": 0.8281,
+      "step": 23663
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9347974880869577e-06,
+      "loss": 1.0742,
+      "step": 23664
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.934064059888697e-06,
+      "loss": 1.0493,
+      "step": 23665
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.933330707590174e-06,
+      "loss": 0.9209,
+      "step": 23666
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9325974311992588e-06,
+      "loss": 1.2998,
+      "step": 23667
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9318642307238365e-06,
+      "loss": 0.9458,
+      "step": 23668
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.931131106171777e-06,
+      "loss": 1.0273,
+      "step": 23669
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.930398057550955e-06,
+      "loss": 1.3994,
+      "step": 23670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9296650848692456e-06,
+      "loss": 1.0859,
+      "step": 23671
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9289321881345257e-06,
+      "loss": 0.9727,
+      "step": 23672
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.928199367354665e-06,
+      "loss": 0.7534,
+      "step": 23673
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9274666225375324e-06,
+      "loss": 1.1475,
+      "step": 23674
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.926733953691001e-06,
+      "loss": 0.7832,
+      "step": 23675
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9260013608229453e-06,
+      "loss": 0.8394,
+      "step": 23676
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.92526884394123e-06,
+      "loss": 1.1221,
+      "step": 23677
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9245364030537202e-06,
+      "loss": 0.8252,
+      "step": 23678
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9238040381682918e-06,
+      "loss": 0.9346,
+      "step": 23679
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.923071749292804e-06,
+      "loss": 0.875,
+      "step": 23680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.922339536435129e-06,
+      "loss": 1.0674,
+      "step": 23681
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.921607399603126e-06,
+      "loss": 0.9541,
+      "step": 23682
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.920875338804666e-06,
+      "loss": 1.0771,
+      "step": 23683
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9201433540476076e-06,
+      "loss": 1.1797,
+      "step": 23684
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9194114453398125e-06,
+      "loss": 0.8857,
+      "step": 23685
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.918679612689145e-06,
+      "loss": 0.7683,
+      "step": 23686
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.917947856103469e-06,
+      "loss": 1.2197,
+      "step": 23687
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9172161755906425e-06,
+      "loss": 0.8232,
+      "step": 23688
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.91648457115852e-06,
+      "loss": 1.1387,
+      "step": 23689
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9157530428149683e-06,
+      "loss": 0.9941,
+      "step": 23690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.915021590567838e-06,
+      "loss": 1.0713,
+      "step": 23691
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.914290214424993e-06,
+      "loss": 1.0015,
+      "step": 23692
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.913558914394282e-06,
+      "loss": 0.7637,
+      "step": 23693
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9128276904835683e-06,
+      "loss": 1.1387,
+      "step": 23694
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.912096542700702e-06,
+      "loss": 1.2559,
+      "step": 23695
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9113654710535346e-06,
+      "loss": 0.8984,
+      "step": 23696
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9106344755499217e-06,
+      "loss": 1.3086,
+      "step": 23697
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9099035561977175e-06,
+      "loss": 1.1694,
+      "step": 23698
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.909172713004772e-06,
+      "loss": 1.0322,
+      "step": 23699
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9084419459789316e-06,
+      "loss": 1.1816,
+      "step": 23700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9077112551280485e-06,
+      "loss": 0.8154,
+      "step": 23701
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9069806404599753e-06,
+      "loss": 1.0918,
+      "step": 23702
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.906250101982557e-06,
+      "loss": 1.29,
+      "step": 23703
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.905519639703639e-06,
+      "loss": 1.334,
+      "step": 23704
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.904789253631071e-06,
+      "loss": 1.0752,
+      "step": 23705
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9040589437726928e-06,
+      "loss": 0.9551,
+      "step": 23706
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9033287101363583e-06,
+      "loss": 0.8108,
+      "step": 23707
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9025985527299016e-06,
+      "loss": 1.0488,
+      "step": 23708
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9018684715611744e-06,
+      "loss": 1.1445,
+      "step": 23709
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.901138466638015e-06,
+      "loss": 1.2314,
+      "step": 23710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.900408537968258e-06,
+      "loss": 0.9302,
+      "step": 23711
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.899678685559757e-06,
+      "loss": 1.1514,
+      "step": 23712
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.898948909420346e-06,
+      "loss": 1.0391,
+      "step": 23713
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.898219209557864e-06,
+      "loss": 0.8125,
+      "step": 23714
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8974895859801455e-06,
+      "loss": 0.8779,
+      "step": 23715
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8967600386950312e-06,
+      "loss": 0.7529,
+      "step": 23716
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8960305677103616e-06,
+      "loss": 1.0205,
+      "step": 23717
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8953011730339643e-06,
+      "loss": 1.0684,
+      "step": 23718
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.894571854673682e-06,
+      "loss": 0.7695,
+      "step": 23719
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8938426126373454e-06,
+      "loss": 0.8447,
+      "step": 23720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.893113446932787e-06,
+      "loss": 0.9956,
+      "step": 23721
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8923843575678335e-06,
+      "loss": 0.9014,
+      "step": 23722
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8916553445503283e-06,
+      "loss": 1.2676,
+      "step": 23723
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8909264078880983e-06,
+      "loss": 1.1729,
+      "step": 23724
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.89019754758897e-06,
+      "loss": 0.7925,
+      "step": 23725
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.889468763660772e-06,
+      "loss": 1.2715,
+      "step": 23726
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8887400561113353e-06,
+      "loss": 0.8374,
+      "step": 23727
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.88801142494849e-06,
+      "loss": 1.0059,
+      "step": 23728
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.887282870180057e-06,
+      "loss": 1.2705,
+      "step": 23729
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8865543918138685e-06,
+      "loss": 1.0117,
+      "step": 23730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8858259898577455e-06,
+      "loss": 1.0586,
+      "step": 23731
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8850976643195094e-06,
+      "loss": 1.1294,
+      "step": 23732
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8843694152069877e-06,
+      "loss": 0.9229,
+      "step": 23733
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8836412425280057e-06,
+      "loss": 0.9404,
+      "step": 23734
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8829131462903815e-06,
+      "loss": 1.2637,
+      "step": 23735
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.882185126501934e-06,
+      "loss": 0.958,
+      "step": 23736
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.881457183170484e-06,
+      "loss": 1.1406,
+      "step": 23737
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8807293163038572e-06,
+      "loss": 1.1709,
+      "step": 23738
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.880001525909867e-06,
+      "loss": 0.9189,
+      "step": 23739
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.879273811996328e-06,
+      "loss": 0.9219,
+      "step": 23740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.878546174571064e-06,
+      "loss": 0.8408,
+      "step": 23741
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.877818613641884e-06,
+      "loss": 0.5225,
+      "step": 23742
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8770911292166095e-06,
+      "loss": 1.0674,
+      "step": 23743
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8763637213030494e-06,
+      "loss": 1.2148,
+      "step": 23744
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8756363899090233e-06,
+      "loss": 0.8965,
+      "step": 23745
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8749091350423406e-06,
+      "loss": 1.2012,
+      "step": 23746
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.87418195671081e-06,
+      "loss": 0.9238,
+      "step": 23747
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8734548549222453e-06,
+      "loss": 0.75,
+      "step": 23748
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8727278296844607e-06,
+      "loss": 1.1924,
+      "step": 23749
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8720008810052625e-06,
+      "loss": 0.8638,
+      "step": 23750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.871274008892455e-06,
+      "loss": 0.9893,
+      "step": 23751
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.870547213353854e-06,
+      "loss": 1.1252,
+      "step": 23752
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8698204943972585e-06,
+      "loss": 1.0063,
+      "step": 23753
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.869093852030482e-06,
+      "loss": 1.2803,
+      "step": 23754
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.868367286261324e-06,
+      "loss": 1.0923,
+      "step": 23755
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.867640797097594e-06,
+      "loss": 1.022,
+      "step": 23756
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8669143845470936e-06,
+      "loss": 0.957,
+      "step": 23757
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8661880486176217e-06,
+      "loss": 1.1152,
+      "step": 23758
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8654617893169836e-06,
+      "loss": 1.0117,
+      "step": 23759
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.864735606652984e-06,
+      "loss": 1.1133,
+      "step": 23760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.86400950063342e-06,
+      "loss": 1.3906,
+      "step": 23761
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8632834712660882e-06,
+      "loss": 1.2056,
+      "step": 23762
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8625575185587908e-06,
+      "loss": 1.3037,
+      "step": 23763
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8618316425193283e-06,
+      "loss": 0.8857,
+      "step": 23764
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.861105843155494e-06,
+      "loss": 0.6211,
+      "step": 23765
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8603801204750826e-06,
+      "loss": 1.1094,
+      "step": 23766
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8596544744858955e-06,
+      "loss": 1.1553,
+      "step": 23767
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8589289051957192e-06,
+      "loss": 0.8936,
+      "step": 23768
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.858203412612356e-06,
+      "loss": 1.1484,
+      "step": 23769
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8574779967435916e-06,
+      "loss": 0.8691,
+      "step": 23770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8567526575972247e-06,
+      "loss": 0.8662,
+      "step": 23771
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8560273951810434e-06,
+      "loss": 1.166,
+      "step": 23772
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.855302209502836e-06,
+      "loss": 1.3447,
+      "step": 23773
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8545771005703925e-06,
+      "loss": 0.7764,
+      "step": 23774
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8538520683915085e-06,
+      "loss": 1.0874,
+      "step": 23775
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8531271129739667e-06,
+      "loss": 0.6689,
+      "step": 23776
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.852402234325551e-06,
+      "loss": 0.6187,
+      "step": 23777
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.851677432454052e-06,
+      "loss": 0.7852,
+      "step": 23778
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.850952707367258e-06,
+      "loss": 0.9248,
+      "step": 23779
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8502280590729504e-06,
+      "loss": 0.9307,
+      "step": 23780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8495034875789106e-06,
+      "loss": 1.1025,
+      "step": 23781
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8487789928929275e-06,
+      "loss": 1.2354,
+      "step": 23782
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8480545750227796e-06,
+      "loss": 0.9668,
+      "step": 23783
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8473302339762467e-06,
+      "loss": 0.7554,
+      "step": 23784
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8466059697611116e-06,
+      "loss": 0.4567,
+      "step": 23785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.845881782385156e-06,
+      "loss": 1.2695,
+      "step": 23786
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8451576718561578e-06,
+      "loss": 0.7881,
+      "step": 23787
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8444336381818915e-06,
+      "loss": 0.8301,
+      "step": 23788
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.843709681370138e-06,
+      "loss": 1.1445,
+      "step": 23789
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8429858014286747e-06,
+      "loss": 0.8052,
+      "step": 23790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8422619983652765e-06,
+      "loss": 0.6689,
+      "step": 23791
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8415382721877137e-06,
+      "loss": 0.959,
+      "step": 23792
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.840814622903768e-06,
+      "loss": 0.7573,
+      "step": 23793
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8400910505212044e-06,
+      "loss": 0.9717,
+      "step": 23794
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8393675550478027e-06,
+      "loss": 0.877,
+      "step": 23795
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8386441364913285e-06,
+      "loss": 0.8379,
+      "step": 23796
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.837920794859558e-06,
+      "loss": 1.1416,
+      "step": 23797
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8371975301602572e-06,
+      "loss": 1.332,
+      "step": 23798
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8364743424011944e-06,
+      "loss": 1.4883,
+      "step": 23799
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.835751231590139e-06,
+      "loss": 0.7925,
+      "step": 23800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8350281977348626e-06,
+      "loss": 1.1689,
+      "step": 23801
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.834305240843127e-06,
+      "loss": 1.1133,
+      "step": 23802
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8335823609226965e-06,
+      "loss": 0.9414,
+      "step": 23803
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.832859557981339e-06,
+      "loss": 0.999,
+      "step": 23804
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8321368320268206e-06,
+      "loss": 0.7529,
+      "step": 23805
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.831414183066903e-06,
+      "loss": 1.1436,
+      "step": 23806
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.830691611109343e-06,
+      "loss": 1.3848,
+      "step": 23807
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8299691161619114e-06,
+      "loss": 0.8398,
+      "step": 23808
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.829246698232362e-06,
+      "loss": 0.8096,
+      "step": 23809
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.828524357328456e-06,
+      "loss": 1.1299,
+      "step": 23810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8278020934579574e-06,
+      "loss": 0.9658,
+      "step": 23811
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8270799066286214e-06,
+      "loss": 0.9468,
+      "step": 23812
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8263577968482035e-06,
+      "loss": 1.2119,
+      "step": 23813
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8256357641244603e-06,
+      "loss": 0.8242,
+      "step": 23814
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8249138084651485e-06,
+      "loss": 0.52,
+      "step": 23815
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8241919298780274e-06,
+      "loss": 1.0532,
+      "step": 23816
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.823470128370843e-06,
+      "loss": 0.8584,
+      "step": 23817
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8227484039513574e-06,
+      "loss": 0.918,
+      "step": 23818
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8220267566273184e-06,
+      "loss": 0.918,
+      "step": 23819
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8213051864064754e-06,
+      "loss": 1.0029,
+      "step": 23820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8205836932965824e-06,
+      "loss": 0.8525,
+      "step": 23821
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8198622773053907e-06,
+      "loss": 1.4141,
+      "step": 23822
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.819140938440649e-06,
+      "loss": 0.832,
+      "step": 23823
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8184196767101035e-06,
+      "loss": 0.9404,
+      "step": 23824
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8176984921214977e-06,
+      "loss": 1.0215,
+      "step": 23825
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8169773846825887e-06,
+      "loss": 1.0986,
+      "step": 23826
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8162563544011166e-06,
+      "loss": 1.3652,
+      "step": 23827
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8155354012848237e-06,
+      "loss": 0.6206,
+      "step": 23828
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8148145253414604e-06,
+      "loss": 0.7441,
+      "step": 23829
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.814093726578765e-06,
+      "loss": 0.6504,
+      "step": 23830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.813373005004484e-06,
+      "loss": 1.4922,
+      "step": 23831
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8126523606263544e-06,
+      "loss": 0.7134,
+      "step": 23832
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.811931793452123e-06,
+      "loss": 1.0391,
+      "step": 23833
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8112113034895273e-06,
+      "loss": 1.2051,
+      "step": 23834
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8104908907463023e-06,
+      "loss": 0.6455,
+      "step": 23835
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8097705552301903e-06,
+      "loss": 0.7471,
+      "step": 23836
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8090502969489306e-06,
+      "loss": 1.0723,
+      "step": 23837
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8083301159102592e-06,
+      "loss": 1.042,
+      "step": 23838
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8076100121219087e-06,
+      "loss": 0.9873,
+      "step": 23839
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8068899855916154e-06,
+      "loss": 1.2314,
+      "step": 23840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.806170036327117e-06,
+      "loss": 0.9424,
+      "step": 23841
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.805450164336145e-06,
+      "loss": 0.8623,
+      "step": 23842
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8047303696264295e-06,
+      "loss": 1.2412,
+      "step": 23843
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.804010652205705e-06,
+      "loss": 0.8955,
+      "step": 23844
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8032910120817034e-06,
+      "loss": 0.666,
+      "step": 23845
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8025714492621505e-06,
+      "loss": 1.0947,
+      "step": 23846
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.801851963754777e-06,
+      "loss": 1.1191,
+      "step": 23847
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.801132555567316e-06,
+      "loss": 0.8291,
+      "step": 23848
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8004132247074923e-06,
+      "loss": 1.0625,
+      "step": 23849
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.799693971183027e-06,
+      "loss": 1.2852,
+      "step": 23850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7989747950016534e-06,
+      "loss": 0.8682,
+      "step": 23851
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.798255696171096e-06,
+      "loss": 1.0303,
+      "step": 23852
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7975366746990785e-06,
+      "loss": 0.7998,
+      "step": 23853
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7968177305933197e-06,
+      "loss": 1.1514,
+      "step": 23854
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7960988638615494e-06,
+      "loss": 1.249,
+      "step": 23855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7953800745114824e-06,
+      "loss": 1.2549,
+      "step": 23856
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7946613625508458e-06,
+      "loss": 1.167,
+      "step": 23857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7939427279873543e-06,
+      "loss": 0.7856,
+      "step": 23858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7932241708287332e-06,
+      "loss": 1.0693,
+      "step": 23859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7925056910826987e-06,
+      "loss": 1.0552,
+      "step": 23860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.791787288756963e-06,
+      "loss": 1.2383,
+      "step": 23861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.791068963859248e-06,
+      "loss": 0.917,
+      "step": 23862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.790350716397272e-06,
+      "loss": 1.4023,
+      "step": 23863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.789632546378748e-06,
+      "loss": 0.6875,
+      "step": 23864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7889144538113867e-06,
+      "loss": 1.0518,
+      "step": 23865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7881964387029038e-06,
+      "loss": 1.0137,
+      "step": 23866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.787478501061015e-06,
+      "loss": 0.8511,
+      "step": 23867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.786760640893431e-06,
+      "loss": 1.1621,
+      "step": 23868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.786042858207858e-06,
+      "loss": 1.0684,
+      "step": 23869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.785325153012014e-06,
+      "loss": 0.7969,
+      "step": 23870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.784607525313603e-06,
+      "loss": 1.168,
+      "step": 23871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.783889975120332e-06,
+      "loss": 0.8906,
+      "step": 23872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7831725024399102e-06,
+      "loss": 0.999,
+      "step": 23873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7824551072800498e-06,
+      "loss": 0.876,
+      "step": 23874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.781737789648452e-06,
+      "loss": 1.0449,
+      "step": 23875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7810205495528186e-06,
+      "loss": 1.0654,
+      "step": 23876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.780303387000859e-06,
+      "loss": 1.4316,
+      "step": 23877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.779586302000278e-06,
+      "loss": 0.6572,
+      "step": 23878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7788692945587758e-06,
+      "loss": 0.6968,
+      "step": 23879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7781523646840515e-06,
+      "loss": 0.9707,
+      "step": 23880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7774355123838113e-06,
+      "loss": 0.9902,
+      "step": 23881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7767187376657502e-06,
+      "loss": 1.04,
+      "step": 23882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7760020405375732e-06,
+      "loss": 0.8569,
+      "step": 23883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7752854210069722e-06,
+      "loss": 0.8799,
+      "step": 23884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7745688790816527e-06,
+      "loss": 0.8311,
+      "step": 23885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7738524147693056e-06,
+      "loss": 0.7407,
+      "step": 23886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7731360280776276e-06,
+      "loss": 1.1719,
+      "step": 23887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.772419719014313e-06,
+      "loss": 1.0918,
+      "step": 23888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7717034875870617e-06,
+      "loss": 0.791,
+      "step": 23889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7709873338035644e-06,
+      "loss": 1.4551,
+      "step": 23890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7702712576715084e-06,
+      "loss": 0.7227,
+      "step": 23891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.769555259198591e-06,
+      "loss": 1.209,
+      "step": 23892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.768839338392505e-06,
+      "loss": 1.1875,
+      "step": 23893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7681234952609383e-06,
+      "loss": 1.2256,
+      "step": 23894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7674077298115752e-06,
+      "loss": 1.1201,
+      "step": 23895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7666920420521127e-06,
+      "loss": 0.6216,
+      "step": 23896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7659764319902315e-06,
+      "loss": 0.8423,
+      "step": 23897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.765260899633624e-06,
+      "loss": 0.9688,
+      "step": 23898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.764545444989971e-06,
+      "loss": 1.0713,
+      "step": 23899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7638300680669625e-06,
+      "loss": 0.8232,
+      "step": 23900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7631147688722814e-06,
+      "loss": 0.7749,
+      "step": 23901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7623995474136068e-06,
+      "loss": 0.5586,
+      "step": 23902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7616844036986256e-06,
+      "loss": 1.1587,
+      "step": 23903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7609693377350213e-06,
+      "loss": 1.0732,
+      "step": 23904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.760254349530472e-06,
+      "loss": 1.0127,
+      "step": 23905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7595394390926557e-06,
+      "loss": 1.0117,
+      "step": 23906
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.758824606429258e-06,
+      "loss": 1.0161,
+      "step": 23907
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.75810985154795e-06,
+      "loss": 1.0342,
+      "step": 23908
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7573951744564132e-06,
+      "loss": 0.8599,
+      "step": 23909
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7566805751623273e-06,
+      "loss": 0.9971,
+      "step": 23910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7559660536733645e-06,
+      "loss": 1.2715,
+      "step": 23911
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7552516099972015e-06,
+      "loss": 1.1328,
+      "step": 23912
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7545372441415054e-06,
+      "loss": 0.7988,
+      "step": 23913
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.753822956113962e-06,
+      "loss": 0.707,
+      "step": 23914
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7531087459222383e-06,
+      "loss": 1.0664,
+      "step": 23915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.752394613574002e-06,
+      "loss": 1.0029,
+      "step": 23916
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7516805590769315e-06,
+      "loss": 1.3145,
+      "step": 23917
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.75096658243869e-06,
+      "loss": 1.0762,
+      "step": 23918
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.750252683666953e-06,
+      "loss": 0.7969,
+      "step": 23919
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.749538862769383e-06,
+      "loss": 0.9756,
+      "step": 23920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.748825119753653e-06,
+      "loss": 0.7266,
+      "step": 23921
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.748111454627428e-06,
+      "loss": 1.3301,
+      "step": 23922
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.747397867398369e-06,
+      "loss": 0.6299,
+      "step": 23923
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7466843580741464e-06,
+      "loss": 0.9214,
+      "step": 23924
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7459709266624258e-06,
+      "loss": 1.0088,
+      "step": 23925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7452575731708685e-06,
+      "loss": 0.9717,
+      "step": 23926
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7445442976071335e-06,
+      "loss": 1.1797,
+      "step": 23927
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7438310999788865e-06,
+      "loss": 0.7874,
+      "step": 23928
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.74311798029379e-06,
+      "loss": 1.0249,
+      "step": 23929
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.742404938559502e-06,
+      "loss": 0.8491,
+      "step": 23930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.741691974783679e-06,
+      "loss": 1.1104,
+      "step": 23931
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.740979088973984e-06,
+      "loss": 0.8662,
+      "step": 23932
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.740266281138073e-06,
+      "loss": 0.7788,
+      "step": 23933
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.739553551283599e-06,
+      "loss": 0.877,
+      "step": 23934
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7388408994182214e-06,
+      "loss": 1.0205,
+      "step": 23935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7381283255495984e-06,
+      "loss": 0.9824,
+      "step": 23936
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7374158296853805e-06,
+      "loss": 0.9316,
+      "step": 23937
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.736703411833218e-06,
+      "loss": 0.6597,
+      "step": 23938
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7359910720007678e-06,
+      "loss": 1.0732,
+      "step": 23939
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.735278810195682e-06,
+      "loss": 0.9385,
+      "step": 23940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.734566626425611e-06,
+      "loss": 1.0488,
+      "step": 23941
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7338545206982003e-06,
+      "loss": 0.7749,
+      "step": 23942
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.733142493021107e-06,
+      "loss": 1.209,
+      "step": 23943
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7324305434019703e-06,
+      "loss": 0.605,
+      "step": 23944
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7317186718484468e-06,
+      "loss": 0.5728,
+      "step": 23945
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7310068783681755e-06,
+      "loss": 0.7607,
+      "step": 23946
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.730295162968809e-06,
+      "loss": 0.7578,
+      "step": 23947
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7295835256579883e-06,
+      "loss": 0.875,
+      "step": 23948
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.728871966443355e-06,
+      "loss": 1.123,
+      "step": 23949
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7281604853325563e-06,
+      "loss": 1.2471,
+      "step": 23950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7274490823332358e-06,
+      "loss": 0.5664,
+      "step": 23951
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.726737757453034e-06,
+      "loss": 1.0674,
+      "step": 23952
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7260265106995874e-06,
+      "loss": 0.8574,
+      "step": 23953
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.72531534208054e-06,
+      "loss": 1.0967,
+      "step": 23954
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7246042516035343e-06,
+      "loss": 0.9365,
+      "step": 23955
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.723893239276204e-06,
+      "loss": 0.9331,
+      "step": 23956
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7231823051061833e-06,
+      "loss": 1.0264,
+      "step": 23957
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7224714491011174e-06,
+      "loss": 1.0693,
+      "step": 23958
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7217606712686328e-06,
+      "loss": 0.9824,
+      "step": 23959
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.721049971616373e-06,
+      "loss": 1.001,
+      "step": 23960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7203393501519647e-06,
+      "loss": 0.8555,
+      "step": 23961
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.719628806883048e-06,
+      "loss": 0.876,
+      "step": 23962
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7189183418172515e-06,
+      "loss": 0.8413,
+      "step": 23963
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7182079549622044e-06,
+      "loss": 1.0381,
+      "step": 23964
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.717497646325539e-06,
+      "loss": 0.9863,
+      "step": 23965
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7167874159148898e-06,
+      "loss": 1.1006,
+      "step": 23966
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.716077263737883e-06,
+      "loss": 1.0,
+      "step": 23967
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.715367189802143e-06,
+      "loss": 0.6841,
+      "step": 23968
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.714657194115303e-06,
+      "loss": 0.9551,
+      "step": 23969
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.713947276684984e-06,
+      "loss": 0.6484,
+      "step": 23970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7132374375188177e-06,
+      "loss": 1.2217,
+      "step": 23971
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7125276766244224e-06,
+      "loss": 1.0664,
+      "step": 23972
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.711817994009428e-06,
+      "loss": 0.9229,
+      "step": 23973
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.711108389681455e-06,
+      "loss": 0.9395,
+      "step": 23974
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.710398863648124e-06,
+      "loss": 0.959,
+      "step": 23975
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.709689415917057e-06,
+      "loss": 0.9463,
+      "step": 23976
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7089800464958794e-06,
+      "loss": 0.9077,
+      "step": 23977
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.708270755392207e-06,
+      "loss": 1.0234,
+      "step": 23978
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.707561542613657e-06,
+      "loss": 0.7314,
+      "step": 23979
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7068524081678493e-06,
+      "loss": 0.876,
+      "step": 23980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7061433520624047e-06,
+      "loss": 0.9004,
+      "step": 23981
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7054343743049373e-06,
+      "loss": 1.0,
+      "step": 23982
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7047254749030582e-06,
+      "loss": 1.0068,
+      "step": 23983
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7040166538643887e-06,
+      "loss": 1.0459,
+      "step": 23984
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7033079111965367e-06,
+      "loss": 0.6631,
+      "step": 23985
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7025992469071217e-06,
+      "loss": 0.709,
+      "step": 23986
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.70189066100375e-06,
+      "loss": 1.127,
+      "step": 23987
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7011821534940388e-06,
+      "loss": 1.0918,
+      "step": 23988
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7004737243855937e-06,
+      "loss": 1.293,
+      "step": 23989
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6997653736860218e-06,
+      "loss": 1.3154,
+      "step": 23990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.699057101402942e-06,
+      "loss": 0.665,
+      "step": 23991
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6983489075439562e-06,
+      "loss": 0.8369,
+      "step": 23992
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6976407921166724e-06,
+      "loss": 1.2139,
+      "step": 23993
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.696932755128693e-06,
+      "loss": 1.5107,
+      "step": 23994
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.69622479658763e-06,
+      "loss": 1.167,
+      "step": 23995
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.695516916501082e-06,
+      "loss": 1.1514,
+      "step": 23996
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6948091148766586e-06,
+      "loss": 1.1367,
+      "step": 23997
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6941013917219573e-06,
+      "loss": 0.9404,
+      "step": 23998
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.693393747044586e-06,
+      "loss": 0.5142,
+      "step": 23999
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.692686180852142e-06,
+      "loss": 1.0254,
+      "step": 24000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6919786931522206e-06,
+      "loss": 1.0596,
+      "step": 24001
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6912712839524336e-06,
+      "loss": 0.9385,
+      "step": 24002
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6905639532603734e-06,
+      "loss": 1.2051,
+      "step": 24003
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6898567010836384e-06,
+      "loss": 0.8247,
+      "step": 24004
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6891495274298218e-06,
+      "loss": 0.8643,
+      "step": 24005
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6884424323065227e-06,
+      "loss": 1.2559,
+      "step": 24006
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68773541572134e-06,
+      "loss": 1.2715,
+      "step": 24007
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.687028477681862e-06,
+      "loss": 0.6343,
+      "step": 24008
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68632161819569e-06,
+      "loss": 1.0801,
+      "step": 24009
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6856148372704107e-06,
+      "loss": 0.9062,
+      "step": 24010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6849081349136152e-06,
+      "loss": 0.874,
+      "step": 24011
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.684201511132898e-06,
+      "loss": 1.0732,
+      "step": 24012
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68349496593585e-06,
+      "loss": 1.0332,
+      "step": 24013
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6827884993300602e-06,
+      "loss": 0.7661,
+      "step": 24014
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6820821113231165e-06,
+      "loss": 1.0479,
+      "step": 24015
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6813758019225998e-06,
+      "loss": 1.0459,
+      "step": 24016
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6806695711361096e-06,
+      "loss": 1.4521,
+      "step": 24017
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6799634189712264e-06,
+      "loss": 0.7783,
+      "step": 24018
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.679257345435532e-06,
+      "loss": 1.2188,
+      "step": 24019
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6785513505366165e-06,
+      "loss": 1.0098,
+      "step": 24020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.677845434282058e-06,
+      "loss": 0.5874,
+      "step": 24021
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6771395966794443e-06,
+      "loss": 0.895,
+      "step": 24022
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.676433837736353e-06,
+      "loss": 1.0312,
+      "step": 24023
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.675728157460369e-06,
+      "loss": 1.0312,
+      "step": 24024
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6750225558590714e-06,
+      "loss": 0.873,
+      "step": 24025
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.674317032940035e-06,
+      "loss": 1.0586,
+      "step": 24026
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6736115887108417e-06,
+      "loss": 0.9883,
+      "step": 24027
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6729062231790726e-06,
+      "loss": 0.7363,
+      "step": 24028
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6722009363523005e-06,
+      "loss": 0.7363,
+      "step": 24029
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6714957282381005e-06,
+      "loss": 0.8228,
+      "step": 24030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.670790598844051e-06,
+      "loss": 0.957,
+      "step": 24031
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.670085548177722e-06,
+      "loss": 0.8066,
+      "step": 24032
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6693805762466916e-06,
+      "loss": 1.0596,
+      "step": 24033
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6686756830585282e-06,
+      "loss": 0.9639,
+      "step": 24034
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6679708686208073e-06,
+      "loss": 1.1201,
+      "step": 24035
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6672661329410987e-06,
+      "loss": 1.4678,
+      "step": 24036
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6665614760269687e-06,
+      "loss": 0.8184,
+      "step": 24037
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.665856897885989e-06,
+      "loss": 0.9824,
+      "step": 24038
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.665152398525731e-06,
+      "loss": 1.1982,
+      "step": 24039
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6644479779537592e-06,
+      "loss": 1.0176,
+      "step": 24040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.663743636177638e-06,
+      "loss": 1.0566,
+      "step": 24041
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.663039373204935e-06,
+      "loss": 0.917,
+      "step": 24042
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6623351890432202e-06,
+      "loss": 0.6133,
+      "step": 24043
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6616310837000527e-06,
+      "loss": 0.8545,
+      "step": 24044
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.660927057182994e-06,
+      "loss": 0.9946,
+      "step": 24045
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.66022310949961e-06,
+      "loss": 1.0303,
+      "step": 24046
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.65951924065746e-06,
+      "loss": 0.9829,
+      "step": 24047
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6588154506641075e-06,
+      "loss": 0.9795,
+      "step": 24048
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6581117395271085e-06,
+      "loss": 1.2266,
+      "step": 24049
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.657408107254027e-06,
+      "loss": 0.8486,
+      "step": 24050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.656704553852418e-06,
+      "loss": 1.1152,
+      "step": 24051
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.656001079329836e-06,
+      "loss": 1.1836,
+      "step": 24052
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6552976836938415e-06,
+      "loss": 0.6807,
+      "step": 24053
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6545943669519914e-06,
+      "loss": 0.7266,
+      "step": 24054
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6538911291118384e-06,
+      "loss": 0.7148,
+      "step": 24055
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.653187970180934e-06,
+      "loss": 0.8721,
+      "step": 24056
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6524848901668363e-06,
+      "loss": 1.0762,
+      "step": 24057
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.651781889077092e-06,
+      "loss": 1.1562,
+      "step": 24058
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6510789669192582e-06,
+      "loss": 0.8086,
+      "step": 24059
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6503761237008795e-06,
+      "loss": 1.002,
+      "step": 24060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.649673359429511e-06,
+      "loss": 1.0723,
+      "step": 24061
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.648970674112701e-06,
+      "loss": 0.7764,
+      "step": 24062
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6482680677579907e-06,
+      "loss": 0.5757,
+      "step": 24063
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.647565540372934e-06,
+      "loss": 1.0869,
+      "step": 24064
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6468630919650772e-06,
+      "loss": 0.9512,
+      "step": 24065
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.646160722541964e-06,
+      "loss": 1.0947,
+      "step": 24066
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6454584321111365e-06,
+      "loss": 0.9854,
+      "step": 24067
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6447562206801413e-06,
+      "loss": 0.9629,
+      "step": 24068
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6440540882565236e-06,
+      "loss": 0.9331,
+      "step": 24069
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6433520348478226e-06,
+      "loss": 1.0449,
+      "step": 24070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.642650060461577e-06,
+      "loss": 0.9355,
+      "step": 24071
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6419481651053325e-06,
+      "loss": 1.0698,
+      "step": 24072
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6412463487866236e-06,
+      "loss": 1.0029,
+      "step": 24073
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.640544611512993e-06,
+      "loss": 1.373,
+      "step": 24074
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6398429532919743e-06,
+      "loss": 1.0098,
+      "step": 24075
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.639141374131111e-06,
+      "loss": 0.8213,
+      "step": 24076
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.638439874037935e-06,
+      "loss": 1.0796,
+      "step": 24077
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.637738453019978e-06,
+      "loss": 1.0391,
+      "step": 24078
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6370371110847782e-06,
+      "loss": 1.5352,
+      "step": 24079
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6363358482398726e-06,
+      "loss": 0.9814,
+      "step": 24080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.63563466449279e-06,
+      "loss": 1.2646,
+      "step": 24081
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6349335598510604e-06,
+      "loss": 0.7537,
+      "step": 24082
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.634232534322216e-06,
+      "loss": 1.0635,
+      "step": 24083
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6335315879137925e-06,
+      "loss": 1.416,
+      "step": 24084
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6328307206333135e-06,
+      "loss": 0.9365,
+      "step": 24085
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6321299324883064e-06,
+      "loss": 0.9912,
+      "step": 24086
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6314292234863046e-06,
+      "loss": 0.9814,
+      "step": 24087
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6307285936348316e-06,
+      "loss": 0.7534,
+      "step": 24088
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.630028042941406e-06,
+      "loss": 0.8442,
+      "step": 24089
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.629327571413567e-06,
+      "loss": 1.2441,
+      "step": 24090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6286271790588312e-06,
+      "loss": 0.9932,
+      "step": 24091
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6279268658847223e-06,
+      "loss": 1.2461,
+      "step": 24092
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6272266318987606e-06,
+      "loss": 0.9668,
+      "step": 24093
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.626526477108471e-06,
+      "loss": 0.9487,
+      "step": 24094
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6258264015213753e-06,
+      "loss": 1.0537,
+      "step": 24095
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.625126405144992e-06,
+      "loss": 0.9131,
+      "step": 24096
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6244264879868377e-06,
+      "loss": 0.8418,
+      "step": 24097
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6237266500544344e-06,
+      "loss": 1.2437,
+      "step": 24098
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.623026891355297e-06,
+      "loss": 1.2402,
+      "step": 24099
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.622327211896941e-06,
+      "loss": 1.0283,
+      "step": 24100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.621627611686889e-06,
+      "loss": 1.0811,
+      "step": 24101
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6209280907326494e-06,
+      "loss": 1.2451,
+      "step": 24102
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6202286490417396e-06,
+      "loss": 1.1973,
+      "step": 24103
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6195292866216635e-06,
+      "loss": 0.9443,
+      "step": 24104
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6188300034799474e-06,
+      "loss": 0.7495,
+      "step": 24105
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.618130799624097e-06,
+      "loss": 0.875,
+      "step": 24106
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.617431675061618e-06,
+      "loss": 0.8506,
+      "step": 24107
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6167326298000272e-06,
+      "loss": 1.0225,
+      "step": 24108
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6160336638468274e-06,
+      "loss": 0.9854,
+      "step": 24109
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.615334777209534e-06,
+      "loss": 0.8359,
+      "step": 24110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6146359698956457e-06,
+      "loss": 1.3848,
+      "step": 24111
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6139372419126775e-06,
+      "loss": 1.2637,
+      "step": 24112
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.613238593268129e-06,
+      "loss": 0.7466,
+      "step": 24113
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.612540023969504e-06,
+      "loss": 1.0605,
+      "step": 24114
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.611841534024309e-06,
+      "loss": 1.4639,
+      "step": 24115
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6111431234400486e-06,
+      "loss": 0.9788,
+      "step": 24116
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.610444792224224e-06,
+      "loss": 0.9736,
+      "step": 24117
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6097465403843315e-06,
+      "loss": 1.0176,
+      "step": 24118
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6090483679278777e-06,
+      "loss": 0.9653,
+      "step": 24119
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6083502748623567e-06,
+      "loss": 1.001,
+      "step": 24120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6076522611952737e-06,
+      "loss": 0.9883,
+      "step": 24121
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.60695432693412e-06,
+      "loss": 0.7998,
+      "step": 24122
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6062564720863982e-06,
+      "loss": 0.8574,
+      "step": 24123
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6055586966596013e-06,
+      "loss": 1.373,
+      "step": 24124
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.604861000661223e-06,
+      "loss": 0.7715,
+      "step": 24125
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.604163384098759e-06,
+      "loss": 1.2949,
+      "step": 24126
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.603465846979706e-06,
+      "loss": 0.8433,
+      "step": 24127
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6027683893115543e-06,
+      "loss": 0.749,
+      "step": 24128
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.602071011101792e-06,
+      "loss": 0.9385,
+      "step": 24129
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.601373712357915e-06,
+      "loss": 0.8423,
+      "step": 24130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.600676493087414e-06,
+      "loss": 0.7812,
+      "step": 24131
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.599979353297777e-06,
+      "loss": 1.1572,
+      "step": 24132
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.599282292996489e-06,
+      "loss": 1.1641,
+      "step": 24133
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5985853121910433e-06,
+      "loss": 1.3066,
+      "step": 24134
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5978884108889202e-06,
+      "loss": 1.0186,
+      "step": 24135
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.597191589097612e-06,
+      "loss": 0.9854,
+      "step": 24136
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5964948468245978e-06,
+      "loss": 1.0068,
+      "step": 24137
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5957981840773684e-06,
+      "loss": 1.1943,
+      "step": 24138
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.595101600863402e-06,
+      "loss": 0.7754,
+      "step": 24139
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.594405097190181e-06,
+      "loss": 0.9268,
+      "step": 24140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5937086730651884e-06,
+      "loss": 0.7896,
+      "step": 24141
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5930123284959063e-06,
+      "loss": 0.9922,
+      "step": 24142
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5923160634898147e-06,
+      "loss": 0.6826,
+      "step": 24143
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5916198780543878e-06,
+      "loss": 0.8789,
+      "step": 24144
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.590923772197107e-06,
+      "loss": 0.7725,
+      "step": 24145
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5902277459254533e-06,
+      "loss": 0.8818,
+      "step": 24146
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5895317992468993e-06,
+      "loss": 0.8394,
+      "step": 24147
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5888359321689183e-06,
+      "loss": 0.6938,
+      "step": 24148
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.588140144698991e-06,
+      "loss": 1.3438,
+      "step": 24149
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.587444436844586e-06,
+      "loss": 0.624,
+      "step": 24150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5867488086131775e-06,
+      "loss": 1.1787,
+      "step": 24151
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.586053260012238e-06,
+      "loss": 1.1611,
+      "step": 24152
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5853577910492424e-06,
+      "loss": 0.8579,
+      "step": 24153
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5846624017316568e-06,
+      "loss": 0.8569,
+      "step": 24154
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5839670920669514e-06,
+      "loss": 0.8525,
+      "step": 24155
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.583271862062594e-06,
+      "loss": 0.7734,
+      "step": 24156
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.582576711726058e-06,
+      "loss": 1.1572,
+      "step": 24157
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.581881641064806e-06,
+      "loss": 0.8936,
+      "step": 24158
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5811866500863025e-06,
+      "loss": 1.2666,
+      "step": 24159
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.580491738798019e-06,
+      "loss": 0.835,
+      "step": 24160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5797969072074114e-06,
+      "loss": 1.1484,
+      "step": 24161
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.579102155321952e-06,
+      "loss": 0.7933,
+      "step": 24162
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5784074831490967e-06,
+      "loss": 0.9863,
+      "step": 24163
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5777128906963143e-06,
+      "loss": 1.1553,
+      "step": 24164
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.577018377971061e-06,
+      "loss": 0.8369,
+      "step": 24165
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5763239449807954e-06,
+      "loss": 0.627,
+      "step": 24166
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.57562959173298e-06,
+      "loss": 1.0166,
+      "step": 24167
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5749353182350746e-06,
+      "loss": 1.0645,
+      "step": 24168
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5742411244945355e-06,
+      "loss": 0.6099,
+      "step": 24169
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5735470105188166e-06,
+      "loss": 1.0869,
+      "step": 24170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5728529763153754e-06,
+      "loss": 0.7373,
+      "step": 24171
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5721590218916703e-06,
+      "loss": 0.8135,
+      "step": 24172
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.571465147255153e-06,
+      "loss": 0.9678,
+      "step": 24173
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.570771352413275e-06,
+      "loss": 0.8408,
+      "step": 24174
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5700776373734927e-06,
+      "loss": 0.8916,
+      "step": 24175
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5693840021432537e-06,
+      "loss": 0.7354,
+      "step": 24176
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5686904467300123e-06,
+      "loss": 1.1797,
+      "step": 24177
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5679969711412143e-06,
+      "loss": 1.2178,
+      "step": 24178
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5673035753843144e-06,
+      "loss": 0.9893,
+      "step": 24179
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5666102594667585e-06,
+      "loss": 1.0352,
+      "step": 24180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5659170233959895e-06,
+      "loss": 0.8232,
+      "step": 24181
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.565223867179457e-06,
+      "loss": 0.8311,
+      "step": 24182
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.564530790824611e-06,
+      "loss": 1.5742,
+      "step": 24183
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5638377943388924e-06,
+      "loss": 0.6836,
+      "step": 24184
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.563144877729742e-06,
+      "loss": 0.8799,
+      "step": 24185
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5624520410046083e-06,
+      "loss": 0.9194,
+      "step": 24186
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5617592841709294e-06,
+      "loss": 0.8364,
+      "step": 24187
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5610666072361477e-06,
+      "loss": 1.0752,
+      "step": 24188
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5603740102077078e-06,
+      "loss": 1.2617,
+      "step": 24189
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5596814930930456e-06,
+      "loss": 1.2012,
+      "step": 24190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5589890558996013e-06,
+      "loss": 0.9336,
+      "step": 24191
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5582966986348057e-06,
+      "loss": 0.6758,
+      "step": 24192
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.557604421306107e-06,
+      "loss": 0.9834,
+      "step": 24193
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5569122239209366e-06,
+      "loss": 1.2832,
+      "step": 24194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5562201064867298e-06,
+      "loss": 1.0146,
+      "step": 24195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5555280690109174e-06,
+      "loss": 0.8271,
+      "step": 24196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.554836111500937e-06,
+      "loss": 1.0078,
+      "step": 24197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5541442339642243e-06,
+      "loss": 0.7153,
+      "step": 24198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5534524364082034e-06,
+      "loss": 0.9307,
+      "step": 24199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.552760718840314e-06,
+      "loss": 1.064,
+      "step": 24200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5520690812679805e-06,
+      "loss": 1.249,
+      "step": 24201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5513775236986316e-06,
+      "loss": 1.1348,
+      "step": 24202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5506860461396977e-06,
+      "loss": 1.1943,
+      "step": 24203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5499946485986094e-06,
+      "loss": 0.7632,
+      "step": 24204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5493033310827896e-06,
+      "loss": 0.873,
+      "step": 24205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.548612093599664e-06,
+      "loss": 0.9004,
+      "step": 24206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5479209361566604e-06,
+      "loss": 0.9199,
+      "step": 24207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.547229858761199e-06,
+      "loss": 0.7656,
+      "step": 24208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.546538861420709e-06,
+      "loss": 0.9302,
+      "step": 24209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5458479441426045e-06,
+      "loss": 0.8447,
+      "step": 24210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.545157106934316e-06,
+      "loss": 0.8711,
+      "step": 24211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5444663498032597e-06,
+      "loss": 0.855,
+      "step": 24212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.543775672756853e-06,
+      "loss": 0.96,
+      "step": 24213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5430850758025184e-06,
+      "loss": 0.8364,
+      "step": 24214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.542394558947676e-06,
+      "loss": 0.7861,
+      "step": 24215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5417041221997398e-06,
+      "loss": 1.002,
+      "step": 24216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.541013765566125e-06,
+      "loss": 1.1016,
+      "step": 24217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5403234890542483e-06,
+      "loss": 1.0137,
+      "step": 24218
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5396332926715284e-06,
+      "loss": 1.0488,
+      "step": 24219
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.538943176425376e-06,
+      "loss": 1.0303,
+      "step": 24220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5382531403232004e-06,
+      "loss": 1.1025,
+      "step": 24221
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5375631843724213e-06,
+      "loss": 0.8828,
+      "step": 24222
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.536873308580442e-06,
+      "loss": 0.7173,
+      "step": 24223
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.536183512954681e-06,
+      "loss": 1.127,
+      "step": 24224
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5354937975025395e-06,
+      "loss": 0.9795,
+      "step": 24225
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5348041622314335e-06,
+      "loss": 0.7925,
+      "step": 24226
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.534114607148769e-06,
+      "loss": 1.2461,
+      "step": 24227
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5334251322619473e-06,
+      "loss": 1.1455,
+      "step": 24228
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.532735737578379e-06,
+      "loss": 0.7349,
+      "step": 24229
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.532046423105472e-06,
+      "loss": 0.854,
+      "step": 24230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5313571888506283e-06,
+      "loss": 0.9258,
+      "step": 24231
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5306680348212476e-06,
+      "loss": 0.8203,
+      "step": 24232
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.529978961024736e-06,
+      "loss": 1.2773,
+      "step": 24233
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.529289967468499e-06,
+      "loss": 1.1211,
+      "step": 24234
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5286010541599326e-06,
+      "loss": 1.0107,
+      "step": 24235
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5279122211064353e-06,
+      "loss": 0.7209,
+      "step": 24236
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.527223468315412e-06,
+      "loss": 0.9893,
+      "step": 24237
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5265347957942577e-06,
+      "loss": 1.1338,
+      "step": 24238
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5258462035503673e-06,
+      "loss": 0.8086,
+      "step": 24239
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5251576915911404e-06,
+      "loss": 0.8928,
+      "step": 24240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.524469259923975e-06,
+      "loss": 1.1689,
+      "step": 24241
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.523780908556265e-06,
+      "loss": 0.9751,
+      "step": 24242
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5230926374953977e-06,
+      "loss": 1.1836,
+      "step": 24243
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.522404446748774e-06,
+      "loss": 0.9111,
+      "step": 24244
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.521716336323785e-06,
+      "loss": 1.2158,
+      "step": 24245
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.521028306227822e-06,
+      "loss": 0.7778,
+      "step": 24246
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5203403564682704e-06,
+      "loss": 0.9902,
+      "step": 24247
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5196524870525274e-06,
+      "loss": 0.9824,
+      "step": 24248
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5189646979879755e-06,
+      "loss": 1.1348,
+      "step": 24249
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.518276989282009e-06,
+      "loss": 0.8789,
+      "step": 24250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5175893609420076e-06,
+      "loss": 1.4023,
+      "step": 24251
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5169018129753654e-06,
+      "loss": 0.8232,
+      "step": 24252
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.516214345389464e-06,
+      "loss": 0.8252,
+      "step": 24253
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5155269581916853e-06,
+      "loss": 0.9946,
+      "step": 24254
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5148396513894145e-06,
+      "loss": 0.7578,
+      "step": 24255
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5141524249900386e-06,
+      "loss": 0.957,
+      "step": 24256
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5134652790009373e-06,
+      "loss": 1.0596,
+      "step": 24257
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5127782134294863e-06,
+      "loss": 0.6104,
+      "step": 24258
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.512091228283071e-06,
+      "loss": 1.3896,
+      "step": 24259
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.511404323569073e-06,
+      "loss": 0.6143,
+      "step": 24260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.510717499294868e-06,
+      "loss": 1.0596,
+      "step": 24261
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5100307554678307e-06,
+      "loss": 1.2246,
+      "step": 24262
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.509344092095344e-06,
+      "loss": 0.6616,
+      "step": 24263
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5086575091847766e-06,
+      "loss": 1.0186,
+      "step": 24264
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.507971006743509e-06,
+      "loss": 0.9434,
+      "step": 24265
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.507284584778913e-06,
+      "loss": 1.0898,
+      "step": 24266
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.506598243298364e-06,
+      "loss": 0.7705,
+      "step": 24267
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5059119823092336e-06,
+      "loss": 1.0454,
+      "step": 24268
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5052258018188903e-06,
+      "loss": 0.9043,
+      "step": 24269
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5045397018347063e-06,
+      "loss": 1.2549,
+      "step": 24270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.503853682364056e-06,
+      "loss": 0.5698,
+      "step": 24271
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.503167743414304e-06,
+      "loss": 0.8628,
+      "step": 24272
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5024818849928167e-06,
+      "loss": 1.083,
+      "step": 24273
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.501796107106966e-06,
+      "loss": 1.0488,
+      "step": 24274
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.501110409764115e-06,
+      "loss": 1.3125,
+      "step": 24275
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.500424792971632e-06,
+      "loss": 1.166,
+      "step": 24276
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4997392567368772e-06,
+      "loss": 1.3223,
+      "step": 24277
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4990538010672205e-06,
+      "loss": 0.8413,
+      "step": 24278
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4983684259700202e-06,
+      "loss": 1.0195,
+      "step": 24279
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.497683131452635e-06,
+      "loss": 0.978,
+      "step": 24280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4969979175224367e-06,
+      "loss": 1.1484,
+      "step": 24281
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4963127841867783e-06,
+      "loss": 0.832,
+      "step": 24282
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.495627731453022e-06,
+      "loss": 1.3188,
+      "step": 24283
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4949427593285215e-06,
+      "loss": 0.9951,
+      "step": 24284
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4942578678206374e-06,
+      "loss": 0.9712,
+      "step": 24285
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.493573056936731e-06,
+      "loss": 0.7549,
+      "step": 24286
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4928883266841507e-06,
+      "loss": 0.9453,
+      "step": 24287
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4922036770702586e-06,
+      "loss": 1.0645,
+      "step": 24288
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.491519108102406e-06,
+      "loss": 0.9844,
+      "step": 24289
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.490834619787943e-06,
+      "loss": 0.812,
+      "step": 24290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4901502121342248e-06,
+      "loss": 1.2266,
+      "step": 24291
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4894658851486076e-06,
+      "loss": 0.9946,
+      "step": 24292
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4887816388384367e-06,
+      "loss": 0.8506,
+      "step": 24293
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.488097473211063e-06,
+      "loss": 0.8877,
+      "step": 24294
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4874133882738305e-06,
+      "loss": 1.0957,
+      "step": 24295
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4867293840340987e-06,
+      "loss": 0.582,
+      "step": 24296
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.48604546049921e-06,
+      "loss": 0.877,
+      "step": 24297
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.485361617676506e-06,
+      "loss": 0.8604,
+      "step": 24298
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4846778555733387e-06,
+      "loss": 1.103,
+      "step": 24299
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.483994174197051e-06,
+      "loss": 1.1367,
+      "step": 24300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4833105735549825e-06,
+      "loss": 0.834,
+      "step": 24301
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4826270536544807e-06,
+      "loss": 1.0996,
+      "step": 24302
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4819436145028885e-06,
+      "loss": 1.0781,
+      "step": 24303
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.481260256107546e-06,
+      "loss": 1.4434,
+      "step": 24304
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4805769784757904e-06,
+      "loss": 1.0371,
+      "step": 24305
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.479893781614963e-06,
+      "loss": 0.9712,
+      "step": 24306
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4792106655324067e-06,
+      "loss": 0.9297,
+      "step": 24307
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4785276302354554e-06,
+      "loss": 0.8311,
+      "step": 24308
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4778446757314445e-06,
+      "loss": 1.124,
+      "step": 24309
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4771618020277144e-06,
+      "loss": 0.9399,
+      "step": 24310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.476479009131595e-06,
+      "loss": 0.7681,
+      "step": 24311
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.475796297050427e-06,
+      "loss": 0.7778,
+      "step": 24312
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4751136657915363e-06,
+      "loss": 0.8726,
+      "step": 24313
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4744311153622636e-06,
+      "loss": 0.8604,
+      "step": 24314
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4737486457699355e-06,
+      "loss": 1.1445,
+      "step": 24315
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4730662570218822e-06,
+      "loss": 0.9678,
+      "step": 24316
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.472383949125434e-06,
+      "loss": 1.2568,
+      "step": 24317
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.471701722087926e-06,
+      "loss": 1.001,
+      "step": 24318
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.471019575916681e-06,
+      "loss": 0.5151,
+      "step": 24319
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.470337510619024e-06,
+      "loss": 0.9868,
+      "step": 24320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4696555262022847e-06,
+      "loss": 0.8994,
+      "step": 24321
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4689736226737916e-06,
+      "loss": 0.874,
+      "step": 24322
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4682918000408673e-06,
+      "loss": 1.0464,
+      "step": 24323
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.467610058310832e-06,
+      "loss": 0.6709,
+      "step": 24324
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.466928397491014e-06,
+      "loss": 1.0439,
+      "step": 24325
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.46624681758873e-06,
+      "loss": 0.853,
+      "step": 24326
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4655653186113083e-06,
+      "loss": 0.6692,
+      "step": 24327
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.464883900566062e-06,
+      "loss": 0.979,
+      "step": 24328
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4642025634603172e-06,
+      "loss": 1.2812,
+      "step": 24329
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4635213073013884e-06,
+      "loss": 1.0674,
+      "step": 24330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4628401320965934e-06,
+      "loss": 0.9277,
+      "step": 24331
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4621590378532488e-06,
+      "loss": 0.8398,
+      "step": 24332
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4614780245786764e-06,
+      "loss": 0.8501,
+      "step": 24333
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.460797092280186e-06,
+      "loss": 1.0605,
+      "step": 24334
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4601162409650903e-06,
+      "loss": 1.0625,
+      "step": 24335
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4594354706407087e-06,
+      "loss": 1.0674,
+      "step": 24336
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4587547813143473e-06,
+      "loss": 0.9814,
+      "step": 24337
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4580741729933246e-06,
+      "loss": 1.1553,
+      "step": 24338
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4573936456849444e-06,
+      "loss": 0.9609,
+      "step": 24339
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.456713199396523e-06,
+      "loss": 1.2705,
+      "step": 24340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4560328341353666e-06,
+      "loss": 1.2891,
+      "step": 24341
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4553525499087816e-06,
+      "loss": 1.125,
+      "step": 24342
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4546723467240775e-06,
+      "loss": 1.0605,
+      "step": 24343
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.453992224588563e-06,
+      "loss": 0.9854,
+      "step": 24344
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.453312183509541e-06,
+      "loss": 1.0908,
+      "step": 24345
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4526322234943133e-06,
+      "loss": 0.7559,
+      "step": 24346
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.451952344550188e-06,
+      "loss": 0.79,
+      "step": 24347
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.45127254668447e-06,
+      "loss": 1.2246,
+      "step": 24348
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.450592829904459e-06,
+      "loss": 1.6484,
+      "step": 24349
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4499131942174527e-06,
+      "loss": 1.2021,
+      "step": 24350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.449233639630757e-06,
+      "loss": 1.1016,
+      "step": 24351
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4485541661516665e-06,
+      "loss": 1.0449,
+      "step": 24352
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4478747737874863e-06,
+      "loss": 0.9265,
+      "step": 24353
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4471954625455065e-06,
+      "loss": 1.0273,
+      "step": 24354
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4465162324330314e-06,
+      "loss": 0.5518,
+      "step": 24355
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4458370834573532e-06,
+      "loss": 0.8521,
+      "step": 24356
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.445158015625765e-06,
+      "loss": 1.1504,
+      "step": 24357
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4444790289455634e-06,
+      "loss": 0.8398,
+      "step": 24358
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.443800123424045e-06,
+      "loss": 0.7871,
+      "step": 24359
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4431212990684994e-06,
+      "loss": 1.1255,
+      "step": 24360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.442442555886214e-06,
+      "loss": 1.1484,
+      "step": 24361
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4417638938844877e-06,
+      "loss": 1.1348,
+      "step": 24362
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4410853130706046e-06,
+      "loss": 0.7471,
+      "step": 24363
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4404068134518578e-06,
+      "loss": 0.9502,
+      "step": 24364
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.43972839503553e-06,
+      "loss": 1.0273,
+      "step": 24365
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4390500578289154e-06,
+      "loss": 0.9521,
+      "step": 24366
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4383718018392977e-06,
+      "loss": 1.2285,
+      "step": 24367
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.437693627073958e-06,
+      "loss": 0.9375,
+      "step": 24368
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4370155335401847e-06,
+      "loss": 1.082,
+      "step": 24369
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4363375212452655e-06,
+      "loss": 1.001,
+      "step": 24370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4356595901964786e-06,
+      "loss": 1.1201,
+      "step": 24371
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4349817404011055e-06,
+      "loss": 0.8535,
+      "step": 24372
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4343039718664273e-06,
+      "loss": 0.9238,
+      "step": 24373
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4336262845997306e-06,
+      "loss": 1.1934,
+      "step": 24374
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.43294867860829e-06,
+      "loss": 0.9648,
+      "step": 24375
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4322711538993803e-06,
+      "loss": 0.8564,
+      "step": 24376
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.431593710480288e-06,
+      "loss": 1.209,
+      "step": 24377
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.430916348358281e-06,
+      "loss": 1.1611,
+      "step": 24378
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4302390675406408e-06,
+      "loss": 0.7969,
+      "step": 24379
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4295618680346434e-06,
+      "loss": 1.1025,
+      "step": 24380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.428884749847561e-06,
+      "loss": 1.3828,
+      "step": 24381
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4282077129866667e-06,
+      "loss": 0.8867,
+      "step": 24382
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.427530757459228e-06,
+      "loss": 0.8408,
+      "step": 24383
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4268538832725275e-06,
+      "loss": 1.1094,
+      "step": 24384
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.42617709043383e-06,
+      "loss": 0.7681,
+      "step": 24385
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.425500378950405e-06,
+      "loss": 0.6558,
+      "step": 24386
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.424823748829519e-06,
+      "loss": 0.9746,
+      "step": 24387
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4241472000784436e-06,
+      "loss": 0.9282,
+      "step": 24388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4234707327044484e-06,
+      "loss": 1.207,
+      "step": 24389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.422794346714794e-06,
+      "loss": 1.1709,
+      "step": 24390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.422118042116752e-06,
+      "loss": 0.7651,
+      "step": 24391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4214418189175837e-06,
+      "loss": 1.1182,
+      "step": 24392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4207656771245525e-06,
+      "loss": 0.7637,
+      "step": 24393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4200896167449163e-06,
+      "loss": 1.1445,
+      "step": 24394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4194136377859478e-06,
+      "loss": 0.6523,
+      "step": 24395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.418737740254904e-06,
+      "loss": 0.7637,
+      "step": 24396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.41806192415904e-06,
+      "loss": 0.6377,
+      "step": 24397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.417386189505623e-06,
+      "loss": 0.9136,
+      "step": 24398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.416710536301904e-06,
+      "loss": 1.1611,
+      "step": 24399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.416034964555147e-06,
+      "loss": 0.9434,
+      "step": 24400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.415359474272604e-06,
+      "loss": 0.9214,
+      "step": 24401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4146840654615356e-06,
+      "loss": 1.1504,
+      "step": 24402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4140087381291944e-06,
+      "loss": 0.9355,
+      "step": 24403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.41333349228283e-06,
+      "loss": 1.1602,
+      "step": 24404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.412658327929701e-06,
+      "loss": 0.8369,
+      "step": 24405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.411983245077062e-06,
+      "loss": 0.9541,
+      "step": 24406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.41130824373216e-06,
+      "loss": 1.0137,
+      "step": 24407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4106333239022452e-06,
+      "loss": 0.5693,
+      "step": 24408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.409958485594568e-06,
+      "loss": 1.1572,
+      "step": 24409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4092837288163807e-06,
+      "loss": 1.1689,
+      "step": 24410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4086090535749298e-06,
+      "loss": 1.0479,
+      "step": 24411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.407934459877458e-06,
+      "loss": 0.9941,
+      "step": 24412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4072599477312176e-06,
+      "loss": 1.0381,
+      "step": 24413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4065855171434495e-06,
+      "loss": 1.0596,
+      "step": 24414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.405911168121402e-06,
+      "loss": 0.9814,
+      "step": 24415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.405236900672314e-06,
+      "loss": 1.3687,
+      "step": 24416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4045627148034335e-06,
+      "loss": 1.0684,
+      "step": 24417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4038886105220005e-06,
+      "loss": 0.8438,
+      "step": 24418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4032145878352518e-06,
+      "loss": 0.752,
+      "step": 24419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4025406467504307e-06,
+      "loss": 0.9531,
+      "step": 24420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4018667872747815e-06,
+      "loss": 0.9863,
+      "step": 24421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4011930094155357e-06,
+      "loss": 1.1396,
+      "step": 24422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4005193131799308e-06,
+      "loss": 0.8164,
+      "step": 24423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3998456985752085e-06,
+      "loss": 0.9219,
+      "step": 24424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3991721656085987e-06,
+      "loss": 0.6646,
+      "step": 24425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3984987142873417e-06,
+      "loss": 1.0938,
+      "step": 24426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3978253446186663e-06,
+      "loss": 0.8335,
+      "step": 24427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3971520566098106e-06,
+      "loss": 1.0391,
+      "step": 24428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3964788502680038e-06,
+      "loss": 1.0225,
+      "step": 24429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3958057256004764e-06,
+      "loss": 1.0947,
+      "step": 24430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3951326826144584e-06,
+      "loss": 1.1641,
+      "step": 24431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.394459721317185e-06,
+      "loss": 0.9736,
+      "step": 24432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3937868417158816e-06,
+      "loss": 0.7207,
+      "step": 24433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3931140438177724e-06,
+      "loss": 0.7231,
+      "step": 24434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3924413276300873e-06,
+      "loss": 0.9829,
+      "step": 24435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3917686931600557e-06,
+      "loss": 0.957,
+      "step": 24436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.391096140414899e-06,
+      "loss": 1.0146,
+      "step": 24437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3904236694018403e-06,
+      "loss": 0.5967,
+      "step": 24438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3897512801281077e-06,
+      "loss": 0.9385,
+      "step": 24439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.389078972600918e-06,
+      "loss": 0.9468,
+      "step": 24440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3884067468274987e-06,
+      "loss": 0.9004,
+      "step": 24441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3877346028150638e-06,
+      "loss": 0.7969,
+      "step": 24442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3870625405708414e-06,
+      "loss": 0.8433,
+      "step": 24443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3863905601020465e-06,
+      "loss": 0.8047,
+      "step": 24444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3857186614158933e-06,
+      "loss": 0.9771,
+      "step": 24445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.385046844519604e-06,
+      "loss": 1.0771,
+      "step": 24446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.384375109420396e-06,
+      "loss": 1.376,
+      "step": 24447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3837034561254837e-06,
+      "loss": 1.123,
+      "step": 24448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3830318846420776e-06,
+      "loss": 1.2744,
+      "step": 24449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3823603949773942e-06,
+      "loss": 1.0049,
+      "step": 24450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3816889871386495e-06,
+      "loss": 0.8647,
+      "step": 24451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.381017661133054e-06,
+      "loss": 0.9238,
+      "step": 24452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.380346416967815e-06,
+      "loss": 0.8569,
+      "step": 24453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.379675254650148e-06,
+      "loss": 0.8506,
+      "step": 24454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.379004174187259e-06,
+      "loss": 1.2422,
+      "step": 24455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.378333175586356e-06,
+      "loss": 0.6855,
+      "step": 24456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.377662258854647e-06,
+      "loss": 1.1982,
+      "step": 24457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3769914239993432e-06,
+      "loss": 0.833,
+      "step": 24458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3763206710276477e-06,
+      "loss": 1.0605,
+      "step": 24459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.375649999946761e-06,
+      "loss": 1.1641,
+      "step": 24460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3749794107638913e-06,
+      "loss": 0.8945,
+      "step": 24461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3743089034862444e-06,
+      "loss": 0.9727,
+      "step": 24462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.37363847812102e-06,
+      "loss": 1.1147,
+      "step": 24463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3729681346754175e-06,
+      "loss": 0.9033,
+      "step": 24464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3722978731566405e-06,
+      "loss": 1.1123,
+      "step": 24465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3716276935718864e-06,
+      "loss": 0.8574,
+      "step": 24466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.370957595928357e-06,
+      "loss": 0.7881,
+      "step": 24467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.370287580233246e-06,
+      "loss": 0.9453,
+      "step": 24468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.369617646493756e-06,
+      "loss": 1.0352,
+      "step": 24469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3689477947170793e-06,
+      "loss": 0.8066,
+      "step": 24470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3682780249104064e-06,
+      "loss": 0.916,
+      "step": 24471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3676083370809442e-06,
+      "loss": 0.8672,
+      "step": 24472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.366938731235878e-06,
+      "loss": 1.2949,
+      "step": 24473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3662692073824035e-06,
+      "loss": 1.0605,
+      "step": 24474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3655997655277076e-06,
+      "loss": 0.9043,
+      "step": 24475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3649304056789834e-06,
+      "loss": 0.9688,
+      "step": 24476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.364261127843426e-06,
+      "loss": 1.0146,
+      "step": 24477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3635919320282164e-06,
+      "loss": 1.1279,
+      "step": 24478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3629228182405504e-06,
+      "loss": 0.8125,
+      "step": 24479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.362253786487613e-06,
+      "loss": 1.1816,
+      "step": 24480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.361584836776586e-06,
+      "loss": 0.7471,
+      "step": 24481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3609159691146577e-06,
+      "loss": 0.687,
+      "step": 24482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.360247183509018e-06,
+      "loss": 0.7773,
+      "step": 24483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3595784799668464e-06,
+      "loss": 0.6375,
+      "step": 24484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3589098584953253e-06,
+      "loss": 1.0566,
+      "step": 24485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3582413191016353e-06,
+      "loss": 0.7383,
+      "step": 24486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3575728617929594e-06,
+      "loss": 0.7939,
+      "step": 24487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3569044865764824e-06,
+      "loss": 1.1182,
+      "step": 24488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.356236193459376e-06,
+      "loss": 1.1426,
+      "step": 24489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3555679824488254e-06,
+      "loss": 0.9873,
+      "step": 24490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3548998535520053e-06,
+      "loss": 1.167,
+      "step": 24491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3542318067760895e-06,
+      "loss": 1.207,
+      "step": 24492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.353563842128257e-06,
+      "loss": 1.1279,
+      "step": 24493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.352895959615685e-06,
+      "loss": 0.6963,
+      "step": 24494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3522281592455475e-06,
+      "loss": 1.0117,
+      "step": 24495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3515604410250116e-06,
+      "loss": 0.7549,
+      "step": 24496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.350892804961253e-06,
+      "loss": 0.9961,
+      "step": 24497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3502252510614486e-06,
+      "loss": 0.5815,
+      "step": 24498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.349557779332764e-06,
+      "loss": 1.1719,
+      "step": 24499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3488903897823656e-06,
+      "loss": 1.3027,
+      "step": 24500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.348223082417429e-06,
+      "loss": 0.8223,
+      "step": 24501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3475558572451174e-06,
+      "loss": 0.8247,
+      "step": 24502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3468887142726017e-06,
+      "loss": 0.9482,
+      "step": 24503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.346221653507044e-06,
+      "loss": 0.6074,
+      "step": 24504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3455546749556134e-06,
+      "loss": 1.3828,
+      "step": 24505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3448877786254742e-06,
+      "loss": 1.2559,
+      "step": 24506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3442209645237846e-06,
+      "loss": 1.0371,
+      "step": 24507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.343554232657711e-06,
+      "loss": 1.0625,
+      "step": 24508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3428875830344188e-06,
+      "loss": 0.7183,
+      "step": 24509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.342221015661066e-06,
+      "loss": 0.8271,
+      "step": 24510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3415545305448084e-06,
+      "loss": 0.8828,
+      "step": 24511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3408881276928087e-06,
+      "loss": 1.1973,
+      "step": 24512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3402218071122285e-06,
+      "loss": 0.8203,
+      "step": 24513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.339555568810221e-06,
+      "loss": 1.2305,
+      "step": 24514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.338889412793941e-06,
+      "loss": 1.3799,
+      "step": 24515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.338223339070551e-06,
+      "loss": 0.7559,
+      "step": 24516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3375573476472003e-06,
+      "loss": 0.9336,
+      "step": 24517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3368914385310415e-06,
+      "loss": 1.2119,
+      "step": 24518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3362256117292293e-06,
+      "loss": 0.811,
+      "step": 24519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.33555986724892e-06,
+      "loss": 0.9199,
+      "step": 24520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.33489420509726e-06,
+      "loss": 1.1016,
+      "step": 24521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3342286252813996e-06,
+      "loss": 1.4385,
+      "step": 24522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.333563127808488e-06,
+      "loss": 1.0713,
+      "step": 24523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.332897712685679e-06,
+      "loss": 0.8926,
+      "step": 24524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3322323799201163e-06,
+      "loss": 1.3633,
+      "step": 24525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3315671295189436e-06,
+      "loss": 0.7178,
+      "step": 24526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3309019614893125e-06,
+      "loss": 0.6299,
+      "step": 24527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.330236875838363e-06,
+      "loss": 0.96,
+      "step": 24528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3295718725732454e-06,
+      "loss": 1.1338,
+      "step": 24529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3289069517010953e-06,
+      "loss": 1.1777,
+      "step": 24530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.328242113229062e-06,
+      "loss": 1.3613,
+      "step": 24531
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3275773571642836e-06,
+      "loss": 1.4277,
+      "step": 24532
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3269126835138988e-06,
+      "loss": 0.915,
+      "step": 24533
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3262480922850484e-06,
+      "loss": 0.8057,
+      "step": 24534
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.325583583484876e-06,
+      "loss": 1.1816,
+      "step": 24535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3249191571205153e-06,
+      "loss": 0.9385,
+      "step": 24536
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3242548131991017e-06,
+      "loss": 0.7915,
+      "step": 24537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.323590551727772e-06,
+      "loss": 0.9072,
+      "step": 24538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3229263727136674e-06,
+      "loss": 1.165,
+      "step": 24539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3222622761639167e-06,
+      "loss": 0.9473,
+      "step": 24540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3215982620856513e-06,
+      "loss": 1.0977,
+      "step": 24541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.320934330486011e-06,
+      "loss": 1.0029,
+      "step": 24542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3202704813721223e-06,
+      "loss": 1.0176,
+      "step": 24543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3196067147511147e-06,
+      "loss": 0.9331,
+      "step": 24544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.318943030630121e-06,
+      "loss": 1.1885,
+      "step": 24545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3182794290162725e-06,
+      "loss": 1.0918,
+      "step": 24546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.317615909916694e-06,
+      "loss": 1.0271,
+      "step": 24547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.316952473338512e-06,
+      "loss": 0.9395,
+      "step": 24548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3162891192888537e-06,
+      "loss": 1.2559,
+      "step": 24549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.315625847774848e-06,
+      "loss": 1.2295,
+      "step": 24550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3149626588036167e-06,
+      "loss": 1.1855,
+      "step": 24551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.314299552382282e-06,
+      "loss": 0.7686,
+      "step": 24552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3136365285179697e-06,
+      "loss": 1.1572,
+      "step": 24553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.312973587217798e-06,
+      "loss": 0.9062,
+      "step": 24554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3123107284888945e-06,
+      "loss": 1.0967,
+      "step": 24555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3116479523383718e-06,
+      "loss": 0.8174,
+      "step": 24556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.310985258773356e-06,
+      "loss": 1.1211,
+      "step": 24557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3103226478009622e-06,
+      "loss": 1.0361,
+      "step": 24558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3096601194283017e-06,
+      "loss": 0.9209,
+      "step": 24559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.308997673662503e-06,
+      "loss": 1.3477,
+      "step": 24560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3083353105106774e-06,
+      "loss": 0.9619,
+      "step": 24561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3076730299799378e-06,
+      "loss": 1.3755,
+      "step": 24562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3070108320773964e-06,
+      "loss": 0.7495,
+      "step": 24563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3063487168101696e-06,
+      "loss": 0.5752,
+      "step": 24564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.305686684185371e-06,
+      "loss": 1.1602,
+      "step": 24565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3050247342101096e-06,
+      "loss": 1.1523,
+      "step": 24566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3043628668914942e-06,
+      "loss": 0.8994,
+      "step": 24567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.303701082236639e-06,
+      "loss": 0.8115,
+      "step": 24568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.303039380252646e-06,
+      "loss": 1.2158,
+      "step": 24569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.302377760946628e-06,
+      "loss": 0.9326,
+      "step": 24570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.301716224325693e-06,
+      "loss": 1.0146,
+      "step": 24571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3010547703969445e-06,
+      "loss": 0.7007,
+      "step": 24572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.300393399167489e-06,
+      "loss": 0.751,
+      "step": 24573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2997321106444248e-06,
+      "loss": 1.4902,
+      "step": 24574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2990709048348616e-06,
+      "loss": 1.0234,
+      "step": 24575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2984097817459016e-06,
+      "loss": 1.3535,
+      "step": 24576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2977487413846433e-06,
+      "loss": 0.9854,
+      "step": 24577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2970877837581917e-06,
+      "loss": 0.9248,
+      "step": 24578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2964269088736437e-06,
+      "loss": 0.9854,
+      "step": 24579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.295766116738095e-06,
+      "loss": 0.8711,
+      "step": 24580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.295105407358648e-06,
+      "loss": 1.1123,
+      "step": 24581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2944447807423996e-06,
+      "loss": 0.7949,
+      "step": 24582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.293784236896447e-06,
+      "loss": 0.9658,
+      "step": 24583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2931237758278825e-06,
+      "loss": 0.6704,
+      "step": 24584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.292463397543797e-06,
+      "loss": 0.9639,
+      "step": 24585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2918031020512945e-06,
+      "loss": 0.8564,
+      "step": 24586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2911428893574606e-06,
+      "loss": 0.8477,
+      "step": 24587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.290482759469387e-06,
+      "loss": 0.7012,
+      "step": 24588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2898227123941675e-06,
+      "loss": 0.9189,
+      "step": 24589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.289162748138888e-06,
+      "loss": 1.4219,
+      "step": 24590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2885028667106435e-06,
+      "loss": 0.8955,
+      "step": 24591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.287843068116515e-06,
+      "loss": 0.9126,
+      "step": 24592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.287183352363598e-06,
+      "loss": 0.8555,
+      "step": 24593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.286523719458974e-06,
+      "loss": 1.0684,
+      "step": 24594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2858641694097263e-06,
+      "loss": 0.5811,
+      "step": 24595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2852047022229417e-06,
+      "loss": 0.9609,
+      "step": 24596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2845453179057076e-06,
+      "loss": 0.8232,
+      "step": 24597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2838860164651045e-06,
+      "loss": 1.0752,
+      "step": 24598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.28322679790821e-06,
+      "loss": 0.7148,
+      "step": 24599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.28256766224211e-06,
+      "loss": 0.5884,
+      "step": 24600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2819086094738865e-06,
+      "loss": 0.8662,
+      "step": 24601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2812496396106155e-06,
+      "loss": 0.9033,
+      "step": 24602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.280590752659373e-06,
+      "loss": 1.0225,
+      "step": 24603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.279931948627242e-06,
+      "loss": 1.2344,
+      "step": 24604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.279273227521297e-06,
+      "loss": 0.8901,
+      "step": 24605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.27861458934861e-06,
+      "loss": 0.8867,
+      "step": 24606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.27795603411626e-06,
+      "loss": 1.2012,
+      "step": 24607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.277297561831322e-06,
+      "loss": 0.8066,
+      "step": 24608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2766391725008673e-06,
+      "loss": 0.7739,
+      "step": 24609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.275980866131966e-06,
+      "loss": 0.9302,
+      "step": 24610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2753226427316897e-06,
+      "loss": 0.7458,
+      "step": 24611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2746645023071144e-06,
+      "loss": 1.0254,
+      "step": 24612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2740064448653067e-06,
+      "loss": 1.54,
+      "step": 24613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.273348470413329e-06,
+      "loss": 0.6885,
+      "step": 24614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2726905789582587e-06,
+      "loss": 0.9951,
+      "step": 24615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2720327705071543e-06,
+      "loss": 0.5527,
+      "step": 24616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2713750450670892e-06,
+      "loss": 0.9805,
+      "step": 24617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.270717402645122e-06,
+      "loss": 0.7148,
+      "step": 24618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2700598432483223e-06,
+      "loss": 1.0234,
+      "step": 24619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2694023668837507e-06,
+      "loss": 0.6714,
+      "step": 24620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2687449735584666e-06,
+      "loss": 1.0879,
+      "step": 24621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2680876632795346e-06,
+      "loss": 0.9297,
+      "step": 24622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.267430436054018e-06,
+      "loss": 1.3652,
+      "step": 24623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.266773291888974e-06,
+      "loss": 1.2549,
+      "step": 24624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2661162307914574e-06,
+      "loss": 1.2422,
+      "step": 24625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2654592527685305e-06,
+      "loss": 1.0039,
+      "step": 24626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2648023578272525e-06,
+      "loss": 1.0605,
+      "step": 24627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2641455459746766e-06,
+      "loss": 0.8916,
+      "step": 24628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2634888172178548e-06,
+      "loss": 1.0195,
+      "step": 24629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.262832171563848e-06,
+      "loss": 1.2832,
+      "step": 24630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2621756090197033e-06,
+      "loss": 0.7944,
+      "step": 24631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.261519129592479e-06,
+      "loss": 0.9775,
+      "step": 24632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.260862733289222e-06,
+      "loss": 0.709,
+      "step": 24633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.260206420116987e-06,
+      "loss": 1.2207,
+      "step": 24634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.259550190082823e-06,
+      "loss": 0.9629,
+      "step": 24635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2588940431937746e-06,
+      "loss": 1.0664,
+      "step": 24636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.258237979456893e-06,
+      "loss": 1.0498,
+      "step": 24637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.257581998879229e-06,
+      "loss": 0.9575,
+      "step": 24638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.256926101467827e-06,
+      "loss": 0.7837,
+      "step": 24639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.256270287229726e-06,
+      "loss": 1.1572,
+      "step": 24640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.25561455617198e-06,
+      "loss": 0.9355,
+      "step": 24641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2549589083016242e-06,
+      "loss": 1.2168,
+      "step": 24642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.254303343625709e-06,
+      "loss": 1.0527,
+      "step": 24643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.253647862151269e-06,
+      "loss": 1.0928,
+      "step": 24644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2529924638853516e-06,
+      "loss": 1.084,
+      "step": 24645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2523371488349954e-06,
+      "loss": 1.0752,
+      "step": 24646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2516819170072336e-06,
+      "loss": 0.7939,
+      "step": 24647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2510267684091104e-06,
+      "loss": 1.208,
+      "step": 24648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2503717030476634e-06,
+      "loss": 0.8184,
+      "step": 24649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2497167209299277e-06,
+      "loss": 0.9082,
+      "step": 24650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.249061822062936e-06,
+      "loss": 0.9775,
+      "step": 24651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2484070064537256e-06,
+      "loss": 0.8447,
+      "step": 24652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2477522741093327e-06,
+      "loss": 1.0332,
+      "step": 24653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.247097625036788e-06,
+      "loss": 0.8545,
+      "step": 24654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2464430592431207e-06,
+      "loss": 1.0303,
+      "step": 24655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.245788576735367e-06,
+      "loss": 0.8691,
+      "step": 24656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2451341775205527e-06,
+      "loss": 1.2129,
+      "step": 24657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2444798616057115e-06,
+      "loss": 1.0693,
+      "step": 24658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2438256289978656e-06,
+      "loss": 1.1182,
+      "step": 24659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2431714797040504e-06,
+      "loss": 0.606,
+      "step": 24660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2425174137312877e-06,
+      "loss": 1.0254,
+      "step": 24661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.241863431086598e-06,
+      "loss": 0.8452,
+      "step": 24662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2412095317770187e-06,
+      "loss": 0.896,
+      "step": 24663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.240555715809567e-06,
+      "loss": 1.124,
+      "step": 24664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.239901983191266e-06,
+      "loss": 0.8545,
+      "step": 24665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.239248333929135e-06,
+      "loss": 1.0801,
+      "step": 24666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.238594768030201e-06,
+      "loss": 0.9971,
+      "step": 24667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.237941285501478e-06,
+      "loss": 0.9146,
+      "step": 24668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2372878863499903e-06,
+      "loss": 1.0669,
+      "step": 24669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2366345705827575e-06,
+      "loss": 0.7593,
+      "step": 24670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2359813382067964e-06,
+      "loss": 1.0059,
+      "step": 24671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2353281892291212e-06,
+      "loss": 0.7344,
+      "step": 24672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2346751236567432e-06,
+      "loss": 1.207,
+      "step": 24673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.234022141496689e-06,
+      "loss": 0.7754,
+      "step": 24674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2333692427559673e-06,
+      "loss": 1.0557,
+      "step": 24675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.232716427441588e-06,
+      "loss": 1.0234,
+      "step": 24676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.232063695560569e-06,
+      "loss": 1.0449,
+      "step": 24677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.231411047119917e-06,
+      "loss": 1.0312,
+      "step": 24678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.230758482126646e-06,
+      "loss": 0.873,
+      "step": 24679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2301060005877627e-06,
+      "loss": 1.1143,
+      "step": 24680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2294536025102785e-06,
+      "loss": 1.0742,
+      "step": 24681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2288012879012022e-06,
+      "loss": 1.0615,
+      "step": 24682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2281490567675346e-06,
+      "loss": 1.249,
+      "step": 24683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2274969091162845e-06,
+      "loss": 0.7734,
+      "step": 24684
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2268448449544635e-06,
+      "loss": 1.4902,
+      "step": 24685
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2261928642890685e-06,
+      "loss": 0.793,
+      "step": 24686
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2255409671271034e-06,
+      "loss": 0.7593,
+      "step": 24687
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2248891534755713e-06,
+      "loss": 1.124,
+      "step": 24688
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2242374233414777e-06,
+      "loss": 0.6284,
+      "step": 24689
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2235857767318202e-06,
+      "loss": 1.3154,
+      "step": 24690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2229342136535946e-06,
+      "loss": 1.0938,
+      "step": 24691
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2222827341138075e-06,
+      "loss": 0.9453,
+      "step": 24692
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.221631338119451e-06,
+      "loss": 0.7593,
+      "step": 24693
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.220980025677526e-06,
+      "loss": 0.963,
+      "step": 24694
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.220328796795024e-06,
+      "loss": 0.793,
+      "step": 24695
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2196776514789455e-06,
+      "loss": 0.7334,
+      "step": 24696
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.219026589736283e-06,
+      "loss": 1.042,
+      "step": 24697
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2183756115740274e-06,
+      "loss": 0.6665,
+      "step": 24698
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.217724716999172e-06,
+      "loss": 1.0693,
+      "step": 24699
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.217073906018713e-06,
+      "loss": 0.7539,
+      "step": 24700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2164231786396383e-06,
+      "loss": 1.0361,
+      "step": 24701
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.215772534868934e-06,
+      "loss": 1.1641,
+      "step": 24702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2151219747135965e-06,
+      "loss": 0.7773,
+      "step": 24703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.214471498180606e-06,
+      "loss": 1.3633,
+      "step": 24704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2138211052769566e-06,
+      "loss": 0.8271,
+      "step": 24705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2131707960096293e-06,
+      "loss": 1.6113,
+      "step": 24706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2125205703856133e-06,
+      "loss": 1.2559,
+      "step": 24707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.211870428411893e-06,
+      "loss": 1.3389,
+      "step": 24708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.211220370095447e-06,
+      "loss": 1.0508,
+      "step": 24709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.210570395443261e-06,
+      "loss": 1.2656,
+      "step": 24710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2099205044623217e-06,
+      "loss": 1.3584,
+      "step": 24711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2092706971596044e-06,
+      "loss": 0.8188,
+      "step": 24712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.208620973542087e-06,
+      "loss": 0.8584,
+      "step": 24713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.207971333616753e-06,
+      "loss": 0.7866,
+      "step": 24714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.207321777390582e-06,
+      "loss": 0.877,
+      "step": 24715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2066723048705486e-06,
+      "loss": 1.1904,
+      "step": 24716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2060229160636273e-06,
+      "loss": 1.0508,
+      "step": 24717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2053736109767974e-06,
+      "loss": 1.1992,
+      "step": 24718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.20472438961703e-06,
+      "loss": 1.0449,
+      "step": 24719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2040752519913033e-06,
+      "loss": 0.8901,
+      "step": 24720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.203426198106584e-06,
+      "loss": 1.0625,
+      "step": 24721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.202777227969851e-06,
+      "loss": 0.8682,
+      "step": 24722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.202128341588071e-06,
+      "loss": 0.9717,
+      "step": 24723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.201479538968212e-06,
+      "loss": 0.8491,
+      "step": 24724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200830820117247e-06,
+      "loss": 0.7598,
+      "step": 24725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2001821850421447e-06,
+      "loss": 1.022,
+      "step": 24726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1995336337498717e-06,
+      "loss": 1.2314,
+      "step": 24727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.198885166247392e-06,
+      "loss": 1.291,
+      "step": 24728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1982367825416752e-06,
+      "loss": 0.8623,
+      "step": 24729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.197588482639681e-06,
+      "loss": 0.9072,
+      "step": 24730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1969402665483787e-06,
+      "loss": 0.9062,
+      "step": 24731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1962921342747268e-06,
+      "loss": 1.0186,
+      "step": 24732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.195644085825691e-06,
+      "loss": 0.7578,
+      "step": 24733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1949961212082303e-06,
+      "loss": 1.0244,
+      "step": 24734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.194348240429304e-06,
+      "loss": 1.1982,
+      "step": 24735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1937004434958707e-06,
+      "loss": 0.9795,
+      "step": 24736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.193052730414895e-06,
+      "loss": 1.1523,
+      "step": 24737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.192405101193329e-06,
+      "loss": 0.9668,
+      "step": 24738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1917575558381275e-06,
+      "loss": 0.8691,
+      "step": 24739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1911100943562503e-06,
+      "loss": 0.9893,
+      "step": 24740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.190462716754652e-06,
+      "loss": 1.5166,
+      "step": 24741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.189815423040287e-06,
+      "loss": 0.9404,
+      "step": 24742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.189168213220103e-06,
+      "loss": 1.4429,
+      "step": 24743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1885210873010586e-06,
+      "loss": 0.5874,
+      "step": 24744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1878740452901003e-06,
+      "loss": 1.2656,
+      "step": 24745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1872270871941824e-06,
+      "loss": 0.96,
+      "step": 24746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1865802130202497e-06,
+      "loss": 0.936,
+      "step": 24747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.185933422775256e-06,
+      "loss": 1.0791,
+      "step": 24748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.185286716466146e-06,
+      "loss": 1.1104,
+      "step": 24749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.184640094099861e-06,
+      "loss": 1.0908,
+      "step": 24750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.183993555683358e-06,
+      "loss": 0.5725,
+      "step": 24751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1833471012235765e-06,
+      "loss": 0.7689,
+      "step": 24752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1827007307274596e-06,
+      "loss": 0.9434,
+      "step": 24753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1820544442019476e-06,
+      "loss": 1.0723,
+      "step": 24754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1814082416539862e-06,
+      "loss": 0.8291,
+      "step": 24755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1807621230905197e-06,
+      "loss": 1.1875,
+      "step": 24756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1801160885184846e-06,
+      "loss": 1.0479,
+      "step": 24757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.179470137944819e-06,
+      "loss": 0.8179,
+      "step": 24758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.178824271376466e-06,
+      "loss": 0.9609,
+      "step": 24759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1781784888203604e-06,
+      "loss": 0.7466,
+      "step": 24760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1775327902834342e-06,
+      "loss": 0.9375,
+      "step": 24761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.176887175772633e-06,
+      "loss": 1.126,
+      "step": 24762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1762416452948875e-06,
+      "loss": 1.0898,
+      "step": 24763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1755961988571316e-06,
+      "loss": 1.3135,
+      "step": 24764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.174950836466296e-06,
+      "loss": 1.1357,
+      "step": 24765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.174305558129315e-06,
+      "loss": 1.0117,
+      "step": 24766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1736603638531218e-06,
+      "loss": 0.9326,
+      "step": 24767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1730152536446437e-06,
+      "loss": 1.0557,
+      "step": 24768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1723702275108136e-06,
+      "loss": 1.0137,
+      "step": 24769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.171725285458559e-06,
+      "loss": 0.6919,
+      "step": 24770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1710804274948037e-06,
+      "loss": 1.1094,
+      "step": 24771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.170435653626477e-06,
+      "loss": 1.1494,
+      "step": 24772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1697909638605096e-06,
+      "loss": 0.9004,
+      "step": 24773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.169146358203822e-06,
+      "loss": 0.4326,
+      "step": 24774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1685018366633348e-06,
+      "loss": 0.7915,
+      "step": 24775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1678573992459773e-06,
+      "loss": 0.8472,
+      "step": 24776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1672130459586706e-06,
+      "loss": 0.8311,
+      "step": 24777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1665687768083366e-06,
+      "loss": 0.9238,
+      "step": 24778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.165924591801891e-06,
+      "loss": 0.855,
+      "step": 24779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.16528049094626e-06,
+      "loss": 1.082,
+      "step": 24780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.164636474248356e-06,
+      "loss": 0.917,
+      "step": 24781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.163992541715102e-06,
+      "loss": 0.7087,
+      "step": 24782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.163348693353411e-06,
+      "loss": 1.0205,
+      "step": 24783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.162704929170204e-06,
+      "loss": 1.2471,
+      "step": 24784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1620612491723923e-06,
+      "loss": 0.6973,
+      "step": 24785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1614176533668874e-06,
+      "loss": 1.0625,
+      "step": 24786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1607741417606066e-06,
+      "loss": 0.8789,
+      "step": 24787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1601307143604643e-06,
+      "loss": 1.1885,
+      "step": 24788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.159487371173369e-06,
+      "loss": 1.1143,
+      "step": 24789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1588441122062274e-06,
+      "loss": 0.9014,
+      "step": 24790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1582009374659573e-06,
+      "loss": 1.1406,
+      "step": 24791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1575578469594606e-06,
+      "loss": 1.0884,
+      "step": 24792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.15691484069365e-06,
+      "loss": 0.791,
+      "step": 24793
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1562719186754287e-06,
+      "loss": 0.8457,
+      "step": 24794
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1556290809117062e-06,
+      "loss": 0.8057,
+      "step": 24795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1549863274093864e-06,
+      "loss": 0.9014,
+      "step": 24796
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1543436581753697e-06,
+      "loss": 1.1094,
+      "step": 24797
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1537010732165632e-06,
+      "loss": 0.8965,
+      "step": 24798
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.153058572539871e-06,
+      "loss": 0.8281,
+      "step": 24799
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1524161561521916e-06,
+      "loss": 1.0488,
+      "step": 24800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1517738240604246e-06,
+      "loss": 1.0479,
+      "step": 24801
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1511315762714713e-06,
+      "loss": 1.2715,
+      "step": 24802
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.150489412792234e-06,
+      "loss": 1.2031,
+      "step": 24803
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1498473336296066e-06,
+      "loss": 0.9497,
+      "step": 24804
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1492053387904844e-06,
+      "loss": 0.5437,
+      "step": 24805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.148563428281768e-06,
+      "loss": 0.8486,
+      "step": 24806
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1479216021103477e-06,
+      "loss": 0.75,
+      "step": 24807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1472798602831237e-06,
+      "loss": 0.7979,
+      "step": 24808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1466382028069833e-06,
+      "loss": 1.1611,
+      "step": 24809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1459966296888235e-06,
+      "loss": 0.6597,
+      "step": 24810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1453551409355353e-06,
+      "loss": 0.7557,
+      "step": 24811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.144713736554005e-06,
+      "loss": 0.834,
+      "step": 24812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.144072416551125e-06,
+      "loss": 1.3926,
+      "step": 24813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1434311809337883e-06,
+      "loss": 1.0576,
+      "step": 24814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1427900297088787e-06,
+      "loss": 1.46,
+      "step": 24815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1421489628832815e-06,
+      "loss": 0.6123,
+      "step": 24816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.141507980463884e-06,
+      "loss": 0.9805,
+      "step": 24817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.140867082457575e-06,
+      "loss": 1.2559,
+      "step": 24818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.140226268871237e-06,
+      "loss": 0.6307,
+      "step": 24819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1395855397117495e-06,
+      "loss": 0.6267,
+      "step": 24820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1389448949860005e-06,
+      "loss": 0.6289,
+      "step": 24821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.138304334700869e-06,
+      "loss": 1.1729,
+      "step": 24822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1376638588632336e-06,
+      "loss": 0.9658,
+      "step": 24823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.137023467479976e-06,
+      "loss": 0.9648,
+      "step": 24824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1363831605579786e-06,
+      "loss": 0.6519,
+      "step": 24825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.135742938104115e-06,
+      "loss": 0.915,
+      "step": 24826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1351028001252614e-06,
+      "loss": 1.0264,
+      "step": 24827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1344627466282954e-06,
+      "loss": 0.6465,
+      "step": 24828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1338227776200948e-06,
+      "loss": 0.9814,
+      "step": 24829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.133182893107533e-06,
+      "loss": 0.9902,
+      "step": 24830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.132543093097479e-06,
+      "loss": 0.6553,
+      "step": 24831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1319033775968102e-06,
+      "loss": 1.0352,
+      "step": 24832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.131263746612394e-06,
+      "loss": 0.9917,
+      "step": 24833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.130624200151107e-06,
+      "loss": 0.7363,
+      "step": 24834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1299847382198125e-06,
+      "loss": 1.1846,
+      "step": 24835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1293453608253844e-06,
+      "loss": 0.623,
+      "step": 24836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1287060679746894e-06,
+      "loss": 0.6431,
+      "step": 24837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.128066859674591e-06,
+      "loss": 1.0371,
+      "step": 24838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1274277359319582e-06,
+      "loss": 1.1787,
+      "step": 24839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1267886967536588e-06,
+      "loss": 1.291,
+      "step": 24840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.126149742146555e-06,
+      "loss": 0.8037,
+      "step": 24841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1255108721175066e-06,
+      "loss": 1.2129,
+      "step": 24842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1248720866733784e-06,
+      "loss": 0.8599,
+      "step": 24843
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.124233385821036e-06,
+      "loss": 0.5886,
+      "step": 24844
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1235947695673374e-06,
+      "loss": 0.9402,
+      "step": 24845
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1229562379191394e-06,
+      "loss": 1.3076,
+      "step": 24846
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.122317790883305e-06,
+      "loss": 1.2441,
+      "step": 24847
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1216794284666888e-06,
+      "loss": 1.0117,
+      "step": 24848
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1210411506761496e-06,
+      "loss": 0.6704,
+      "step": 24849
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.120402957518546e-06,
+      "loss": 0.8008,
+      "step": 24850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1197648490007304e-06,
+      "loss": 1.1396,
+      "step": 24851
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1191268251295583e-06,
+      "loss": 0.8496,
+      "step": 24852
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.118488885911879e-06,
+      "loss": 0.959,
+      "step": 24853
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1178510313545485e-06,
+      "loss": 1.1123,
+      "step": 24854
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1172132614644214e-06,
+      "loss": 0.7822,
+      "step": 24855
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1165755762483455e-06,
+      "loss": 0.9844,
+      "step": 24856
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.115937975713167e-06,
+      "loss": 0.9155,
+      "step": 24857
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1153004598657413e-06,
+      "loss": 0.8193,
+      "step": 24858
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.11466302871291e-06,
+      "loss": 0.8271,
+      "step": 24859
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1140256822615233e-06,
+      "loss": 0.9355,
+      "step": 24860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1133884205184298e-06,
+      "loss": 0.7812,
+      "step": 24861
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.112751243490473e-06,
+      "loss": 1.0747,
+      "step": 24862
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1121141511844967e-06,
+      "loss": 0.6968,
+      "step": 24863
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1114771436073377e-06,
+      "loss": 0.8213,
+      "step": 24864
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.110840220765852e-06,
+      "loss": 0.9658,
+      "step": 24865
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.110203382666873e-06,
+      "loss": 1.1104,
+      "step": 24866
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.10956662931724e-06,
+      "loss": 1.1025,
+      "step": 24867
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.108929960723799e-06,
+      "loss": 0.8955,
+      "step": 24868
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1082933768933812e-06,
+      "loss": 0.8555,
+      "step": 24869
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.107656877832832e-06,
+      "loss": 0.7822,
+      "step": 24870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.107020463548982e-06,
+      "loss": 1.0039,
+      "step": 24871
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1063841340486736e-06,
+      "loss": 0.8232,
+      "step": 24872
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1057478893387374e-06,
+      "loss": 1.2002,
+      "step": 24873
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1051117294260072e-06,
+      "loss": 0.96,
+      "step": 24874
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1044756543173184e-06,
+      "loss": 1.042,
+      "step": 24875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.103839664019506e-06,
+      "loss": 0.7988,
+      "step": 24876
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1032037585393994e-06,
+      "loss": 0.7441,
+      "step": 24877
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1025679378838247e-06,
+      "loss": 0.8994,
+      "step": 24878
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.10193220205962e-06,
+      "loss": 0.9561,
+      "step": 24879
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.101296551073606e-06,
+      "loss": 0.8286,
+      "step": 24880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1006609849326176e-06,
+      "loss": 1.0898,
+      "step": 24881
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.100025503643477e-06,
+      "loss": 1.2803,
+      "step": 24882
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.099390107213014e-06,
+      "loss": 1.2109,
+      "step": 24883
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0987547956480515e-06,
+      "loss": 0.6562,
+      "step": 24884
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0981195689554114e-06,
+      "loss": 0.9268,
+      "step": 24885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0974844271419206e-06,
+      "loss": 0.9736,
+      "step": 24886
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0968493702144032e-06,
+      "loss": 1.0771,
+      "step": 24887
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.096214398179678e-06,
+      "loss": 0.9873,
+      "step": 24888
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.095579511044563e-06,
+      "loss": 0.9502,
+      "step": 24889
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0949447088158812e-06,
+      "loss": 0.7686,
+      "step": 24890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.094309991500455e-06,
+      "loss": 0.8232,
+      "step": 24891
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0936753591050974e-06,
+      "loss": 0.9355,
+      "step": 24892
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.093040811636623e-06,
+      "loss": 1.0107,
+      "step": 24893
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0924063491018554e-06,
+      "loss": 1.0703,
+      "step": 24894
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0917719715076024e-06,
+      "loss": 0.9521,
+      "step": 24895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.091137678860684e-06,
+      "loss": 0.8633,
+      "step": 24896
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0905034711679083e-06,
+      "loss": 1.2021,
+      "step": 24897
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.089869348436093e-06,
+      "loss": 0.6299,
+      "step": 24898
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0892353106720485e-06,
+      "loss": 0.8145,
+      "step": 24899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0886013578825803e-06,
+      "loss": 1.1191,
+      "step": 24900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.087967490074502e-06,
+      "loss": 1.3652,
+      "step": 24901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.087333707254625e-06,
+      "loss": 1.042,
+      "step": 24902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0867000094297542e-06,
+      "loss": 0.7893,
+      "step": 24903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.086066396606694e-06,
+      "loss": 0.7354,
+      "step": 24904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.085432868792253e-06,
+      "loss": 1.2324,
+      "step": 24905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.084799425993239e-06,
+      "loss": 0.9507,
+      "step": 24906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0841660682164542e-06,
+      "loss": 0.8945,
+      "step": 24907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0835327954686977e-06,
+      "loss": 1.2212,
+      "step": 24908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.08289960775678e-06,
+      "loss": 1.1562,
+      "step": 24909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0822665050874967e-06,
+      "loss": 0.9512,
+      "step": 24910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.081633487467648e-06,
+      "loss": 1.0244,
+      "step": 24911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0810005549040346e-06,
+      "loss": 0.6177,
+      "step": 24912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.08036770740346e-06,
+      "loss": 0.6895,
+      "step": 24913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.079734944972717e-06,
+      "loss": 1.1875,
+      "step": 24914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0791022676186013e-06,
+      "loss": 1.1309,
+      "step": 24915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0784696753479106e-06,
+      "loss": 1.3271,
+      "step": 24916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0778371681674436e-06,
+      "loss": 1.0225,
+      "step": 24917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0772047460839908e-06,
+      "loss": 1.2852,
+      "step": 24918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0765724091043438e-06,
+      "loss": 0.793,
+      "step": 24919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0759401572352987e-06,
+      "loss": 0.7998,
+      "step": 24920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.075307990483645e-06,
+      "loss": 1.0176,
+      "step": 24921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.074675908856174e-06,
+      "loss": 0.9307,
+      "step": 24922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0740439123596735e-06,
+      "loss": 0.8789,
+      "step": 24923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.073412001000936e-06,
+      "loss": 1.3901,
+      "step": 24924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.072780174786746e-06,
+      "loss": 1.1689,
+      "step": 24925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0721484337238895e-06,
+      "loss": 1.1572,
+      "step": 24926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0715167778191535e-06,
+      "loss": 1.2119,
+      "step": 24927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.070885207079326e-06,
+      "loss": 1.1035,
+      "step": 24928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0702537215111895e-06,
+      "loss": 0.7393,
+      "step": 24929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.069622321121524e-06,
+      "loss": 0.9971,
+      "step": 24930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.068991005917114e-06,
+      "loss": 0.9561,
+      "step": 24931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0683597759047424e-06,
+      "loss": 0.7998,
+      "step": 24932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0677286310911894e-06,
+      "loss": 1.1982,
+      "step": 24933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.067097571483231e-06,
+      "loss": 0.9482,
+      "step": 24934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0664665970876496e-06,
+      "loss": 0.791,
+      "step": 24935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.06583570791122e-06,
+      "loss": 1.2705,
+      "step": 24936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0652049039607224e-06,
+      "loss": 0.7847,
+      "step": 24937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.064574185242929e-06,
+      "loss": 0.772,
+      "step": 24938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0639435517646187e-06,
+      "loss": 1.2959,
+      "step": 24939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.063313003532563e-06,
+      "loss": 0.7764,
+      "step": 24940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0626825405535343e-06,
+      "loss": 1.0518,
+      "step": 24941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.062052162834305e-06,
+      "loss": 1.2354,
+      "step": 24942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.061421870381649e-06,
+      "loss": 0.832,
+      "step": 24943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.060791663202336e-06,
+      "loss": 0.9512,
+      "step": 24944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.060161541303132e-06,
+      "loss": 0.8359,
+      "step": 24945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0595315046908094e-06,
+      "loss": 0.9297,
+      "step": 24946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0589015533721324e-06,
+      "loss": 0.8047,
+      "step": 24947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0582716873538698e-06,
+      "loss": 0.9805,
+      "step": 24948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0576419066427898e-06,
+      "loss": 1.0898,
+      "step": 24949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0570122112456547e-06,
+      "loss": 1.168,
+      "step": 24950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.056382601169229e-06,
+      "loss": 1.0732,
+      "step": 24951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0557530764202693e-06,
+      "loss": 0.8608,
+      "step": 24952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0551236370055494e-06,
+      "loss": 0.9414,
+      "step": 24953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.054494282931825e-06,
+      "loss": 0.7104,
+      "step": 24954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.053865014205856e-06,
+      "loss": 0.5576,
+      "step": 24955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0532358308343993e-06,
+      "loss": 0.7295,
+      "step": 24956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.052606732824216e-06,
+      "loss": 0.8521,
+      "step": 24957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.051977720182067e-06,
+      "loss": 1.2832,
+      "step": 24958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0513487929147013e-06,
+      "loss": 1.1924,
+      "step": 24959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.050719951028883e-06,
+      "loss": 1.3047,
+      "step": 24960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.050091194531363e-06,
+      "loss": 1.1621,
+      "step": 24961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.049462523428891e-06,
+      "loss": 1.0195,
+      "step": 24962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.048833937728225e-06,
+      "loss": 0.9238,
+      "step": 24963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.048205437436118e-06,
+      "loss": 0.3354,
+      "step": 24964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0475770225593204e-06,
+      "loss": 0.6973,
+      "step": 24965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.046948693104577e-06,
+      "loss": 1.1602,
+      "step": 24966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0463204490786417e-06,
+      "loss": 1.0391,
+      "step": 24967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0456922904882657e-06,
+      "loss": 0.771,
+      "step": 24968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0450642173401924e-06,
+      "loss": 1.1133,
+      "step": 24969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.044436229641167e-06,
+      "loss": 0.8066,
+      "step": 24970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.043808327397939e-06,
+      "loss": 0.9663,
+      "step": 24971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.043180510617252e-06,
+      "loss": 0.6084,
+      "step": 24972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0425527793058454e-06,
+      "loss": 1.3096,
+      "step": 24973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.041925133470467e-06,
+      "loss": 0.8125,
+      "step": 24974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.04129757311786e-06,
+      "loss": 0.834,
+      "step": 24975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.040670098254762e-06,
+      "loss": 1.1738,
+      "step": 24976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0400427088879115e-06,
+      "loss": 1.1523,
+      "step": 24977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0394154050240513e-06,
+      "loss": 0.8164,
+      "step": 24978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.03878818666992e-06,
+      "loss": 0.8818,
+      "step": 24979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.038161053832254e-06,
+      "loss": 1.04,
+      "step": 24980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.037534006517786e-06,
+      "loss": 0.8066,
+      "step": 24981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0369070447332583e-06,
+      "loss": 1.0381,
+      "step": 24982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0362801684853994e-06,
+      "loss": 0.8018,
+      "step": 24983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0356533777809475e-06,
+      "loss": 1.1572,
+      "step": 24984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.035026672626631e-06,
+      "loss": 1.0386,
+      "step": 24985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0344000530291875e-06,
+      "loss": 1.0654,
+      "step": 24986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0337735189953435e-06,
+      "loss": 0.8906,
+      "step": 24987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0331470705318278e-06,
+      "loss": 1.4482,
+      "step": 24988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0325207076453725e-06,
+      "loss": 0.834,
+      "step": 24989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.031894430342707e-06,
+      "loss": 1.0439,
+      "step": 24990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0312682386305573e-06,
+      "loss": 1.2812,
+      "step": 24991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0306421325156445e-06,
+      "loss": 1.1362,
+      "step": 24992
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0300161120047002e-06,
+      "loss": 0.876,
+      "step": 24993
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0293901771044487e-06,
+      "loss": 1.1416,
+      "step": 24994
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.028764327821613e-06,
+      "loss": 0.7803,
+      "step": 24995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0281385641629126e-06,
+      "loss": 1.083,
+      "step": 24996
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0275128861350725e-06,
+      "loss": 0.9883,
+      "step": 24997
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0268872937448113e-06,
+      "loss": 0.9189,
+      "step": 24998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0262617869988522e-06,
+      "loss": 0.778,
+      "step": 24999
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.025636365903908e-06,
+      "loss": 0.8535,
+      "step": 25000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0250110304667046e-06,
+      "loss": 0.7842,
+      "step": 25001
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.024385780693956e-06,
+      "loss": 1.6211,
+      "step": 25002
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.023760616592374e-06,
+      "loss": 0.918,
+      "step": 25003
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.023135538168678e-06,
+      "loss": 0.8999,
+      "step": 25004
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0225105454295835e-06,
+      "loss": 0.916,
+      "step": 25005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.021885638381804e-06,
+      "loss": 1.1982,
+      "step": 25006
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0212608170320467e-06,
+      "loss": 0.9326,
+      "step": 25007
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0206360813870298e-06,
+      "loss": 0.4768,
+      "step": 25008
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.020011431453458e-06,
+      "loss": 1.3242,
+      "step": 25009
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.019386867238048e-06,
+      "loss": 1.1104,
+      "step": 25010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0187623887475015e-06,
+      "loss": 1.1963,
+      "step": 25011
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0181379959885316e-06,
+      "loss": 0.9346,
+      "step": 25012
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0175136889678447e-06,
+      "loss": 0.8481,
+      "step": 25013
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0168894676921425e-06,
+      "loss": 0.9883,
+      "step": 25014
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0162653321681323e-06,
+      "loss": 0.7266,
+      "step": 25015
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.015641282402523e-06,
+      "loss": 0.5996,
+      "step": 25016
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.015017318402014e-06,
+      "loss": 0.7939,
+      "step": 25017
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0143934401733055e-06,
+      "loss": 0.9365,
+      "step": 25018
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0137696477231017e-06,
+      "loss": 0.8291,
+      "step": 25019
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0131459410581043e-06,
+      "loss": 1.1748,
+      "step": 25020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0125223201850118e-06,
+      "loss": 1.292,
+      "step": 25021
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0118987851105207e-06,
+      "loss": 1.1133,
+      "step": 25022
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.011275335841333e-06,
+      "loss": 0.7534,
+      "step": 25023
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0106519723841412e-06,
+      "loss": 0.4656,
+      "step": 25024
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0100286947456447e-06,
+      "loss": 0.5356,
+      "step": 25025
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0094055029325355e-06,
+      "loss": 0.7261,
+      "step": 25026
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.008782396951511e-06,
+      "loss": 1.1152,
+      "step": 25027
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0081593768092634e-06,
+      "loss": 0.8359,
+      "step": 25028
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.007536442512481e-06,
+      "loss": 0.812,
+      "step": 25029
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0069135940678585e-06,
+      "loss": 0.75,
+      "step": 25030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0062908314820893e-06,
+      "loss": 0.8662,
+      "step": 25031
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0056681547618596e-06,
+      "loss": 1.1201,
+      "step": 25032
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0050455639138555e-06,
+      "loss": 0.7671,
+      "step": 25033
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0044230589447698e-06,
+      "loss": 0.8188,
+      "step": 25034
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0038006398612844e-06,
+      "loss": 0.8809,
+      "step": 25035
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0031783066700903e-06,
+      "loss": 0.8545,
+      "step": 25036
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.002556059377866e-06,
+      "loss": 0.8457,
+      "step": 25037
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.001933897991303e-06,
+      "loss": 0.833,
+      "step": 25038
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.001311822517079e-06,
+      "loss": 0.9131,
+      "step": 25039
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0006898329618717e-06,
+      "loss": 0.9043,
+      "step": 25040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.000067929332374e-06,
+      "loss": 1.0811,
+      "step": 25041
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9994461116352603e-06,
+      "loss": 1.0693,
+      "step": 25042
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.99882437987721e-06,
+      "loss": 1.126,
+      "step": 25043
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.998202734064899e-06,
+      "loss": 0.9707,
+      "step": 25044
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.997581174205007e-06,
+      "loss": 0.9883,
+      "step": 25045
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.996959700304212e-06,
+      "loss": 1.1699,
+      "step": 25046
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9963383123691873e-06,
+      "loss": 0.8916,
+      "step": 25047
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9957170104066104e-06,
+      "loss": 0.999,
+      "step": 25048
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.995095794423153e-06,
+      "loss": 0.9688,
+      "step": 25049
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9944746644254864e-06,
+      "loss": 0.6748,
+      "step": 25050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.993853620420284e-06,
+      "loss": 0.9141,
+      "step": 25051
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9932326624142206e-06,
+      "loss": 0.8008,
+      "step": 25052
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.992611790413963e-06,
+      "loss": 1.2256,
+      "step": 25053
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9919910044261805e-06,
+      "loss": 1.1963,
+      "step": 25054
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.991370304457535e-06,
+      "loss": 1.0518,
+      "step": 25055
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9907496905147062e-06,
+      "loss": 0.9863,
+      "step": 25056
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.990129162604354e-06,
+      "loss": 0.8438,
+      "step": 25057
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9895087207331422e-06,
+      "loss": 1.1699,
+      "step": 25058
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9888883649077397e-06,
+      "loss": 1.0322,
+      "step": 25059
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9882680951348054e-06,
+      "loss": 1.083,
+      "step": 25060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.987647911421007e-06,
+      "loss": 1.1602,
+      "step": 25061
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9870278137730016e-06,
+      "loss": 0.8672,
+      "step": 25062
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9864078021974543e-06,
+      "loss": 0.7793,
+      "step": 25063
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.985787876701024e-06,
+      "loss": 1.3906,
+      "step": 25064
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9851680372903646e-06,
+      "loss": 1.2236,
+      "step": 25065
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.98454828397214e-06,
+      "loss": 1.0215,
+      "step": 25066
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.983928616753007e-06,
+      "loss": 0.8457,
+      "step": 25067
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9833090356396212e-06,
+      "loss": 0.7324,
+      "step": 25068
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9826895406386347e-06,
+      "loss": 0.8833,
+      "step": 25069
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.982070131756707e-06,
+      "loss": 0.5703,
+      "step": 25070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.981450809000486e-06,
+      "loss": 1.1738,
+      "step": 25071
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.98083157237663e-06,
+      "loss": 0.9805,
+      "step": 25072
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9802124218917853e-06,
+      "loss": 0.9531,
+      "step": 25073
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9795933575526073e-06,
+      "loss": 1.0527,
+      "step": 25074
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9789743793657445e-06,
+      "loss": 0.5768,
+      "step": 25075
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.978355487337842e-06,
+      "loss": 0.8579,
+      "step": 25076
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9777366814755507e-06,
+      "loss": 1.1582,
+      "step": 25077
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9771179617855196e-06,
+      "loss": 1.4609,
+      "step": 25078
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9764993282743927e-06,
+      "loss": 1.2334,
+      "step": 25079
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.975880780948812e-06,
+      "loss": 0.8301,
+      "step": 25080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9752623198154265e-06,
+      "loss": 0.9551,
+      "step": 25081
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9746439448808785e-06,
+      "loss": 0.7217,
+      "step": 25082
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.974025656151811e-06,
+      "loss": 1.0181,
+      "step": 25083
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9734074536348604e-06,
+      "loss": 1.0029,
+      "step": 25084
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9727893373366746e-06,
+      "loss": 0.8828,
+      "step": 25085
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.972171307263886e-06,
+      "loss": 1.1709,
+      "step": 25086
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9715533634231398e-06,
+      "loss": 1.0684,
+      "step": 25087
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9709355058210676e-06,
+      "loss": 1.1377,
+      "step": 25088
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.970317734464312e-06,
+      "loss": 0.5884,
+      "step": 25089
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9697000493595054e-06,
+      "loss": 0.7842,
+      "step": 25090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.969082450513282e-06,
+      "loss": 1.0195,
+      "step": 25091
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.968464937932277e-06,
+      "loss": 1.1289,
+      "step": 25092
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9678475116231267e-06,
+      "loss": 0.7446,
+      "step": 25093
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.967230171592459e-06,
+      "loss": 0.7935,
+      "step": 25094
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.966612917846905e-06,
+      "loss": 1.125,
+      "step": 25095
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9659957503930995e-06,
+      "loss": 0.6963,
+      "step": 25096
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9653786692376655e-06,
+      "loss": 0.9888,
+      "step": 25097
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.964761674387238e-06,
+      "loss": 0.915,
+      "step": 25098
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9641447658484382e-06,
+      "loss": 0.9873,
+      "step": 25099
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9635279436278985e-06,
+      "loss": 0.8525,
+      "step": 25100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.962911207732243e-06,
+      "loss": 0.7217,
+      "step": 25101
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9622945581680918e-06,
+      "loss": 0.793,
+      "step": 25102
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.961677994942073e-06,
+      "loss": 0.9092,
+      "step": 25103
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9610615180608116e-06,
+      "loss": 1.2725,
+      "step": 25104
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.960445127530927e-06,
+      "loss": 0.855,
+      "step": 25105
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.959828823359037e-06,
+      "loss": 0.8911,
+      "step": 25106
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9592126055517646e-06,
+      "loss": 0.832,
+      "step": 25107
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9585964741157317e-06,
+      "loss": 0.9395,
+      "step": 25108
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9579804290575544e-06,
+      "loss": 1.3438,
+      "step": 25109
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9573644703838467e-06,
+      "loss": 0.6543,
+      "step": 25110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.956748598101231e-06,
+      "loss": 0.751,
+      "step": 25111
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.956132812216318e-06,
+      "loss": 1.0337,
+      "step": 25112
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.955517112735725e-06,
+      "loss": 1.1006,
+      "step": 25113
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.954901499666064e-06,
+      "loss": 0.7383,
+      "step": 25114
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9542859730139497e-06,
+      "loss": 1.2939,
+      "step": 25115
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9536705327859927e-06,
+      "loss": 1.0684,
+      "step": 25116
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9530551789888007e-06,
+      "loss": 0.8872,
+      "step": 25117
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9524399116289874e-06,
+      "loss": 0.6477,
+      "step": 25118
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9518247307131634e-06,
+      "loss": 1.1396,
+      "step": 25119
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9512096362479336e-06,
+      "loss": 0.4756,
+      "step": 25120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9505946282399035e-06,
+      "loss": 0.8955,
+      "step": 25121
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9499797066956815e-06,
+      "loss": 1.1934,
+      "step": 25122
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.949364871621876e-06,
+      "loss": 0.9678,
+      "step": 25123
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.948750123025088e-06,
+      "loss": 0.7471,
+      "step": 25124
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9481354609119186e-06,
+      "loss": 0.8867,
+      "step": 25125
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.947520885288976e-06,
+      "loss": 0.8076,
+      "step": 25126
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9469063961628576e-06,
+      "loss": 1.0698,
+      "step": 25127
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9462919935401635e-06,
+      "loss": 1.1729,
+      "step": 25128
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9456776774274942e-06,
+      "loss": 1.0635,
+      "step": 25129
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.945063447831452e-06,
+      "loss": 0.8496,
+      "step": 25130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9444493047586333e-06,
+      "loss": 1.3438,
+      "step": 25131
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.94383524821563e-06,
+      "loss": 1.0703,
+      "step": 25132
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.943221278209041e-06,
+      "loss": 1.2949,
+      "step": 25133
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.942607394745466e-06,
+      "loss": 0.9922,
+      "step": 25134
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9419935978314943e-06,
+      "loss": 0.7305,
+      "step": 25135
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9413798874737177e-06,
+      "loss": 1.2178,
+      "step": 25136
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.940766263678733e-06,
+      "loss": 0.7476,
+      "step": 25137
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9401527264531274e-06,
+      "loss": 1.0483,
+      "step": 25138
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9395392758034913e-06,
+      "loss": 0.9688,
+      "step": 25139
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.93892591173642e-06,
+      "loss": 0.917,
+      "step": 25140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9383126342584992e-06,
+      "loss": 0.9941,
+      "step": 25141
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.937699443376314e-06,
+      "loss": 1.001,
+      "step": 25142
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9370863390964467e-06,
+      "loss": 1.2246,
+      "step": 25143
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.936473321425495e-06,
+      "loss": 1.2109,
+      "step": 25144
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.935860390370037e-06,
+      "loss": 1.0811,
+      "step": 25145
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.935247545936655e-06,
+      "loss": 1.0498,
+      "step": 25146
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.934634788131936e-06,
+      "loss": 0.8975,
+      "step": 25147
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9340221169624575e-06,
+      "loss": 1.1846,
+      "step": 25148
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.933409532434807e-06,
+      "loss": 0.9102,
+      "step": 25149
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.932797034555557e-06,
+      "loss": 0.9863,
+      "step": 25150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9321846233312925e-06,
+      "loss": 0.9639,
+      "step": 25151
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9315722987685913e-06,
+      "loss": 1.0312,
+      "step": 25152
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.930960060874025e-06,
+      "loss": 0.7314,
+      "step": 25153
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9303479096541767e-06,
+      "loss": 0.8174,
+      "step": 25154
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9297358451156202e-06,
+      "loss": 0.8818,
+      "step": 25155
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9291238672649305e-06,
+      "loss": 0.9229,
+      "step": 25156
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.928511976108677e-06,
+      "loss": 0.9922,
+      "step": 25157
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9279001716534386e-06,
+      "loss": 0.9785,
+      "step": 25158
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.927288453905781e-06,
+      "loss": 1.0244,
+      "step": 25159
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.926676822872281e-06,
+      "loss": 0.8154,
+      "step": 25160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9260652785595035e-06,
+      "loss": 1.124,
+      "step": 25161
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9254538209740216e-06,
+      "loss": 0.9131,
+      "step": 25162
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9248424501224016e-06,
+      "loss": 1.1646,
+      "step": 25163
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.924231166011208e-06,
+      "loss": 1.082,
+      "step": 25164
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.923619968647009e-06,
+      "loss": 0.8013,
+      "step": 25165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9230088580363727e-06,
+      "loss": 0.9141,
+      "step": 25166
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.922397834185862e-06,
+      "loss": 1.0439,
+      "step": 25167
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.921786897102036e-06,
+      "loss": 0.5952,
+      "step": 25168
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9211760467914596e-06,
+      "loss": 0.8618,
+      "step": 25169
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.920565283260699e-06,
+      "loss": 0.8057,
+      "step": 25170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.919954606516311e-06,
+      "loss": 1.042,
+      "step": 25171
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9193440165648524e-06,
+      "loss": 1.0186,
+      "step": 25172
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.918733513412887e-06,
+      "loss": 0.9097,
+      "step": 25173
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9181230970669683e-06,
+      "loss": 0.8525,
+      "step": 25174
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.917512767533658e-06,
+      "loss": 0.9146,
+      "step": 25175
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.916902524819506e-06,
+      "loss": 0.9648,
+      "step": 25176
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9162923689310743e-06,
+      "loss": 0.8955,
+      "step": 25177
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9156822998749126e-06,
+      "loss": 1.0303,
+      "step": 25178
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9150723176575724e-06,
+      "loss": 1.1699,
+      "step": 25179
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.914462422285608e-06,
+      "loss": 0.7627,
+      "step": 25180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9138526137655733e-06,
+      "loss": 0.958,
+      "step": 25181
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.913242892104017e-06,
+      "loss": 0.9849,
+      "step": 25182
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9126332573074848e-06,
+      "loss": 1.0996,
+      "step": 25183
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.912023709382529e-06,
+      "loss": 1.3311,
+      "step": 25184
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.911414248335698e-06,
+      "loss": 0.9414,
+      "step": 25185
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9108048741735364e-06,
+      "loss": 0.8154,
+      "step": 25186
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.910195586902588e-06,
+      "loss": 1.1816,
+      "step": 25187
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.909586386529402e-06,
+      "loss": 0.7559,
+      "step": 25188
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9089772730605204e-06,
+      "loss": 0.8179,
+      "step": 25189
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9083682465024833e-06,
+      "loss": 0.6904,
+      "step": 25190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9077593068618337e-06,
+      "loss": 1.0732,
+      "step": 25191
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9071504541451181e-06,
+      "loss": 0.6992,
+      "step": 25192
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9065416883588717e-06,
+      "loss": 0.9795,
+      "step": 25193
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9059330095096318e-06,
+      "loss": 0.5747,
+      "step": 25194
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9053244176039387e-06,
+      "loss": 0.8408,
+      "step": 25195
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9047159126483329e-06,
+      "loss": 1.0215,
+      "step": 25196
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9041074946493477e-06,
+      "loss": 0.7075,
+      "step": 25197
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9034991636135158e-06,
+      "loss": 1.0254,
+      "step": 25198
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9028909195473787e-06,
+      "loss": 1.002,
+      "step": 25199
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9022827624574614e-06,
+      "loss": 1.0527,
+      "step": 25200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.901674692350305e-06,
+      "loss": 1.0312,
+      "step": 25201
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9010667092324342e-06,
+      "loss": 0.8965,
+      "step": 25202
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9004588131103841e-06,
+      "loss": 1.0938,
+      "step": 25203
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8998510039906847e-06,
+      "loss": 0.873,
+      "step": 25204
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.89924328187986e-06,
+      "loss": 0.8442,
+      "step": 25205
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8986356467844402e-06,
+      "loss": 1.1045,
+      "step": 25206
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8980280987109566e-06,
+      "loss": 1.1045,
+      "step": 25207
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.897420637665932e-06,
+      "loss": 1.4023,
+      "step": 25208
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8968132636558889e-06,
+      "loss": 1.0078,
+      "step": 25209
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8962059766873541e-06,
+      "loss": 1.5332,
+      "step": 25210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8955987767668527e-06,
+      "loss": 1.1426,
+      "step": 25211
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8949916639009057e-06,
+      "loss": 0.676,
+      "step": 25212
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8943846380960317e-06,
+      "loss": 0.7178,
+      "step": 25213
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8937776993587554e-06,
+      "loss": 1.1245,
+      "step": 25214
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8931708476955912e-06,
+      "loss": 1.2178,
+      "step": 25215
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8925640831130642e-06,
+      "loss": 0.7568,
+      "step": 25216
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8919574056176848e-06,
+      "loss": 1.3486,
+      "step": 25217
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8913508152159777e-06,
+      "loss": 0.6836,
+      "step": 25218
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.890744311914453e-06,
+      "loss": 0.9863,
+      "step": 25219
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.890137895719626e-06,
+      "loss": 1.5,
+      "step": 25220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8895315666380098e-06,
+      "loss": 0.6863,
+      "step": 25221
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.888925324676123e-06,
+      "loss": 1.0908,
+      "step": 25222
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8883191698404735e-06,
+      "loss": 0.9893,
+      "step": 25223
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8877131021375705e-06,
+      "loss": 0.8604,
+      "step": 25224
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.887107121573928e-06,
+      "loss": 0.7661,
+      "step": 25225
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8865012281560523e-06,
+      "loss": 1.0244,
+      "step": 25226
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8858954218904546e-06,
+      "loss": 0.791,
+      "step": 25227
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8852897027836381e-06,
+      "loss": 1.0225,
+      "step": 25228
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8846840708421144e-06,
+      "loss": 0.874,
+      "step": 25229
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8840785260723871e-06,
+      "loss": 1.4141,
+      "step": 25230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8834730684809544e-06,
+      "loss": 1.1523,
+      "step": 25231
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8828676980743298e-06,
+      "loss": 0.9375,
+      "step": 25232
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.882262414859013e-06,
+      "loss": 0.8433,
+      "step": 25233
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8816572188415028e-06,
+      "loss": 0.6304,
+      "step": 25234
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8810521100283009e-06,
+      "loss": 1.0322,
+      "step": 25235
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8804470884259063e-06,
+      "loss": 0.9023,
+      "step": 25236
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.879842154040823e-06,
+      "loss": 0.8955,
+      "step": 25237
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8792373068795422e-06,
+      "loss": 1.0117,
+      "step": 25238
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8786325469485666e-06,
+      "loss": 1.0205,
+      "step": 25239
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.878027874254391e-06,
+      "loss": 1.042,
+      "step": 25240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.877423288803506e-06,
+      "loss": 0.5933,
+      "step": 25241
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8768187906024104e-06,
+      "loss": 1.083,
+      "step": 25242
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8762143796575982e-06,
+      "loss": 0.6172,
+      "step": 25243
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8756100559755596e-06,
+      "loss": 1.1699,
+      "step": 25244
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8750058195627874e-06,
+      "loss": 1.0146,
+      "step": 25245
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8744016704257684e-06,
+      "loss": 1.0762,
+      "step": 25246
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8737976085709942e-06,
+      "loss": 0.751,
+      "step": 25247
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8731936340049572e-06,
+      "loss": 1.0713,
+      "step": 25248
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8725897467341393e-06,
+      "loss": 0.999,
+      "step": 25249
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8719859467650314e-06,
+      "loss": 1.1348,
+      "step": 25250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.87138223410412e-06,
+      "loss": 1.0352,
+      "step": 25251
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8707786087578838e-06,
+      "loss": 0.8496,
+      "step": 25252
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8701750707328104e-06,
+      "loss": 1.1816,
+      "step": 25253
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8695716200353863e-06,
+      "loss": 1.1836,
+      "step": 25254
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8689682566720902e-06,
+      "loss": 0.6709,
+      "step": 25255
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8683649806494009e-06,
+      "loss": 0.7251,
+      "step": 25256
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8677617919738e-06,
+      "loss": 1.2178,
+      "step": 25257
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8671586906517713e-06,
+      "loss": 0.8188,
+      "step": 25258
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8665556766897897e-06,
+      "loss": 1.168,
+      "step": 25259
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8659527500943297e-06,
+      "loss": 0.9414,
+      "step": 25260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8653499108718732e-06,
+      "loss": 0.9697,
+      "step": 25261
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8647471590288902e-06,
+      "loss": 1.0205,
+      "step": 25262
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.86414449457186e-06,
+      "loss": 1.2793,
+      "step": 25263
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8635419175072511e-06,
+      "loss": 1.0771,
+      "step": 25264
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8629394278415435e-06,
+      "loss": 0.9531,
+      "step": 25265
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8623370255812034e-06,
+      "loss": 0.8213,
+      "step": 25266
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.861734710732701e-06,
+      "loss": 0.5103,
+      "step": 25267
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8611324833025069e-06,
+      "loss": 0.9199,
+      "step": 25268
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8605303432970946e-06,
+      "loss": 1.0205,
+      "step": 25269
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8599282907229287e-06,
+      "loss": 1.0137,
+      "step": 25270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8593263255864725e-06,
+      "loss": 1.0498,
+      "step": 25271
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8587244478941967e-06,
+      "loss": 0.9375,
+      "step": 25272
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.858122657652568e-06,
+      "loss": 1.0508,
+      "step": 25273
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8575209548680472e-06,
+      "loss": 0.6343,
+      "step": 25274
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8569193395470964e-06,
+      "loss": 1.1396,
+      "step": 25275
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8563178116961821e-06,
+      "loss": 0.5882,
+      "step": 25276
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8557163713217642e-06,
+      "loss": 0.8191,
+      "step": 25277
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8551150184303002e-06,
+      "loss": 0.7593,
+      "step": 25278
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8545137530282509e-06,
+      "loss": 0.8301,
+      "step": 25279
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8539125751220788e-06,
+      "loss": 1.0254,
+      "step": 25280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8533114847182387e-06,
+      "loss": 0.8618,
+      "step": 25281
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8527104818231857e-06,
+      "loss": 1.063,
+      "step": 25282
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8521095664433752e-06,
+      "loss": 1.2393,
+      "step": 25283
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8515087385852681e-06,
+      "loss": 1.0059,
+      "step": 25284
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8509079982553135e-06,
+      "loss": 0.792,
+      "step": 25285
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.850307345459962e-06,
+      "loss": 1.0137,
+      "step": 25286
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8497067802056711e-06,
+      "loss": 0.5835,
+      "step": 25287
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8491063024988876e-06,
+      "loss": 0.8467,
+      "step": 25288
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8485059123460646e-06,
+      "loss": 0.8438,
+      "step": 25289
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8479056097536485e-06,
+      "loss": 1.2334,
+      "step": 25290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8473053947280906e-06,
+      "loss": 1.1855,
+      "step": 25291
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.846705267275838e-06,
+      "loss": 0.8877,
+      "step": 25292
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8461052274033319e-06,
+      "loss": 1.1758,
+      "step": 25293
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8455052751170211e-06,
+      "loss": 1.3496,
+      "step": 25294
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8449054104233543e-06,
+      "loss": 0.8745,
+      "step": 25295
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8443056333287702e-06,
+      "loss": 0.9326,
+      "step": 25296
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8437059438397098e-06,
+      "loss": 0.9917,
+      "step": 25297
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8431063419626172e-06,
+      "loss": 0.7646,
+      "step": 25298
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8425068277039371e-06,
+      "loss": 0.7646,
+      "step": 25299
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8419074010701044e-06,
+      "loss": 1.1006,
+      "step": 25300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.841308062067556e-06,
+      "loss": 1.0615,
+      "step": 25301
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.840708810702735e-06,
+      "loss": 1.0439,
+      "step": 25302
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8401096469820746e-06,
+      "loss": 0.8228,
+      "step": 25303
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8395105709120143e-06,
+      "loss": 0.999,
+      "step": 25304
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8389115824989835e-06,
+      "loss": 0.8389,
+      "step": 25305
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8383126817494234e-06,
+      "loss": 1.167,
+      "step": 25306
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8377138686697637e-06,
+      "loss": 1.001,
+      "step": 25307
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8371151432664348e-06,
+      "loss": 0.8721,
+      "step": 25308
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8365165055458678e-06,
+      "loss": 1.415,
+      "step": 25309
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8359179555144991e-06,
+      "loss": 1.1333,
+      "step": 25310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8353194931787544e-06,
+      "loss": 1.1162,
+      "step": 25311
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8347211185450576e-06,
+      "loss": 0.7974,
+      "step": 25312
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8341228316198446e-06,
+      "loss": 0.9072,
+      "step": 25313
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8335246324095346e-06,
+      "loss": 0.8213,
+      "step": 25314
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8329265209205583e-06,
+      "loss": 0.7192,
+      "step": 25315
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.832328497159337e-06,
+      "loss": 0.9219,
+      "step": 25316
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.831730561132299e-06,
+      "loss": 0.9756,
+      "step": 25317
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8311327128458634e-06,
+      "loss": 0.9092,
+      "step": 25318
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8305349523064475e-06,
+      "loss": 0.7417,
+      "step": 25319
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8299372795204828e-06,
+      "loss": 0.8047,
+      "step": 25320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8293396944943841e-06,
+      "loss": 0.855,
+      "step": 25321
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8287421972345697e-06,
+      "loss": 0.7852,
+      "step": 25322
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8281447877474567e-06,
+      "loss": 0.748,
+      "step": 25323
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8275474660394642e-06,
+      "loss": 0.6812,
+      "step": 25324
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8269502321170097e-06,
+      "loss": 1.041,
+      "step": 25325
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8263530859865075e-06,
+      "loss": 0.8066,
+      "step": 25326
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8257560276543685e-06,
+      "loss": 0.9775,
+      "step": 25327
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8251590571270117e-06,
+      "loss": 1.0566,
+      "step": 25328
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8245621744108443e-06,
+      "loss": 1.3105,
+      "step": 25329
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8239653795122804e-06,
+      "loss": 0.7861,
+      "step": 25330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.823368672437732e-06,
+      "loss": 0.9043,
+      "step": 25331
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8227720531936077e-06,
+      "loss": 0.7959,
+      "step": 25332
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8221755217863156e-06,
+      "loss": 1.0986,
+      "step": 25333
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8215790782222587e-06,
+      "loss": 0.8691,
+      "step": 25334
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8209827225078537e-06,
+      "loss": 0.7637,
+      "step": 25335
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.820386454649501e-06,
+      "loss": 0.7197,
+      "step": 25336
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8197902746536022e-06,
+      "loss": 0.8237,
+      "step": 25337
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.819194182526568e-06,
+      "loss": 1.2725,
+      "step": 25338
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8185981782747987e-06,
+      "loss": 0.9785,
+      "step": 25339
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8180022619046933e-06,
+      "loss": 0.8359,
+      "step": 25340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.817406433422655e-06,
+      "loss": 0.9746,
+      "step": 25341
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.816810692835087e-06,
+      "loss": 0.5723,
+      "step": 25342
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8162150401483858e-06,
+      "loss": 0.644,
+      "step": 25343
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8156194753689505e-06,
+      "loss": 1.2183,
+      "step": 25344
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8150239985031715e-06,
+      "loss": 1.0381,
+      "step": 25345
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.814428609557458e-06,
+      "loss": 0.731,
+      "step": 25346
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8138333085381988e-06,
+      "loss": 0.7627,
+      "step": 25347
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8132380954517848e-06,
+      "loss": 0.8105,
+      "step": 25348
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8126429703046166e-06,
+      "loss": 0.9282,
+      "step": 25349
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8120479331030817e-06,
+      "loss": 0.9707,
+      "step": 25350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8114529838535754e-06,
+      "loss": 1.4531,
+      "step": 25351
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8108581225624843e-06,
+      "loss": 0.9258,
+      "step": 25352
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8102633492362032e-06,
+      "loss": 0.9443,
+      "step": 25353
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.809668663881118e-06,
+      "loss": 0.959,
+      "step": 25354
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8090740665036144e-06,
+      "loss": 0.9316,
+      "step": 25355
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.808479557110081e-06,
+      "loss": 0.876,
+      "step": 25356
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8078851357069083e-06,
+      "loss": 1.0317,
+      "step": 25357
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8072908023004777e-06,
+      "loss": 1.4111,
+      "step": 25358
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8066965568971705e-06,
+      "loss": 1.0493,
+      "step": 25359
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8061023995033723e-06,
+      "loss": 1.082,
+      "step": 25360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8055083301254683e-06,
+      "loss": 1.5527,
+      "step": 25361
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8049143487698374e-06,
+      "loss": 0.9023,
+      "step": 25362
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8043204554428572e-06,
+      "loss": 1.2109,
+      "step": 25363
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8037266501509121e-06,
+      "loss": 0.8369,
+      "step": 25364
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8031329329003755e-06,
+      "loss": 0.9502,
+      "step": 25365
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8025393036976302e-06,
+      "loss": 0.5752,
+      "step": 25366
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8019457625490476e-06,
+      "loss": 0.9746,
+      "step": 25367
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8013523094610076e-06,
+      "loss": 1.1416,
+      "step": 25368
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8007589444398832e-06,
+      "loss": 0.9678,
+      "step": 25369
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8001656674920453e-06,
+      "loss": 0.9341,
+      "step": 25370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.79957247862387e-06,
+      "loss": 1.1016,
+      "step": 25371
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7989793778417298e-06,
+      "loss": 0.9707,
+      "step": 25372
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.798386365151995e-06,
+      "loss": 0.96,
+      "step": 25373
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7977934405610309e-06,
+      "loss": 1.0684,
+      "step": 25374
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7972006040752144e-06,
+      "loss": 0.877,
+      "step": 25375
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7966078557009059e-06,
+      "loss": 1.2812,
+      "step": 25376
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7960151954444782e-06,
+      "loss": 0.7261,
+      "step": 25377
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7954226233122928e-06,
+      "loss": 1.2129,
+      "step": 25378
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7948301393107204e-06,
+      "loss": 0.9277,
+      "step": 25379
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7942377434461212e-06,
+      "loss": 1.0601,
+      "step": 25380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7936454357248578e-06,
+      "loss": 1.1064,
+      "step": 25381
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7930532161532943e-06,
+      "loss": 0.9714,
+      "step": 25382
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7924610847377943e-06,
+      "loss": 0.8223,
+      "step": 25383
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.791869041484715e-06,
+      "loss": 1.4199,
+      "step": 25384
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7912770864004158e-06,
+      "loss": 1.0742,
+      "step": 25385
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7906852194912562e-06,
+      "loss": 1.2393,
+      "step": 25386
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7900934407635961e-06,
+      "loss": 0.6992,
+      "step": 25387
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7895017502237899e-06,
+      "loss": 0.7939,
+      "step": 25388
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7889101478781913e-06,
+      "loss": 0.8271,
+      "step": 25389
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7883186337331594e-06,
+      "loss": 1.2314,
+      "step": 25390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7877272077950435e-06,
+      "loss": 0.7314,
+      "step": 25391
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7871358700702014e-06,
+      "loss": 1.0039,
+      "step": 25392
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7865446205649795e-06,
+      "loss": 1.0454,
+      "step": 25393
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7859534592857353e-06,
+      "loss": 1.0361,
+      "step": 25394
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7853623862388148e-06,
+      "loss": 1.0435,
+      "step": 25395
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.784771401430565e-06,
+      "loss": 0.9834,
+      "step": 25396
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7841805048673378e-06,
+      "loss": 0.541,
+      "step": 25397
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.78358969655548e-06,
+      "loss": 0.6787,
+      "step": 25398
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.782998976501339e-06,
+      "loss": 0.7271,
+      "step": 25399
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7824083447112538e-06,
+      "loss": 0.7749,
+      "step": 25400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7818178011915765e-06,
+      "loss": 0.9741,
+      "step": 25401
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7812273459486451e-06,
+      "loss": 0.8579,
+      "step": 25402
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7806369789888068e-06,
+      "loss": 1.0039,
+      "step": 25403
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7800467003183975e-06,
+      "loss": 0.8286,
+      "step": 25404
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7794565099437632e-06,
+      "loss": 0.8086,
+      "step": 25405
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7788664078712414e-06,
+      "loss": 0.9463,
+      "step": 25406
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7782763941071667e-06,
+      "loss": 0.4941,
+      "step": 25407
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.777686468657882e-06,
+      "loss": 1.0674,
+      "step": 25408
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7770966315297245e-06,
+      "loss": 0.9355,
+      "step": 25409
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7765068827290277e-06,
+      "loss": 0.9707,
+      "step": 25410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7759172222621245e-06,
+      "loss": 1.1221,
+      "step": 25411
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7753276501353522e-06,
+      "loss": 1.1484,
+      "step": 25412
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7747381663550445e-06,
+      "loss": 0.6299,
+      "step": 25413
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7741487709275318e-06,
+      "loss": 1.1094,
+      "step": 25414
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7735594638591424e-06,
+      "loss": 1.0098,
+      "step": 25415
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7729702451562115e-06,
+      "loss": 1.0537,
+      "step": 25416
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7723811148250635e-06,
+      "loss": 0.9463,
+      "step": 25417
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7717920728720284e-06,
+      "loss": 1.0332,
+      "step": 25418
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7712031193034373e-06,
+      "loss": 1.0059,
+      "step": 25419
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7706142541256134e-06,
+      "loss": 0.9932,
+      "step": 25420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.770025477344881e-06,
+      "loss": 1.0801,
+      "step": 25421
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7694367889675613e-06,
+      "loss": 0.9531,
+      "step": 25422
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7688481889999864e-06,
+      "loss": 0.8984,
+      "step": 25423
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.768259677448475e-06,
+      "loss": 0.7656,
+      "step": 25424
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7676712543193475e-06,
+      "loss": 1.0547,
+      "step": 25425
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.767082919618922e-06,
+      "loss": 0.8574,
+      "step": 25426
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7664946733535227e-06,
+      "loss": 1.187,
+      "step": 25427
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7659065155294686e-06,
+      "loss": 0.8657,
+      "step": 25428
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7653184461530737e-06,
+      "loss": 0.916,
+      "step": 25429
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7647304652306597e-06,
+      "loss": 0.6328,
+      "step": 25430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.764142572768539e-06,
+      "loss": 0.8989,
+      "step": 25431
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.763554768773027e-06,
+      "loss": 1.2168,
+      "step": 25432
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7629670532504328e-06,
+      "loss": 0.6724,
+      "step": 25433
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7623794262070805e-06,
+      "loss": 0.8896,
+      "step": 25434
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7617918876492757e-06,
+      "loss": 1.0371,
+      "step": 25435
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.761204437583327e-06,
+      "loss": 0.8438,
+      "step": 25436
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7606170760155516e-06,
+      "loss": 0.9805,
+      "step": 25437
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7600298029522512e-06,
+      "loss": 1.498,
+      "step": 25438
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7594426183997393e-06,
+      "loss": 0.8994,
+      "step": 25439
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.75885552236432e-06,
+      "loss": 0.9761,
+      "step": 25440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7582685148523037e-06,
+      "loss": 0.9375,
+      "step": 25441
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7576815958699932e-06,
+      "loss": 0.7832,
+      "step": 25442
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7570947654236914e-06,
+      "loss": 0.8311,
+      "step": 25443
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7565080235197017e-06,
+      "loss": 1.1553,
+      "step": 25444
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7559213701643329e-06,
+      "loss": 0.9082,
+      "step": 25445
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7553348053638808e-06,
+      "loss": 0.7188,
+      "step": 25446
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.754748329124646e-06,
+      "loss": 0.9004,
+      "step": 25447
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7541619414529288e-06,
+      "loss": 1.2617,
+      "step": 25448
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7535756423550321e-06,
+      "loss": 1.2412,
+      "step": 25449
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7529894318372498e-06,
+      "loss": 1.2085,
+      "step": 25450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.752403309905878e-06,
+      "loss": 0.8511,
+      "step": 25451
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.751817276567216e-06,
+      "loss": 1.0391,
+      "step": 25452
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7512313318275532e-06,
+      "loss": 0.5967,
+      "step": 25453
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7506454756931911e-06,
+      "loss": 1.2646,
+      "step": 25454
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7500597081704162e-06,
+      "loss": 1.0977,
+      "step": 25455
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7494740292655254e-06,
+      "loss": 1.208,
+      "step": 25456
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7488884389848072e-06,
+      "loss": 0.8347,
+      "step": 25457
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7483029373345505e-06,
+      "loss": 1.0586,
+      "step": 25458
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7477175243210465e-06,
+      "loss": 1.0767,
+      "step": 25459
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7471321999505853e-06,
+      "loss": 0.7109,
+      "step": 25460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.746546964229453e-06,
+      "loss": 1.2227,
+      "step": 25461
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7459618171639326e-06,
+      "loss": 1.1045,
+      "step": 25462
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7453767587603144e-06,
+      "loss": 0.7617,
+      "step": 25463
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7447917890248788e-06,
+      "loss": 1.0352,
+      "step": 25464
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7442069079639135e-06,
+      "loss": 0.8232,
+      "step": 25465
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7436221155836953e-06,
+      "loss": 0.9189,
+      "step": 25466
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7430374118905136e-06,
+      "loss": 0.7168,
+      "step": 25467
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7424527968906434e-06,
+      "loss": 1.2148,
+      "step": 25468
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7418682705903645e-06,
+      "loss": 0.9316,
+      "step": 25469
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.741283832995956e-06,
+      "loss": 1.252,
+      "step": 25470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7406994841136993e-06,
+      "loss": 0.8428,
+      "step": 25471
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7401152239498697e-06,
+      "loss": 1.2188,
+      "step": 25472
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7395310525107389e-06,
+      "loss": 1.0488,
+      "step": 25473
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7389469698025851e-06,
+      "loss": 0.9189,
+      "step": 25474
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7383629758316856e-06,
+      "loss": 0.7822,
+      "step": 25475
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7377790706043095e-06,
+      "loss": 0.8965,
+      "step": 25476
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7371952541267267e-06,
+      "loss": 1.0645,
+      "step": 25477
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7366115264052141e-06,
+      "loss": 1.0938,
+      "step": 25478
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7360278874460368e-06,
+      "loss": 0.9526,
+      "step": 25479
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7354443372554674e-06,
+      "loss": 0.7888,
+      "step": 25480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7348608758397723e-06,
+      "loss": 1.0781,
+      "step": 25481
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7342775032052205e-06,
+      "loss": 1.2031,
+      "step": 25482
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7336942193580786e-06,
+      "loss": 0.7617,
+      "step": 25483
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7331110243046078e-06,
+      "loss": 1.1807,
+      "step": 25484
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7325279180510756e-06,
+      "loss": 0.9795,
+      "step": 25485
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.731944900603748e-06,
+      "loss": 1.1304,
+      "step": 25486
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7313619719688857e-06,
+      "loss": 1.0391,
+      "step": 25487
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7307791321527466e-06,
+      "loss": 1.3496,
+      "step": 25488
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.730196381161594e-06,
+      "loss": 0.9746,
+      "step": 25489
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7296137190016916e-06,
+      "loss": 1.1836,
+      "step": 25490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.729031145679294e-06,
+      "loss": 0.8887,
+      "step": 25491
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7284486612006569e-06,
+      "loss": 1.0996,
+      "step": 25492
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7278662655720424e-06,
+      "loss": 0.9209,
+      "step": 25493
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7272839587997048e-06,
+      "loss": 1.0967,
+      "step": 25494
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7267017408898946e-06,
+      "loss": 0.9932,
+      "step": 25495
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.72611961184887e-06,
+      "loss": 0.6787,
+      "step": 25496
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7255375716828848e-06,
+      "loss": 0.6357,
+      "step": 25497
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.72495562039819e-06,
+      "loss": 1.2793,
+      "step": 25498
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7243737580010334e-06,
+      "loss": 1.0146,
+      "step": 25499
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7237919844976681e-06,
+      "loss": 0.8506,
+      "step": 25500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7232102998943457e-06,
+      "loss": 1.0996,
+      "step": 25501
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7226287041973123e-06,
+      "loss": 0.9316,
+      "step": 25502
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7220471974128117e-06,
+      "loss": 1.4521,
+      "step": 25503
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.721465779547097e-06,
+      "loss": 0.7129,
+      "step": 25504
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.720884450606406e-06,
+      "loss": 1.0615,
+      "step": 25505
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7203032105969897e-06,
+      "loss": 0.8066,
+      "step": 25506
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7197220595250875e-06,
+      "loss": 0.9253,
+      "step": 25507
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7191409973969452e-06,
+      "loss": 0.9336,
+      "step": 25508
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7185600242188028e-06,
+      "loss": 1.1665,
+      "step": 25509
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7179791399968948e-06,
+      "loss": 0.999,
+      "step": 25510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.717398344737472e-06,
+      "loss": 0.825,
+      "step": 25511
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7168176384467682e-06,
+      "loss": 1.0459,
+      "step": 25512
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7162370211310209e-06,
+      "loss": 0.7949,
+      "step": 25513
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7156564927964637e-06,
+      "loss": 1.1953,
+      "step": 25514
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7150760534493348e-06,
+      "loss": 0.8838,
+      "step": 25515
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7144957030958732e-06,
+      "loss": 0.814,
+      "step": 25516
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7139154417423064e-06,
+      "loss": 0.9952,
+      "step": 25517
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7133352693948725e-06,
+      "loss": 1.1768,
+      "step": 25518
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7127551860598014e-06,
+      "loss": 1.002,
+      "step": 25519
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7121751917433216e-06,
+      "loss": 1.2129,
+      "step": 25520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.711595286451665e-06,
+      "loss": 0.6987,
+      "step": 25521
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7110154701910642e-06,
+      "loss": 0.8774,
+      "step": 25522
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7104357429677443e-06,
+      "loss": 0.9805,
+      "step": 25523
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7098561047879337e-06,
+      "loss": 0.5581,
+      "step": 25524
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.709276555657854e-06,
+      "loss": 0.9326,
+      "step": 25525
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7086970955837344e-06,
+      "loss": 1.3203,
+      "step": 25526
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7081177245718018e-06,
+      "loss": 1.0811,
+      "step": 25527
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.707538442628275e-06,
+      "loss": 1.0381,
+      "step": 25528
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.70695924975938e-06,
+      "loss": 0.8711,
+      "step": 25529
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7063801459713358e-06,
+      "loss": 1.1045,
+      "step": 25530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7058011312703626e-06,
+      "loss": 0.9795,
+      "step": 25531
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.70522220566268e-06,
+      "loss": 1.3379,
+      "step": 25532
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7046433691545105e-06,
+      "loss": 1.0752,
+      "step": 25533
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7040646217520685e-06,
+      "loss": 0.9863,
+      "step": 25534
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7034859634615697e-06,
+      "loss": 1.0195,
+      "step": 25535
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7029073942892304e-06,
+      "loss": 0.5407,
+      "step": 25536
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7023289142412691e-06,
+      "loss": 1.0098,
+      "step": 25537
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7017505233238962e-06,
+      "loss": 1.0889,
+      "step": 25538
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7011722215433223e-06,
+      "loss": 0.8389,
+      "step": 25539
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.700594008905766e-06,
+      "loss": 0.6284,
+      "step": 25540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.700015885417431e-06,
+      "loss": 1.2607,
+      "step": 25541
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6994378510845332e-06,
+      "loss": 0.9561,
+      "step": 25542
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.698859905913276e-06,
+      "loss": 0.8379,
+      "step": 25543
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.698282049909874e-06,
+      "loss": 0.9336,
+      "step": 25544
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6977042830805302e-06,
+      "loss": 1.2217,
+      "step": 25545
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6971266054314494e-06,
+      "loss": 1.0986,
+      "step": 25546
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6965490169688392e-06,
+      "loss": 1.334,
+      "step": 25547
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6959715176989056e-06,
+      "loss": 0.9346,
+      "step": 25548
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.695394107627849e-06,
+      "loss": 1.1201,
+      "step": 25549
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6948167867618714e-06,
+      "loss": 1.1367,
+      "step": 25550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6942395551071745e-06,
+      "loss": 1.0742,
+      "step": 25551
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.693662412669962e-06,
+      "loss": 1.1465,
+      "step": 25552
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6930853594564312e-06,
+      "loss": 0.7891,
+      "step": 25553
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6925083954727773e-06,
+      "loss": 0.9941,
+      "step": 25554
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.691931520725204e-06,
+      "loss": 1.2881,
+      "step": 25555
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6913547352199044e-06,
+      "loss": 0.7192,
+      "step": 25556
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6907780389630713e-06,
+      "loss": 1.1758,
+      "step": 25557
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.690201431960903e-06,
+      "loss": 0.9756,
+      "step": 25558
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6896249142195952e-06,
+      "loss": 1.0869,
+      "step": 25559
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6890484857453382e-06,
+      "loss": 1.0234,
+      "step": 25560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.688472146544321e-06,
+      "loss": 1.1006,
+      "step": 25561
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6878958966227366e-06,
+      "loss": 0.5918,
+      "step": 25562
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6873197359867788e-06,
+      "loss": 1.0156,
+      "step": 25563
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6867436646426328e-06,
+      "loss": 0.8403,
+      "step": 25564
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6861676825964845e-06,
+      "loss": 1.3086,
+      "step": 25565
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.685591789854526e-06,
+      "loss": 1.0605,
+      "step": 25566
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6850159864229377e-06,
+      "loss": 0.6948,
+      "step": 25567
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6844402723079101e-06,
+      "loss": 0.875,
+      "step": 25568
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6838646475156216e-06,
+      "loss": 0.8975,
+      "step": 25569
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.683289112052262e-06,
+      "loss": 1.2188,
+      "step": 25570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6827136659240095e-06,
+      "loss": 1.2314,
+      "step": 25571
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6821383091370435e-06,
+      "loss": 1.1748,
+      "step": 25572
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.681563041697547e-06,
+      "loss": 1.4082,
+      "step": 25573
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6809878636117005e-06,
+      "loss": 0.7983,
+      "step": 25574
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6804127748856814e-06,
+      "loss": 0.7344,
+      "step": 25575
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6798377755256645e-06,
+      "loss": 0.8828,
+      "step": 25576
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6792628655378274e-06,
+      "loss": 0.8433,
+      "step": 25577
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6786880449283493e-06,
+      "loss": 0.7852,
+      "step": 25578
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6781133137034022e-06,
+      "loss": 0.6304,
+      "step": 25579
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6775386718691566e-06,
+      "loss": 0.4241,
+      "step": 25580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6769641194317898e-06,
+      "loss": 1.0205,
+      "step": 25581
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6763896563974714e-06,
+      "loss": 0.7598,
+      "step": 25582
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6758152827723694e-06,
+      "loss": 1.0225,
+      "step": 25583
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6752409985626572e-06,
+      "loss": 0.7158,
+      "step": 25584
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.674666803774504e-06,
+      "loss": 0.7358,
+      "step": 25585
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6740926984140771e-06,
+      "loss": 1.0127,
+      "step": 25586
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6735186824875394e-06,
+      "loss": 0.8525,
+      "step": 25587
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6729447560010593e-06,
+      "loss": 0.9814,
+      "step": 25588
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.672370918960805e-06,
+      "loss": 0.9639,
+      "step": 25589
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6717971713729374e-06,
+      "loss": 0.9272,
+      "step": 25590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6712235132436184e-06,
+      "loss": 1.4316,
+      "step": 25591
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6706499445790125e-06,
+      "loss": 0.752,
+      "step": 25592
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6700764653852774e-06,
+      "loss": 0.9932,
+      "step": 25593
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6695030756685782e-06,
+      "loss": 1.1055,
+      "step": 25594
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6689297754350687e-06,
+      "loss": 1.1445,
+      "step": 25595
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6683565646909116e-06,
+      "loss": 0.9043,
+      "step": 25596
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6677834434422625e-06,
+      "loss": 0.9805,
+      "step": 25597
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6672104116952748e-06,
+      "loss": 0.8159,
+      "step": 25598
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6666374694561073e-06,
+      "loss": 0.918,
+      "step": 25599
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6660646167309147e-06,
+      "loss": 1.2773,
+      "step": 25600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.665491853525849e-06,
+      "loss": 1.459,
+      "step": 25601
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.664919179847061e-06,
+      "loss": 1.3613,
+      "step": 25602
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6643465957007032e-06,
+      "loss": 0.835,
+      "step": 25603
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6637741010929286e-06,
+      "loss": 1.4238,
+      "step": 25604
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6632016960298859e-06,
+      "loss": 0.8457,
+      "step": 25605
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6626293805177197e-06,
+      "loss": 1.2422,
+      "step": 25606
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.662057154562583e-06,
+      "loss": 1.3193,
+      "step": 25607
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6614850181706165e-06,
+      "loss": 1.0078,
+      "step": 25608
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.660912971347971e-06,
+      "loss": 1.1865,
+      "step": 25609
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6603410141007913e-06,
+      "loss": 1.1006,
+      "step": 25610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6597691464352196e-06,
+      "loss": 0.7246,
+      "step": 25611
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.659197368357398e-06,
+      "loss": 0.7646,
+      "step": 25612
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6586256798734667e-06,
+      "loss": 0.999,
+      "step": 25613
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6580540809895684e-06,
+      "loss": 0.5459,
+      "step": 25614
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6574825717118459e-06,
+      "loss": 0.8643,
+      "step": 25615
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6569111520464354e-06,
+      "loss": 1.0718,
+      "step": 25616
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656339821999473e-06,
+      "loss": 0.8916,
+      "step": 25617
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6557685815771006e-06,
+      "loss": 0.5168,
+      "step": 25618
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655197430785448e-06,
+      "loss": 0.9219,
+      "step": 25619
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6546263696306542e-06,
+      "loss": 0.8604,
+      "step": 25620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6540553981188556e-06,
+      "loss": 0.8608,
+      "step": 25621
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6534845162561818e-06,
+      "loss": 0.7432,
+      "step": 25622
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652913724048767e-06,
+      "loss": 1.0059,
+      "step": 25623
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6523430215027358e-06,
+      "loss": 0.9658,
+      "step": 25624
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6517724086242282e-06,
+      "loss": 0.9414,
+      "step": 25625
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6512018854193702e-06,
+      "loss": 0.8511,
+      "step": 25626
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6506314518942868e-06,
+      "loss": 1.0005,
+      "step": 25627
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.65006110805511e-06,
+      "loss": 0.7939,
+      "step": 25628
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6494908539079613e-06,
+      "loss": 1.0537,
+      "step": 25629
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6489206894589716e-06,
+      "loss": 1.1182,
+      "step": 25630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6483506147142603e-06,
+      "loss": 0.9502,
+      "step": 25631
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.647780629679956e-06,
+      "loss": 0.8223,
+      "step": 25632
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.647210734362178e-06,
+      "loss": 0.4502,
+      "step": 25633
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.646640928767047e-06,
+      "loss": 0.9277,
+      "step": 25634
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.646071212900684e-06,
+      "loss": 0.9629,
+      "step": 25635
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6455015867692137e-06,
+      "loss": 0.7583,
+      "step": 25636
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6449320503787503e-06,
+      "loss": 1.373,
+      "step": 25637
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6443626037354089e-06,
+      "loss": 0.7803,
+      "step": 25638
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.643793246845311e-06,
+      "loss": 1.0449,
+      "step": 25639
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6432239797145734e-06,
+      "loss": 1.0859,
+      "step": 25640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6426548023493072e-06,
+      "loss": 1.1143,
+      "step": 25641
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6420857147556268e-06,
+      "loss": 0.8379,
+      "step": 25642
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6415167169396473e-06,
+      "loss": 0.8252,
+      "step": 25643
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6409478089074804e-06,
+      "loss": 0.8545,
+      "step": 25644
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6403789906652334e-06,
+      "loss": 0.7617,
+      "step": 25645
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6398102622190193e-06,
+      "loss": 1.0024,
+      "step": 25646
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6392416235749486e-06,
+      "loss": 1.0449,
+      "step": 25647
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6386730747391288e-06,
+      "loss": 0.9629,
+      "step": 25648
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6381046157176628e-06,
+      "loss": 1.2363,
+      "step": 25649
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.63753624651666e-06,
+      "loss": 0.9586,
+      "step": 25650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6369679671422277e-06,
+      "loss": 0.8379,
+      "step": 25651
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6363997776004691e-06,
+      "loss": 1.3379,
+      "step": 25652
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6358316778974836e-06,
+      "loss": 1.2227,
+      "step": 25653
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6352636680393785e-06,
+      "loss": 0.8545,
+      "step": 25654
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6346957480322512e-06,
+      "loss": 0.7559,
+      "step": 25655
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6341279178822057e-06,
+      "loss": 1.4492,
+      "step": 25656
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6335601775953369e-06,
+      "loss": 1.2275,
+      "step": 25657
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6329925271777492e-06,
+      "loss": 0.9692,
+      "step": 25658
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6324249666355362e-06,
+      "loss": 0.6221,
+      "step": 25659
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6318574959747935e-06,
+      "loss": 1.0996,
+      "step": 25660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.631290115201617e-06,
+      "loss": 0.9268,
+      "step": 25661
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6307228243221062e-06,
+      "loss": 1.3057,
+      "step": 25662
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.630155623342351e-06,
+      "loss": 0.9209,
+      "step": 25663
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6295885122684407e-06,
+      "loss": 0.9697,
+      "step": 25664
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6290214911064717e-06,
+      "loss": 1.0078,
+      "step": 25665
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6284545598625357e-06,
+      "loss": 0.8018,
+      "step": 25666
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.62788771854272e-06,
+      "loss": 1.2656,
+      "step": 25667
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.627320967153111e-06,
+      "loss": 0.9517,
+      "step": 25668
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6267543056998025e-06,
+      "loss": 0.8701,
+      "step": 25669
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6261877341888754e-06,
+      "loss": 1.2598,
+      "step": 25670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6256212526264204e-06,
+      "loss": 1.0938,
+      "step": 25671
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.625054861018518e-06,
+      "loss": 1.041,
+      "step": 25672
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6244885593712566e-06,
+      "loss": 0.9648,
+      "step": 25673
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6239223476907184e-06,
+      "loss": 0.8398,
+      "step": 25674
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6233562259829806e-06,
+      "loss": 0.9619,
+      "step": 25675
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.622790194254128e-06,
+      "loss": 0.7617,
+      "step": 25676
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6222242525102437e-06,
+      "loss": 0.9629,
+      "step": 25677
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6216584007574033e-06,
+      "loss": 1.2388,
+      "step": 25678
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6210926390016823e-06,
+      "loss": 0.7563,
+      "step": 25679
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6205269672491641e-06,
+      "loss": 1.0278,
+      "step": 25680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6199613855059194e-06,
+      "loss": 0.9858,
+      "step": 25681
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6193958937780285e-06,
+      "loss": 1.1416,
+      "step": 25682
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6188304920715602e-06,
+      "loss": 1.126,
+      "step": 25683
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6182651803925942e-06,
+      "loss": 1.0869,
+      "step": 25684
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6176999587471987e-06,
+      "loss": 0.8955,
+      "step": 25685
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6171348271414432e-06,
+      "loss": 1.0586,
+      "step": 25686
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6165697855814012e-06,
+      "loss": 1.1514,
+      "step": 25687
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6160048340731439e-06,
+      "loss": 0.7559,
+      "step": 25688
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6154399726227377e-06,
+      "loss": 0.6963,
+      "step": 25689
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6148752012362467e-06,
+      "loss": 0.8228,
+      "step": 25690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6143105199197417e-06,
+      "loss": 1.0688,
+      "step": 25691
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.61374592867929e-06,
+      "loss": 0.9385,
+      "step": 25692
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6131814275209546e-06,
+      "loss": 0.7329,
+      "step": 25693
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6126170164507949e-06,
+      "loss": 0.9209,
+      "step": 25694
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6120526954748795e-06,
+      "loss": 0.8687,
+      "step": 25695
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6114884645992646e-06,
+      "loss": 0.8359,
+      "step": 25696
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6109243238300177e-06,
+      "loss": 1.002,
+      "step": 25697
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6103602731731916e-06,
+      "loss": 1.0654,
+      "step": 25698
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6097963126348503e-06,
+      "loss": 1.083,
+      "step": 25699
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6092324422210503e-06,
+      "loss": 1.1221,
+      "step": 25700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6086686619378432e-06,
+      "loss": 1.0635,
+      "step": 25701
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6081049717912944e-06,
+      "loss": 1.1514,
+      "step": 25702
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6075413717874545e-06,
+      "loss": 0.6768,
+      "step": 25703
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.606977861932376e-06,
+      "loss": 0.9424,
+      "step": 25704
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6064144422321114e-06,
+      "loss": 1.0342,
+      "step": 25705
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6058511126927178e-06,
+      "loss": 1.186,
+      "step": 25706
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6052878733202393e-06,
+      "loss": 0.8823,
+      "step": 25707
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6047247241207297e-06,
+      "loss": 0.8779,
+      "step": 25708
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.60416166510024e-06,
+      "loss": 0.8442,
+      "step": 25709
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6035986962648176e-06,
+      "loss": 1.2334,
+      "step": 25710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6030358176205075e-06,
+      "loss": 0.8438,
+      "step": 25711
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6024730291733525e-06,
+      "loss": 1.0869,
+      "step": 25712
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.601910330929407e-06,
+      "loss": 0.8174,
+      "step": 25713
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6013477228947104e-06,
+      "loss": 1.4092,
+      "step": 25714
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.600785205075307e-06,
+      "loss": 0.6475,
+      "step": 25715
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.600222777477235e-06,
+      "loss": 1.0303,
+      "step": 25716
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5996604401065396e-06,
+      "loss": 0.9033,
+      "step": 25717
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.599098192969264e-06,
+      "loss": 0.8545,
+      "step": 25718
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5985360360714418e-06,
+      "loss": 0.8828,
+      "step": 25719
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5979739694191165e-06,
+      "loss": 0.7891,
+      "step": 25720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.597411993018323e-06,
+      "loss": 0.9951,
+      "step": 25721
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5968501068750953e-06,
+      "loss": 1.2637,
+      "step": 25722
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.596288310995473e-06,
+      "loss": 0.8828,
+      "step": 25723
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5957266053854913e-06,
+      "loss": 1.3594,
+      "step": 25724
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5951649900511812e-06,
+      "loss": 0.999,
+      "step": 25725
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5946034649985754e-06,
+      "loss": 0.873,
+      "step": 25726
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5940420302337056e-06,
+      "loss": 0.7539,
+      "step": 25727
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5934806857626062e-06,
+      "loss": 0.7949,
+      "step": 25728
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5929194315913032e-06,
+      "loss": 1.1035,
+      "step": 25729
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5923582677258254e-06,
+      "loss": 0.595,
+      "step": 25730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5917971941722032e-06,
+      "loss": 0.8071,
+      "step": 25731
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5912362109364588e-06,
+      "loss": 0.9292,
+      "step": 25732
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5906753180246238e-06,
+      "loss": 0.813,
+      "step": 25733
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.590114515442719e-06,
+      "loss": 0.8491,
+      "step": 25734
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5895538031967706e-06,
+      "loss": 0.7285,
+      "step": 25735
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5889931812928016e-06,
+      "loss": 0.9062,
+      "step": 25736
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5884326497368309e-06,
+      "loss": 1.0684,
+      "step": 25737
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5878722085348808e-06,
+      "loss": 1.0439,
+      "step": 25738
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5873118576929748e-06,
+      "loss": 0.8696,
+      "step": 25739
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.58675159721713e-06,
+      "loss": 0.938,
+      "step": 25740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5861914271133617e-06,
+      "loss": 0.9102,
+      "step": 25741
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5856313473876916e-06,
+      "loss": 1.1523,
+      "step": 25742
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5850713580461308e-06,
+      "loss": 1.0493,
+      "step": 25743
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5845114590946998e-06,
+      "loss": 0.9199,
+      "step": 25744
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.583951650539408e-06,
+      "loss": 1.1416,
+      "step": 25745
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5833919323862734e-06,
+      "loss": 0.7559,
+      "step": 25746
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5828323046413051e-06,
+      "loss": 1.1074,
+      "step": 25747
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5822727673105143e-06,
+      "loss": 1.3564,
+      "step": 25748
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5817133203999112e-06,
+      "loss": 1.0059,
+      "step": 25749
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5811539639155094e-06,
+      "loss": 1.0918,
+      "step": 25750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5805946978633136e-06,
+      "loss": 0.6919,
+      "step": 25751
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.580035522249329e-06,
+      "loss": 0.9502,
+      "step": 25752
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5794764370795657e-06,
+      "loss": 0.7607,
+      "step": 25753
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5789174423600307e-06,
+      "loss": 0.9336,
+      "step": 25754
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5783585380967248e-06,
+      "loss": 1.0195,
+      "step": 25755
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5777997242956522e-06,
+      "loss": 1.0049,
+      "step": 25756
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.577241000962817e-06,
+      "loss": 1.126,
+      "step": 25757
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5766823681042188e-06,
+      "loss": 0.7324,
+      "step": 25758
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5761238257258615e-06,
+      "loss": 1.1309,
+      "step": 25759
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5755653738337395e-06,
+      "loss": 0.7949,
+      "step": 25760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5750070124338579e-06,
+      "loss": 1.0039,
+      "step": 25761
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5744487415322107e-06,
+      "loss": 1.2627,
+      "step": 25762
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5738905611347933e-06,
+      "loss": 0.9902,
+      "step": 25763
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5733324712476027e-06,
+      "loss": 1.1455,
+      "step": 25764
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.572774471876637e-06,
+      "loss": 0.9639,
+      "step": 25765
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5722165630278873e-06,
+      "loss": 1.082,
+      "step": 25766
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5716587447073438e-06,
+      "loss": 1.25,
+      "step": 25767
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5711010169210039e-06,
+      "loss": 0.6855,
+      "step": 25768
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5705433796748525e-06,
+      "loss": 1.1292,
+      "step": 25769
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5699858329748852e-06,
+      "loss": 0.8623,
+      "step": 25770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5694283768270868e-06,
+      "loss": 1.0308,
+      "step": 25771
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5688710112374495e-06,
+      "loss": 0.9336,
+      "step": 25772
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5683137362119572e-06,
+      "loss": 0.8274,
+      "step": 25773
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.567756551756594e-06,
+      "loss": 0.719,
+      "step": 25774
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5671994578773475e-06,
+      "loss": 0.9668,
+      "step": 25775
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5666424545802038e-06,
+      "loss": 1.2236,
+      "step": 25776
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.566085541871145e-06,
+      "loss": 1.1904,
+      "step": 25777
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5655287197561497e-06,
+      "loss": 0.7886,
+      "step": 25778
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5649719882412008e-06,
+      "loss": 0.855,
+      "step": 25779
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.564415347332283e-06,
+      "loss": 0.8086,
+      "step": 25780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5638587970353714e-06,
+      "loss": 1.3828,
+      "step": 25781
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5633023373564415e-06,
+      "loss": 1.1963,
+      "step": 25782
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5627459683014778e-06,
+      "loss": 1.1611,
+      "step": 25783
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5621896898764487e-06,
+      "loss": 1.0391,
+      "step": 25784
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5616335020873373e-06,
+      "loss": 0.877,
+      "step": 25785
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5610774049401112e-06,
+      "loss": 0.8853,
+      "step": 25786
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5605213984407496e-06,
+      "loss": 1.3604,
+      "step": 25787
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5599654825952215e-06,
+      "loss": 1.1348,
+      "step": 25788
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.559409657409494e-06,
+      "loss": 0.957,
+      "step": 25789
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5588539228895482e-06,
+      "loss": 0.7363,
+      "step": 25790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.558298279041347e-06,
+      "loss": 1.0664,
+      "step": 25791
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.55774272587086e-06,
+      "loss": 0.9189,
+      "step": 25792
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5571872633840534e-06,
+      "loss": 1.0293,
+      "step": 25793
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.556631891586894e-06,
+      "loss": 0.7261,
+      "step": 25794
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5560766104853508e-06,
+      "loss": 1.0186,
+      "step": 25795
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5555214200853851e-06,
+      "loss": 1.0186,
+      "step": 25796
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.554966320392961e-06,
+      "loss": 0.7876,
+      "step": 25797
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5544113114140424e-06,
+      "loss": 1.1338,
+      "step": 25798
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5538563931545913e-06,
+      "loss": 0.9053,
+      "step": 25799
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5533015656205618e-06,
+      "loss": 1.2002,
+      "step": 25800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5527468288179249e-06,
+      "loss": 1.0186,
+      "step": 25801
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5521921827526332e-06,
+      "loss": 0.7754,
+      "step": 25802
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5516376274306444e-06,
+      "loss": 0.8301,
+      "step": 25803
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5510831628579148e-06,
+      "loss": 1.3887,
+      "step": 25804
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5505287890404009e-06,
+      "loss": 1.0708,
+      "step": 25805
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.549974505984061e-06,
+      "loss": 1.1221,
+      "step": 25806
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5494203136948428e-06,
+      "loss": 0.7959,
+      "step": 25807
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5488662121787057e-06,
+      "loss": 0.8306,
+      "step": 25808
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5483122014415974e-06,
+      "loss": 0.7998,
+      "step": 25809
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5477582814894688e-06,
+      "loss": 1.1943,
+      "step": 25810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5472044523282704e-06,
+      "loss": 0.7124,
+      "step": 25811
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5466507139639552e-06,
+      "loss": 0.8462,
+      "step": 25812
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5460970664024665e-06,
+      "loss": 0.7036,
+      "step": 25813
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.545543509649754e-06,
+      "loss": 0.9883,
+      "step": 25814
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5449900437117582e-06,
+      "loss": 0.8438,
+      "step": 25815
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5444366685944323e-06,
+      "loss": 1.082,
+      "step": 25816
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5438833843037183e-06,
+      "loss": 1.1484,
+      "step": 25817
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5433301908455544e-06,
+      "loss": 0.9033,
+      "step": 25818
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5427770882258886e-06,
+      "loss": 0.7842,
+      "step": 25819
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.542224076450658e-06,
+      "loss": 1.0361,
+      "step": 25820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5416711555258078e-06,
+      "loss": 0.8611,
+      "step": 25821
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5411183254572714e-06,
+      "loss": 0.9448,
+      "step": 25822
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5405655862509915e-06,
+      "loss": 1.0981,
+      "step": 25823
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5400129379129047e-06,
+      "loss": 1.0518,
+      "step": 25824
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5394603804489439e-06,
+      "loss": 1.1572,
+      "step": 25825
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5389079138650464e-06,
+      "loss": 1.1914,
+      "step": 25826
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.53835553816715e-06,
+      "loss": 0.7549,
+      "step": 25827
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5378032533611865e-06,
+      "loss": 1.1416,
+      "step": 25828
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5372510594530831e-06,
+      "loss": 0.8306,
+      "step": 25829
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5366989564487789e-06,
+      "loss": 0.9521,
+      "step": 25830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5361469443541987e-06,
+      "loss": 1.1973,
+      "step": 25831
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.535595023175276e-06,
+      "loss": 0.741,
+      "step": 25832
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5350431929179344e-06,
+      "loss": 0.8071,
+      "step": 25833
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5344914535881084e-06,
+      "loss": 1.3232,
+      "step": 25834
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5339398051917199e-06,
+      "loss": 0.5215,
+      "step": 25835
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.533388247734693e-06,
+      "loss": 0.8613,
+      "step": 25836
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.532836781222955e-06,
+      "loss": 1.2598,
+      "step": 25837
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5322854056624315e-06,
+      "loss": 0.6797,
+      "step": 25838
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5317341210590432e-06,
+      "loss": 1.2949,
+      "step": 25839
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5311829274187096e-06,
+      "loss": 0.9648,
+      "step": 25840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5306318247473529e-06,
+      "loss": 0.835,
+      "step": 25841
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5300808130508948e-06,
+      "loss": 0.8721,
+      "step": 25842
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.529529892335254e-06,
+      "loss": 0.9443,
+      "step": 25843
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5289790626063438e-06,
+      "loss": 0.9717,
+      "step": 25844
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.528428323870087e-06,
+      "loss": 1.123,
+      "step": 25845
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5278776761323933e-06,
+      "loss": 0.8467,
+      "step": 25846
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5273271193991835e-06,
+      "loss": 1.1182,
+      "step": 25847
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5267766536763673e-06,
+      "loss": 0.7764,
+      "step": 25848
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.52622627896986e-06,
+      "loss": 0.9424,
+      "step": 25849
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5256759952855737e-06,
+      "loss": 0.9902,
+      "step": 25850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5251258026294147e-06,
+      "loss": 0.6584,
+      "step": 25851
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.524575701007298e-06,
+      "loss": 1.0889,
+      "step": 25852
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5240256904251328e-06,
+      "loss": 1.1104,
+      "step": 25853
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5234757708888248e-06,
+      "loss": 1.0459,
+      "step": 25854
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5229259424042797e-06,
+      "loss": 0.9785,
+      "step": 25855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.522376204977406e-06,
+      "loss": 0.9746,
+      "step": 25856
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.52182655861411e-06,
+      "loss": 0.8223,
+      "step": 25857
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.521277003320295e-06,
+      "loss": 0.5581,
+      "step": 25858
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5207275391018605e-06,
+      "loss": 1.0801,
+      "step": 25859
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.520178165964713e-06,
+      "loss": 0.8438,
+      "step": 25860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5196288839147533e-06,
+      "loss": 0.9561,
+      "step": 25861
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5190796929578778e-06,
+      "loss": 0.835,
+      "step": 25862
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5185305930999872e-06,
+      "loss": 1.1582,
+      "step": 25863
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5179815843469837e-06,
+      "loss": 0.7158,
+      "step": 25864
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5174326667047623e-06,
+      "loss": 1.1934,
+      "step": 25865
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.516883840179215e-06,
+      "loss": 0.9492,
+      "step": 25866
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.516335104776242e-06,
+      "loss": 0.751,
+      "step": 25867
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.515786460501738e-06,
+      "loss": 1.29,
+      "step": 25868
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5152379073615942e-06,
+      "loss": 0.9707,
+      "step": 25869
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.514689445361701e-06,
+      "loss": 0.9512,
+      "step": 25870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5141410745079555e-06,
+      "loss": 1.3389,
+      "step": 25871
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5135927948062423e-06,
+      "loss": 1.0347,
+      "step": 25872
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5130446062624559e-06,
+      "loss": 0.9678,
+      "step": 25873
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5124965088824795e-06,
+      "loss": 0.8416,
+      "step": 25874
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5119485026722048e-06,
+      "loss": 1.2334,
+      "step": 25875
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.511400587637517e-06,
+      "loss": 0.999,
+      "step": 25876
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5108527637842996e-06,
+      "loss": 0.7061,
+      "step": 25877
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5103050311184386e-06,
+      "loss": 0.8916,
+      "step": 25878
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5097573896458207e-06,
+      "loss": 1.0176,
+      "step": 25879
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5092098393723253e-06,
+      "loss": 1.2939,
+      "step": 25880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5086623803038313e-06,
+      "loss": 0.8672,
+      "step": 25881
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.508115012446223e-06,
+      "loss": 1.2051,
+      "step": 25882
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5075677358053809e-06,
+      "loss": 1.0615,
+      "step": 25883
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5070205503871827e-06,
+      "loss": 1.0762,
+      "step": 25884
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5064734561975024e-06,
+      "loss": 1.083,
+      "step": 25885
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5059264532422224e-06,
+      "loss": 1.0039,
+      "step": 25886
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5053795415272132e-06,
+      "loss": 1.127,
+      "step": 25887
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5048327210583525e-06,
+      "loss": 0.9453,
+      "step": 25888
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5042859918415154e-06,
+      "loss": 0.8564,
+      "step": 25889
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5037393538825728e-06,
+      "loss": 0.5786,
+      "step": 25890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5031928071873968e-06,
+      "loss": 0.6692,
+      "step": 25891
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.502646351761856e-06,
+      "loss": 0.7773,
+      "step": 25892
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5020999876118225e-06,
+      "loss": 0.7842,
+      "step": 25893
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5015537147431657e-06,
+      "loss": 1.2832,
+      "step": 25894
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5010075331617535e-06,
+      "loss": 1.124,
+      "step": 25895
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5004614428734498e-06,
+      "loss": 0.8887,
+      "step": 25896
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4999154438841246e-06,
+      "loss": 0.998,
+      "step": 25897
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4993695361996374e-06,
+      "loss": 0.7827,
+      "step": 25898
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.498823719825857e-06,
+      "loss": 0.8301,
+      "step": 25899
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4982779947686465e-06,
+      "loss": 0.5864,
+      "step": 25900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4977323610338668e-06,
+      "loss": 0.8291,
+      "step": 25901
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4971868186273785e-06,
+      "loss": 0.8506,
+      "step": 25902
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4966413675550361e-06,
+      "loss": 0.8501,
+      "step": 25903
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4960960078227093e-06,
+      "loss": 1.0156,
+      "step": 25904
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.49555073943625e-06,
+      "loss": 0.8936,
+      "step": 25905
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4950055624015147e-06,
+      "loss": 0.6733,
+      "step": 25906
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4944604767243631e-06,
+      "loss": 0.8335,
+      "step": 25907
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.493915482410646e-06,
+      "loss": 0.9429,
+      "step": 25908
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4933705794662223e-06,
+      "loss": 0.645,
+      "step": 25909
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4928257678969394e-06,
+      "loss": 0.7666,
+      "step": 25910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4922810477086558e-06,
+      "loss": 1.1104,
+      "step": 25911
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.491736418907219e-06,
+      "loss": 0.7112,
+      "step": 25912
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4911918814984784e-06,
+      "loss": 1.0176,
+      "step": 25913
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4906474354882827e-06,
+      "loss": 0.667,
+      "step": 25914
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4901030808824856e-06,
+      "loss": 1.2637,
+      "step": 25915
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4895588176869313e-06,
+      "loss": 0.833,
+      "step": 25916
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4890146459074617e-06,
+      "loss": 1.1973,
+      "step": 25917
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4884705655499299e-06,
+      "loss": 1.0078,
+      "step": 25918
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4879265766201724e-06,
+      "loss": 1.3486,
+      "step": 25919
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4873826791240398e-06,
+      "loss": 1.0225,
+      "step": 25920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4868388730673677e-06,
+      "loss": 0.7266,
+      "step": 25921
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4862951584560037e-06,
+      "loss": 1.001,
+      "step": 25922
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4857515352957853e-06,
+      "loss": 0.7568,
+      "step": 25923
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4852080035925498e-06,
+      "loss": 1.126,
+      "step": 25924
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4846645633521373e-06,
+      "loss": 1.1211,
+      "step": 25925
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4841212145803884e-06,
+      "loss": 0.8311,
+      "step": 25926
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4835779572831365e-06,
+      "loss": 1.2163,
+      "step": 25927
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4830347914662157e-06,
+      "loss": 0.9873,
+      "step": 25928
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4824917171354614e-06,
+      "loss": 0.9463,
+      "step": 25929
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4819487342967098e-06,
+      "loss": 0.959,
+      "step": 25930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4814058429557932e-06,
+      "loss": 1.3125,
+      "step": 25931
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4808630431185378e-06,
+      "loss": 0.9238,
+      "step": 25932
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4803203347907801e-06,
+      "loss": 0.8066,
+      "step": 25933
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4797777179783457e-06,
+      "loss": 0.96,
+      "step": 25934
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4792351926870663e-06,
+      "loss": 0.9561,
+      "step": 25935
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4786927589227663e-06,
+      "loss": 0.7109,
+      "step": 25936
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4781504166912774e-06,
+      "loss": 1.3262,
+      "step": 25937
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.477608165998421e-06,
+      "loss": 1.3135,
+      "step": 25938
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4770660068500197e-06,
+      "loss": 0.8704,
+      "step": 25939
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4765239392519015e-06,
+      "loss": 0.9707,
+      "step": 25940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4759819632098893e-06,
+      "loss": 1.1826,
+      "step": 25941
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.475440078729804e-06,
+      "loss": 1.1133,
+      "step": 25942
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4748982858174631e-06,
+      "loss": 1.1943,
+      "step": 25943
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4743565844786889e-06,
+      "loss": 0.791,
+      "step": 25944
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.473814974719302e-06,
+      "loss": 1.041,
+      "step": 25945
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.473273456545119e-06,
+      "loss": 1.0479,
+      "step": 25946
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.472732029961953e-06,
+      "loss": 0.9805,
+      "step": 25947
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4721906949756248e-06,
+      "loss": 0.627,
+      "step": 25948
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.471649451591949e-06,
+      "loss": 0.8774,
+      "step": 25949
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4711082998167336e-06,
+      "loss": 1.2266,
+      "step": 25950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4705672396557968e-06,
+      "loss": 1.019,
+      "step": 25951
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4700262711149504e-06,
+      "loss": 1.0361,
+      "step": 25952
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4694853942000053e-06,
+      "loss": 0.8535,
+      "step": 25953
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.468944608916767e-06,
+      "loss": 0.8545,
+      "step": 25954
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4684039152710483e-06,
+      "loss": 0.77,
+      "step": 25955
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.467863313268658e-06,
+      "loss": 0.7285,
+      "step": 25956
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4673228029154019e-06,
+      "loss": 0.835,
+      "step": 25957
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4667823842170837e-06,
+      "loss": 1.3145,
+      "step": 25958
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4662420571795121e-06,
+      "loss": 0.9492,
+      "step": 25959
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4657018218084874e-06,
+      "loss": 1.3135,
+      "step": 25960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4651616781098156e-06,
+      "loss": 1.2979,
+      "step": 25961
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4646216260892966e-06,
+      "loss": 0.8066,
+      "step": 25962
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4640816657527346e-06,
+      "loss": 0.96,
+      "step": 25963
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4635417971059263e-06,
+      "loss": 0.9521,
+      "step": 25964
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4630020201546713e-06,
+      "loss": 0.9463,
+      "step": 25965
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4624623349047673e-06,
+      "loss": 1.0068,
+      "step": 25966
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4619227413620153e-06,
+      "loss": 0.7773,
+      "step": 25967
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4613832395322081e-06,
+      "loss": 1.2354,
+      "step": 25968
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4608438294211402e-06,
+      "loss": 0.7729,
+      "step": 25969
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.460304511034606e-06,
+      "loss": 0.7422,
+      "step": 25970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4597652843784016e-06,
+      "loss": 1.5107,
+      "step": 25971
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.459226149458317e-06,
+      "loss": 0.9355,
+      "step": 25972
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4586871062801423e-06,
+      "loss": 1.3027,
+      "step": 25973
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.45814815484967e-06,
+      "loss": 1.1592,
+      "step": 25974
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4576092951726873e-06,
+      "loss": 1.0596,
+      "step": 25975
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4570705272549846e-06,
+      "loss": 1.2598,
+      "step": 25976
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4565318511023463e-06,
+      "loss": 0.834,
+      "step": 25977
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4559932667205623e-06,
+      "loss": 1.1279,
+      "step": 25978
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4554547741154158e-06,
+      "loss": 1.043,
+      "step": 25979
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4549163732926885e-06,
+      "loss": 1.0879,
+      "step": 25980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4543780642581672e-06,
+      "loss": 0.6855,
+      "step": 25981
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4538398470176352e-06,
+      "loss": 0.8745,
+      "step": 25982
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.453301721576873e-06,
+      "loss": 0.9961,
+      "step": 25983
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4527636879416562e-06,
+      "loss": 0.7964,
+      "step": 25984
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4522257461177713e-06,
+      "loss": 1.1396,
+      "step": 25985
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4516878961109905e-06,
+      "loss": 0.8984,
+      "step": 25986
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4511501379270964e-06,
+      "loss": 0.833,
+      "step": 25987
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4506124715718605e-06,
+      "loss": 0.7021,
+      "step": 25988
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4500748970510636e-06,
+      "loss": 1.0791,
+      "step": 25989
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4495374143704766e-06,
+      "loss": 1.0752,
+      "step": 25990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4490000235358692e-06,
+      "loss": 0.7666,
+      "step": 25991
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4484627245530237e-06,
+      "loss": 0.8477,
+      "step": 25992
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4479255174277051e-06,
+      "loss": 0.8701,
+      "step": 25993
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447388402165686e-06,
+      "loss": 0.9644,
+      "step": 25994
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446851378772731e-06,
+      "loss": 0.5854,
+      "step": 25995
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4463144472546143e-06,
+      "loss": 0.9814,
+      "step": 25996
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4457776076171026e-06,
+      "loss": 0.7134,
+      "step": 25997
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4452408598659595e-06,
+      "loss": 1.248,
+      "step": 25998
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4447042040069547e-06,
+      "loss": 0.8662,
+      "step": 25999
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4441676400458503e-06,
+      "loss": 0.6924,
+      "step": 26000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4436311679884073e-06,
+      "loss": 1.2236,
+      "step": 26001
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4430947878403922e-06,
+      "loss": 0.7881,
+      "step": 26002
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.442558499607567e-06,
+      "loss": 1.0068,
+      "step": 26003
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4420223032956914e-06,
+      "loss": 0.7026,
+      "step": 26004
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4414861989105223e-06,
+      "loss": 0.7183,
+      "step": 26005
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4409501864578202e-06,
+      "loss": 1.1348,
+      "step": 26006
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.440414265943345e-06,
+      "loss": 1.2451,
+      "step": 26007
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4398784373728514e-06,
+      "loss": 0.9033,
+      "step": 26008
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4393427007520933e-06,
+      "loss": 0.7515,
+      "step": 26009
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4388070560868284e-06,
+      "loss": 1.0566,
+      "step": 26010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.43827150338281e-06,
+      "loss": 1.1338,
+      "step": 26011
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4377360426457876e-06,
+      "loss": 0.8306,
+      "step": 26012
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4372006738815148e-06,
+      "loss": 1.0142,
+      "step": 26013
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4366653970957455e-06,
+      "loss": 0.5771,
+      "step": 26014
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4361302122942266e-06,
+      "loss": 1.0645,
+      "step": 26015
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4355951194827055e-06,
+      "loss": 1.1436,
+      "step": 26016
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.435060118666931e-06,
+      "loss": 0.8232,
+      "step": 26017
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4345252098526518e-06,
+      "loss": 0.8516,
+      "step": 26018
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4339903930456133e-06,
+      "loss": 1.0957,
+      "step": 26019
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4334556682515576e-06,
+      "loss": 1.0225,
+      "step": 26020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4329210354762312e-06,
+      "loss": 1.2383,
+      "step": 26021
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.432386494725374e-06,
+      "loss": 0.6494,
+      "step": 26022
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4318520460047325e-06,
+      "loss": 0.9248,
+      "step": 26023
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4313176893200432e-06,
+      "loss": 0.8975,
+      "step": 26024
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.430783424677049e-06,
+      "loss": 1.0684,
+      "step": 26025
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.430249252081487e-06,
+      "loss": 1.1367,
+      "step": 26026
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4297151715390945e-06,
+      "loss": 1.0439,
+      "step": 26027
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.429181183055608e-06,
+      "loss": 0.9951,
+      "step": 26028
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4286472866367674e-06,
+      "loss": 1.0264,
+      "step": 26029
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4281134822883058e-06,
+      "loss": 1.0059,
+      "step": 26030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4275797700159532e-06,
+      "loss": 0.9307,
+      "step": 26031
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4270461498254462e-06,
+      "loss": 0.8604,
+      "step": 26032
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4265126217225189e-06,
+      "loss": 1.1172,
+      "step": 26033
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4259791857128991e-06,
+      "loss": 0.9297,
+      "step": 26034
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4254458418023154e-06,
+      "loss": 0.948,
+      "step": 26035
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4249125899965012e-06,
+      "loss": 0.6055,
+      "step": 26036
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4243794303011793e-06,
+      "loss": 0.9165,
+      "step": 26037
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.423846362722081e-06,
+      "loss": 1.0176,
+      "step": 26038
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4233133872649296e-06,
+      "loss": 1.1182,
+      "step": 26039
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4227805039354525e-06,
+      "loss": 0.9082,
+      "step": 26040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.422247712739373e-06,
+      "loss": 0.9756,
+      "step": 26041
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4217150136824121e-06,
+      "loss": 1.1445,
+      "step": 26042
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4211824067702917e-06,
+      "loss": 0.8484,
+      "step": 26043
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4206498920087386e-06,
+      "loss": 0.9414,
+      "step": 26044
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.420117469403468e-06,
+      "loss": 0.9199,
+      "step": 26045
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4195851389601978e-06,
+      "loss": 0.8262,
+      "step": 26046
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4190529006846499e-06,
+      "loss": 0.7114,
+      "step": 26047
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4185207545825375e-06,
+      "loss": 1.0176,
+      "step": 26048
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4179887006595817e-06,
+      "loss": 0.9175,
+      "step": 26049
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4174567389214922e-06,
+      "loss": 1.0427,
+      "step": 26050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.416924869373988e-06,
+      "loss": 1.1787,
+      "step": 26051
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4163930920227808e-06,
+      "loss": 0.7598,
+      "step": 26052
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4158614068735788e-06,
+      "loss": 0.8179,
+      "step": 26053
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.415329813932097e-06,
+      "loss": 0.7422,
+      "step": 26054
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4147983132040465e-06,
+      "loss": 0.8536,
+      "step": 26055
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4142669046951351e-06,
+      "loss": 1.1729,
+      "step": 26056
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.413735588411068e-06,
+      "loss": 0.9951,
+      "step": 26057
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4132043643575566e-06,
+      "loss": 1.0205,
+      "step": 26058
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.412673232540307e-06,
+      "loss": 1.0742,
+      "step": 26059
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4121421929650237e-06,
+      "loss": 1.0176,
+      "step": 26060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4116112456374075e-06,
+      "loss": 1.1855,
+      "step": 26061
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4110803905631665e-06,
+      "loss": 1.0557,
+      "step": 26062
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4105496277480002e-06,
+      "loss": 1.1777,
+      "step": 26063
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.410018957197612e-06,
+      "loss": 1.0186,
+      "step": 26064
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4094883789176984e-06,
+      "loss": 1.3828,
+      "step": 26065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4089578929139635e-06,
+      "loss": 1.2031,
+      "step": 26066
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.408427499192103e-06,
+      "loss": 1.0215,
+      "step": 26067
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4078971977578126e-06,
+      "loss": 0.9414,
+      "step": 26068
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4073669886167907e-06,
+      "loss": 0.9355,
+      "step": 26069
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4068368717747339e-06,
+      "loss": 1.124,
+      "step": 26070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4063068472373343e-06,
+      "loss": 1.5,
+      "step": 26071
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.405776915010285e-06,
+      "loss": 1.0273,
+      "step": 26072
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4052470750992808e-06,
+      "loss": 0.8081,
+      "step": 26073
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.404717327510009e-06,
+      "loss": 1.2324,
+      "step": 26074
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4041876722481652e-06,
+      "loss": 1.0264,
+      "step": 26075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4036581093194335e-06,
+      "loss": 0.957,
+      "step": 26076
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4031286387295074e-06,
+      "loss": 0.9307,
+      "step": 26077
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.402599260484071e-06,
+      "loss": 0.8623,
+      "step": 26078
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4020699745888067e-06,
+      "loss": 0.8428,
+      "step": 26079
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4015407810494096e-06,
+      "loss": 0.9639,
+      "step": 26080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.401011679871559e-06,
+      "loss": 0.9043,
+      "step": 26081
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4004826710609376e-06,
+      "loss": 0.7661,
+      "step": 26082
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3999537546232279e-06,
+      "loss": 0.7588,
+      "step": 26083
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3994249305641105e-06,
+      "loss": 1.1309,
+      "step": 26084
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3988961988892703e-06,
+      "loss": 0.7485,
+      "step": 26085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3983675596043834e-06,
+      "loss": 1.0801,
+      "step": 26086
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3978390127151265e-06,
+      "loss": 0.8242,
+      "step": 26087
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3973105582271818e-06,
+      "loss": 0.8521,
+      "step": 26088
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3967821961462202e-06,
+      "loss": 1.2949,
+      "step": 26089
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3962539264779206e-06,
+      "loss": 0.7312,
+      "step": 26090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3957257492279596e-06,
+      "loss": 1.0649,
+      "step": 26091
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3951976644020071e-06,
+      "loss": 1.1533,
+      "step": 26092
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3946696720057372e-06,
+      "loss": 1.3516,
+      "step": 26093
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.394141772044817e-06,
+      "loss": 0.8516,
+      "step": 26094
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3936139645249247e-06,
+      "loss": 1.0137,
+      "step": 26095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.393086249451725e-06,
+      "loss": 0.9019,
+      "step": 26096
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3925586268308867e-06,
+      "loss": 0.8467,
+      "step": 26097
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3920310966680794e-06,
+      "loss": 0.6692,
+      "step": 26098
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3915036589689657e-06,
+      "loss": 1.1416,
+      "step": 26099
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3909763137392163e-06,
+      "loss": 0.9658,
+      "step": 26100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.390449060984491e-06,
+      "loss": 1.1602,
+      "step": 26101
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.389921900710458e-06,
+      "loss": 1.29,
+      "step": 26102
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3893948329227768e-06,
+      "loss": 1.2749,
+      "step": 26103
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3888678576271075e-06,
+      "loss": 0.8496,
+      "step": 26104
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3883409748291132e-06,
+      "loss": 1.0322,
+      "step": 26105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.387814184534454e-06,
+      "loss": 1.2031,
+      "step": 26106
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3872874867487885e-06,
+      "loss": 0.9473,
+      "step": 26107
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3867608814777711e-06,
+      "loss": 0.6169,
+      "step": 26108
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3862343687270619e-06,
+      "loss": 1.0391,
+      "step": 26109
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.385707948502314e-06,
+      "loss": 0.8545,
+      "step": 26110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.385181620809185e-06,
+      "loss": 0.8389,
+      "step": 26111
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.384655385653324e-06,
+      "loss": 0.8838,
+      "step": 26112
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3841292430403896e-06,
+      "loss": 0.792,
+      "step": 26113
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3836031929760296e-06,
+      "loss": 1.4492,
+      "step": 26114
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3830772354658927e-06,
+      "loss": 0.7578,
+      "step": 26115
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3825513705156312e-06,
+      "loss": 1.1934,
+      "step": 26116
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3820255981308962e-06,
+      "loss": 0.834,
+      "step": 26117
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3814999183173328e-06,
+      "loss": 0.937,
+      "step": 26118
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3809743310805856e-06,
+      "loss": 1.0293,
+      "step": 26119
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3804488364263024e-06,
+      "loss": 1.2925,
+      "step": 26120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3799234343601298e-06,
+      "loss": 0.9346,
+      "step": 26121
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3793981248877086e-06,
+      "loss": 0.5962,
+      "step": 26122
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3788729080146812e-06,
+      "loss": 1.3486,
+      "step": 26123
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.378347783746692e-06,
+      "loss": 0.8477,
+      "step": 26124
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3778227520893784e-06,
+      "loss": 0.9546,
+      "step": 26125
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.377297813048384e-06,
+      "loss": 0.7266,
+      "step": 26126
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3767729666293428e-06,
+      "loss": 1.1719,
+      "step": 26127
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3762482128378973e-06,
+      "loss": 0.9321,
+      "step": 26128
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.375723551679683e-06,
+      "loss": 0.5259,
+      "step": 26129
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3751989831603308e-06,
+      "loss": 0.9834,
+      "step": 26130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3746745072854805e-06,
+      "loss": 1.168,
+      "step": 26131
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3741501240607658e-06,
+      "loss": 1.1079,
+      "step": 26132
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3736258334918185e-06,
+      "loss": 1.3945,
+      "step": 26133
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3731016355842686e-06,
+      "loss": 0.8438,
+      "step": 26134
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3725775303437495e-06,
+      "loss": 1.1572,
+      "step": 26135
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3720535177758875e-06,
+      "loss": 0.9058,
+      "step": 26136
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3715295978863162e-06,
+      "loss": 0.9639,
+      "step": 26137
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.371005770680659e-06,
+      "loss": 0.9004,
+      "step": 26138
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3704820361645466e-06,
+      "loss": 1.0884,
+      "step": 26139
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3699583943436013e-06,
+      "loss": 0.9258,
+      "step": 26140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3694348452234473e-06,
+      "loss": 0.9971,
+      "step": 26141
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3689113888097116e-06,
+      "loss": 0.7053,
+      "step": 26142
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3683880251080162e-06,
+      "loss": 1.4512,
+      "step": 26143
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.367864754123983e-06,
+      "loss": 0.7031,
+      "step": 26144
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3673415758632303e-06,
+      "loss": 0.7478,
+      "step": 26145
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3668184903313785e-06,
+      "loss": 1.1201,
+      "step": 26146
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3662954975340503e-06,
+      "loss": 1.1631,
+      "step": 26147
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.365772597476861e-06,
+      "loss": 0.9121,
+      "step": 26148
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.365249790165425e-06,
+      "loss": 1.105,
+      "step": 26149
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.364727075605362e-06,
+      "loss": 0.7427,
+      "step": 26150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3642044538022825e-06,
+      "loss": 0.8486,
+      "step": 26151
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.363681924761806e-06,
+      "loss": 1.5449,
+      "step": 26152
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3631594884895394e-06,
+      "loss": 0.791,
+      "step": 26153
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3626371449911003e-06,
+      "loss": 0.5635,
+      "step": 26154
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3621148942720963e-06,
+      "loss": 0.8174,
+      "step": 26155
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3615927363381342e-06,
+      "loss": 0.75,
+      "step": 26156
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3610706711948273e-06,
+      "loss": 0.9658,
+      "step": 26157
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3605486988477845e-06,
+      "loss": 1.001,
+      "step": 26158
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.36002681930261e-06,
+      "loss": 1.0459,
+      "step": 26159
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3595050325649073e-06,
+      "loss": 1.0112,
+      "step": 26160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.358983338640285e-06,
+      "loss": 0.5781,
+      "step": 26161
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3584617375343478e-06,
+      "loss": 1.0244,
+      "step": 26162
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3579402292526967e-06,
+      "loss": 0.7646,
+      "step": 26163
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3574188138009303e-06,
+      "loss": 0.8047,
+      "step": 26164
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3568974911846566e-06,
+      "loss": 0.6914,
+      "step": 26165
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.35637626140947e-06,
+      "loss": 0.9258,
+      "step": 26166
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3558551244809693e-06,
+      "loss": 0.7061,
+      "step": 26167
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3553340804047544e-06,
+      "loss": 0.875,
+      "step": 26168
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.354813129186423e-06,
+      "loss": 1.1597,
+      "step": 26169
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3542922708315708e-06,
+      "loss": 0.7837,
+      "step": 26170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3537715053457889e-06,
+      "loss": 0.9551,
+      "step": 26171
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3532508327346738e-06,
+      "loss": 0.8843,
+      "step": 26172
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3527302530038211e-06,
+      "loss": 0.96,
+      "step": 26173
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3522097661588195e-06,
+      "loss": 1.0234,
+      "step": 26174
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3516893722052582e-06,
+      "loss": 1.3652,
+      "step": 26175
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3511690711487314e-06,
+      "loss": 0.9531,
+      "step": 26176
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3506488629948244e-06,
+      "loss": 1.2754,
+      "step": 26177
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3501287477491255e-06,
+      "loss": 1.002,
+      "step": 26178
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3496087254172253e-06,
+      "loss": 1.0879,
+      "step": 26179
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3490887960047071e-06,
+      "loss": 0.9561,
+      "step": 26180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3485689595171558e-06,
+      "loss": 0.8604,
+      "step": 26181
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.348049215960151e-06,
+      "loss": 0.7451,
+      "step": 26182
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.347529565339284e-06,
+      "loss": 1.0498,
+      "step": 26183
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3470100076601323e-06,
+      "loss": 0.9492,
+      "step": 26184
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3464905429282771e-06,
+      "loss": 0.8613,
+      "step": 26185
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3459711711492962e-06,
+      "loss": 1.626,
+      "step": 26186
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3454518923287708e-06,
+      "loss": 0.8379,
+      "step": 26187
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3449327064722807e-06,
+      "loss": 0.8975,
+      "step": 26188
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.344413613585398e-06,
+      "loss": 0.6377,
+      "step": 26189
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.343894613673703e-06,
+      "loss": 0.9873,
+      "step": 26190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.34337570674277e-06,
+      "loss": 1.1299,
+      "step": 26191
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3428568927981689e-06,
+      "loss": 0.7861,
+      "step": 26192
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3423381718454754e-06,
+      "loss": 1.125,
+      "step": 26193
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3418195438902637e-06,
+      "loss": 1.002,
+      "step": 26194
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3413010089381028e-06,
+      "loss": 0.7148,
+      "step": 26195
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3407825669945607e-06,
+      "loss": 1.0879,
+      "step": 26196
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3402642180652092e-06,
+      "loss": 0.832,
+      "step": 26197
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.339745962155613e-06,
+      "loss": 0.8252,
+      "step": 26198
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3392277992713442e-06,
+      "loss": 1.0078,
+      "step": 26199
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3387097294179619e-06,
+      "loss": 1.1592,
+      "step": 26200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.338191752601037e-06,
+      "loss": 1.1108,
+      "step": 26201
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3376738688261316e-06,
+      "loss": 0.7861,
+      "step": 26202
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3371560780988058e-06,
+      "loss": 0.792,
+      "step": 26203
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.336638380424623e-06,
+      "loss": 1.1396,
+      "step": 26204
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3361207758091489e-06,
+      "loss": 0.7588,
+      "step": 26205
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3356032642579376e-06,
+      "loss": 0.9219,
+      "step": 26206
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.335085845776548e-06,
+      "loss": 0.9209,
+      "step": 26207
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3345685203705405e-06,
+      "loss": 1.252,
+      "step": 26208
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3340512880454738e-06,
+      "loss": 0.8506,
+      "step": 26209
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3335341488068997e-06,
+      "loss": 0.9307,
+      "step": 26210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3330171026603744e-06,
+      "loss": 0.8223,
+      "step": 26211
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3325001496114542e-06,
+      "loss": 1.2559,
+      "step": 26212
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.331983289665687e-06,
+      "loss": 1.2183,
+      "step": 26213
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3314665228286305e-06,
+      "loss": 1.0215,
+      "step": 26214
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3309498491058314e-06,
+      "loss": 1.1738,
+      "step": 26215
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3304332685028421e-06,
+      "loss": 0.8867,
+      "step": 26216
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3299167810252113e-06,
+      "loss": 1.1523,
+      "step": 26217
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3294003866784845e-06,
+      "loss": 1.127,
+      "step": 26218
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3288840854682094e-06,
+      "loss": 1.083,
+      "step": 26219
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3283678773999354e-06,
+      "loss": 1.0498,
+      "step": 26220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3278517624792052e-06,
+      "loss": 0.623,
+      "step": 26221
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3273357407115594e-06,
+      "loss": 0.8076,
+      "step": 26222
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3268198121025444e-06,
+      "loss": 1.0654,
+      "step": 26223
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3263039766577034e-06,
+      "loss": 0.8047,
+      "step": 26224
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3257882343825768e-06,
+      "loss": 0.7344,
+      "step": 26225
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3252725852826998e-06,
+      "loss": 0.9316,
+      "step": 26226
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.324757029363617e-06,
+      "loss": 0.7783,
+      "step": 26227
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.324241566630865e-06,
+      "loss": 1.0029,
+      "step": 26228
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3237261970899773e-06,
+      "loss": 0.9338,
+      "step": 26229
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3232109207464915e-06,
+      "loss": 0.7656,
+      "step": 26230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3226957376059457e-06,
+      "loss": 1.0098,
+      "step": 26231
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.322180647673872e-06,
+      "loss": 1.1416,
+      "step": 26232
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3216656509558001e-06,
+      "loss": 1.0146,
+      "step": 26233
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.321150747457265e-06,
+      "loss": 0.832,
+      "step": 26234
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3206359371837995e-06,
+      "loss": 1.1006,
+      "step": 26235
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3201212201409308e-06,
+      "loss": 0.7961,
+      "step": 26236
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3196065963341865e-06,
+      "loss": 1.2715,
+      "step": 26237
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3190920657690976e-06,
+      "loss": 1.5273,
+      "step": 26238
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3185776284511875e-06,
+      "loss": 1.2656,
+      "step": 26239
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3180632843859865e-06,
+      "loss": 0.877,
+      "step": 26240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3175490335790154e-06,
+      "loss": 1.0791,
+      "step": 26241
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3170348760358021e-06,
+      "loss": 0.9688,
+      "step": 26242
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3165208117618666e-06,
+      "loss": 1.1582,
+      "step": 26243
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3160068407627303e-06,
+      "loss": 1.0859,
+      "step": 26244
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.315492963043914e-06,
+      "loss": 0.8037,
+      "step": 26245
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3149791786109422e-06,
+      "loss": 1.0215,
+      "step": 26246
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3144654874693297e-06,
+      "loss": 0.7798,
+      "step": 26247
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3139518896245929e-06,
+      "loss": 0.8945,
+      "step": 26248
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3134383850822507e-06,
+      "loss": 0.7046,
+      "step": 26249
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3129249738478222e-06,
+      "loss": 0.7109,
+      "step": 26250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3124116559268185e-06,
+      "loss": 0.9707,
+      "step": 26251
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3118984313247529e-06,
+      "loss": 1.1855,
+      "step": 26252
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.311385300047141e-06,
+      "loss": 0.9805,
+      "step": 26253
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3108722620994918e-06,
+      "loss": 1.3516,
+      "step": 26254
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3103593174873185e-06,
+      "loss": 1.0889,
+      "step": 26255
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3098464662161293e-06,
+      "loss": 0.7168,
+      "step": 26256
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3093337082914359e-06,
+      "loss": 0.8486,
+      "step": 26257
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.308821043718743e-06,
+      "loss": 1.2695,
+      "step": 26258
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3083084725035578e-06,
+      "loss": 0.8154,
+      "step": 26259
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3077959946513862e-06,
+      "loss": 0.8433,
+      "step": 26260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3072836101677367e-06,
+      "loss": 1.1733,
+      "step": 26261
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.30677131905811e-06,
+      "loss": 0.9814,
+      "step": 26262
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3062591213280062e-06,
+      "loss": 1.1611,
+      "step": 26263
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3057470169829334e-06,
+      "loss": 1.3184,
+      "step": 26264
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.305235006028386e-06,
+      "loss": 0.9814,
+      "step": 26265
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3047230884698702e-06,
+      "loss": 0.8037,
+      "step": 26266
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.304211264312879e-06,
+      "loss": 1.3213,
+      "step": 26267
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3036995335629144e-06,
+      "loss": 1.2637,
+      "step": 26268
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.303187896225473e-06,
+      "loss": 0.7598,
+      "step": 26269
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3026763523060438e-06,
+      "loss": 1.1719,
+      "step": 26270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3021649018101324e-06,
+      "loss": 1.0049,
+      "step": 26271
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3016535447432265e-06,
+      "loss": 0.9209,
+      "step": 26272
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.301142281110821e-06,
+      "loss": 1.1426,
+      "step": 26273
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3006311109184034e-06,
+      "loss": 1.0811,
+      "step": 26274
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3001200341714682e-06,
+      "loss": 0.6726,
+      "step": 26275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2996090508755056e-06,
+      "loss": 1.0312,
+      "step": 26276
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.299098161036002e-06,
+      "loss": 0.9805,
+      "step": 26277
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.298587364658449e-06,
+      "loss": 0.7671,
+      "step": 26278
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.298076661748331e-06,
+      "loss": 1.2607,
+      "step": 26279
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.297566052311131e-06,
+      "loss": 1.0635,
+      "step": 26280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2970555363523374e-06,
+      "loss": 1.3555,
+      "step": 26281
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2965451138774343e-06,
+      "loss": 0.9795,
+      "step": 26282
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2960347848919042e-06,
+      "loss": 1.0479,
+      "step": 26283
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.295524549401228e-06,
+      "loss": 0.9756,
+      "step": 26284
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.295014407410884e-06,
+      "loss": 0.7729,
+      "step": 26285
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.294504358926354e-06,
+      "loss": 1.0557,
+      "step": 26286
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2939944039531193e-06,
+      "loss": 0.8601,
+      "step": 26287
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2934845424966546e-06,
+      "loss": 0.959,
+      "step": 26288
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.292974774562439e-06,
+      "loss": 0.9102,
+      "step": 26289
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2924651001559463e-06,
+      "loss": 1.0352,
+      "step": 26290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2919555192826494e-06,
+      "loss": 0.9658,
+      "step": 26291
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.291446031948025e-06,
+      "loss": 1.2969,
+      "step": 26292
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2909366381575472e-06,
+      "loss": 0.9697,
+      "step": 26293
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2904273379166854e-06,
+      "loss": 0.7996,
+      "step": 26294
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2899181312309094e-06,
+      "loss": 1.0078,
+      "step": 26295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2894090181056896e-06,
+      "loss": 0.896,
+      "step": 26296
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2888999985464979e-06,
+      "loss": 0.9502,
+      "step": 26297
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2883910725587989e-06,
+      "loss": 1.3643,
+      "step": 26298
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2878822401480573e-06,
+      "loss": 0.936,
+      "step": 26299
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2873735013197442e-06,
+      "loss": 0.8594,
+      "step": 26300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2868648560793184e-06,
+      "loss": 1.0068,
+      "step": 26301
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2863563044322491e-06,
+      "loss": 0.7871,
+      "step": 26302
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.285847846383994e-06,
+      "loss": 0.877,
+      "step": 26303
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.28533948194002e-06,
+      "loss": 1.2158,
+      "step": 26304
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2848312111057849e-06,
+      "loss": 0.7461,
+      "step": 26305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2843230338867452e-06,
+      "loss": 0.8838,
+      "step": 26306
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2838149502883634e-06,
+      "loss": 0.915,
+      "step": 26307
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2833069603160986e-06,
+      "loss": 0.8428,
+      "step": 26308
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2827990639754062e-06,
+      "loss": 0.8325,
+      "step": 26309
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2822912612717387e-06,
+      "loss": 1.2871,
+      "step": 26310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2817835522105514e-06,
+      "loss": 0.9375,
+      "step": 26311
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2812759367973037e-06,
+      "loss": 1.4043,
+      "step": 26312
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2807684150374433e-06,
+      "loss": 0.7222,
+      "step": 26313
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.28026098693642e-06,
+      "loss": 0.876,
+      "step": 26314
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2797536524996902e-06,
+      "loss": 0.959,
+      "step": 26315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2792464117326997e-06,
+      "loss": 0.9229,
+      "step": 26316
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2787392646408959e-06,
+      "loss": 0.9893,
+      "step": 26317
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2782322112297274e-06,
+      "loss": 0.7275,
+      "step": 26318
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2777252515046445e-06,
+      "loss": 0.7913,
+      "step": 26319
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2772183854710896e-06,
+      "loss": 0.9043,
+      "step": 26320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.276711613134506e-06,
+      "loss": 1.0288,
+      "step": 26321
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.276204934500338e-06,
+      "loss": 0.9917,
+      "step": 26322
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2756983495740316e-06,
+      "loss": 1.3047,
+      "step": 26323
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2751918583610258e-06,
+      "loss": 0.7959,
+      "step": 26324
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.274685460866758e-06,
+      "loss": 0.7271,
+      "step": 26325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2741791570966743e-06,
+      "loss": 0.7617,
+      "step": 26326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.273672947056207e-06,
+      "loss": 1.1494,
+      "step": 26327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2731668307507983e-06,
+      "loss": 1.2578,
+      "step": 26328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2726608081858804e-06,
+      "loss": 1.0537,
+      "step": 26329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2721548793668936e-06,
+      "loss": 0.6377,
+      "step": 26330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2716490442992701e-06,
+      "loss": 0.9226,
+      "step": 26331
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2711433029884412e-06,
+      "loss": 0.8623,
+      "step": 26332
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2706376554398414e-06,
+      "loss": 0.9033,
+      "step": 26333
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2701321016589052e-06,
+      "loss": 1.1289,
+      "step": 26334
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2696266416510594e-06,
+      "loss": 1.1211,
+      "step": 26335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2691212754217318e-06,
+      "loss": 0.9971,
+      "step": 26336
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.268616002976354e-06,
+      "loss": 1.0088,
+      "step": 26337
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2681108243203543e-06,
+      "loss": 1.3252,
+      "step": 26338
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2676057394591578e-06,
+      "loss": 0.9473,
+      "step": 26339
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.267100748398188e-06,
+      "loss": 0.7607,
+      "step": 26340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2665958511428722e-06,
+      "loss": 1.1396,
+      "step": 26341
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2660910476986322e-06,
+      "loss": 1.0205,
+      "step": 26342
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.265586338070891e-06,
+      "loss": 1.0811,
+      "step": 26343
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2650817222650692e-06,
+      "loss": 1.1924,
+      "step": 26344
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2645772002865898e-06,
+      "loss": 0.7759,
+      "step": 26345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2640727721408696e-06,
+      "loss": 1.2051,
+      "step": 26346
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.263568437833327e-06,
+      "loss": 1.1201,
+      "step": 26347
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.263064197369378e-06,
+      "loss": 0.7244,
+      "step": 26348
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2625600507544445e-06,
+      "loss": 0.813,
+      "step": 26349
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2620559979939384e-06,
+      "loss": 0.6152,
+      "step": 26350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2615520390932722e-06,
+      "loss": 1.123,
+      "step": 26351
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2610481740578629e-06,
+      "loss": 0.9912,
+      "step": 26352
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.260544402893118e-06,
+      "loss": 0.7339,
+      "step": 26353
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2600407256044544e-06,
+      "loss": 0.8623,
+      "step": 26354
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.259537142197278e-06,
+      "loss": 1.0127,
+      "step": 26355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.259033652677003e-06,
+      "loss": 0.9912,
+      "step": 26356
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2585302570490331e-06,
+      "loss": 0.8104,
+      "step": 26357
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2580269553187763e-06,
+      "loss": 1.2842,
+      "step": 26358
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.257523747491639e-06,
+      "loss": 1.4082,
+      "step": 26359
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2570206335730294e-06,
+      "loss": 0.8467,
+      "step": 26360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2565176135683498e-06,
+      "loss": 1.2363,
+      "step": 26361
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2560146874830014e-06,
+      "loss": 0.9424,
+      "step": 26362
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2555118553223888e-06,
+      "loss": 0.8887,
+      "step": 26363
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2550091170919143e-06,
+      "loss": 1.3008,
+      "step": 26364
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2545064727969758e-06,
+      "loss": 1.0234,
+      "step": 26365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2540039224429724e-06,
+      "loss": 1.1348,
+      "step": 26366
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2535014660353051e-06,
+      "loss": 0.5889,
+      "step": 26367
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2529991035793664e-06,
+      "loss": 1.0664,
+      "step": 26368
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2524968350805555e-06,
+      "loss": 1.248,
+      "step": 26369
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.25199466054427e-06,
+      "loss": 1.0815,
+      "step": 26370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2514925799759015e-06,
+      "loss": 0.9814,
+      "step": 26371
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.250990593380843e-06,
+      "loss": 1.21,
+      "step": 26372
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2504887007644828e-06,
+      "loss": 0.9707,
+      "step": 26373
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2499869021322197e-06,
+      "loss": 1.0625,
+      "step": 26374
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2494851974894417e-06,
+      "loss": 0.4463,
+      "step": 26375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2489835868415323e-06,
+      "loss": 0.8428,
+      "step": 26376
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.248482070193887e-06,
+      "loss": 0.9746,
+      "step": 26377
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2479806475518896e-06,
+      "loss": 0.9443,
+      "step": 26378
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2474793189209244e-06,
+      "loss": 0.8008,
+      "step": 26379
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2469780843063784e-06,
+      "loss": 0.7461,
+      "step": 26380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.246476943713637e-06,
+      "loss": 0.9257,
+      "step": 26381
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2459758971480829e-06,
+      "loss": 1.1016,
+      "step": 26382
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.245474944615097e-06,
+      "loss": 0.7285,
+      "step": 26383
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2449740861200554e-06,
+      "loss": 0.8589,
+      "step": 26384
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.244473321668348e-06,
+      "loss": 0.9922,
+      "step": 26385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2439726512653495e-06,
+      "loss": 1.3018,
+      "step": 26386
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2434720749164354e-06,
+      "loss": 0.9453,
+      "step": 26387
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.242971592626987e-06,
+      "loss": 0.9609,
+      "step": 26388
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2424712044023757e-06,
+      "loss": 0.9941,
+      "step": 26389
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2419709102479804e-06,
+      "loss": 0.9199,
+      "step": 26390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2414707101691725e-06,
+      "loss": 0.7285,
+      "step": 26391
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2409706041713276e-06,
+      "loss": 0.7432,
+      "step": 26392
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.240470592259817e-06,
+      "loss": 0.9336,
+      "step": 26393
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2399706744400075e-06,
+      "loss": 0.6108,
+      "step": 26394
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2394708507172737e-06,
+      "loss": 1.4023,
+      "step": 26395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2389711210969834e-06,
+      "loss": 1.0664,
+      "step": 26396
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.238471485584506e-06,
+      "loss": 1.127,
+      "step": 26397
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2379719441852035e-06,
+      "loss": 0.8525,
+      "step": 26398
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.237472496904446e-06,
+      "loss": 1.2002,
+      "step": 26399
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2369731437475985e-06,
+      "loss": 0.6045,
+      "step": 26400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2364738847200242e-06,
+      "loss": 0.9941,
+      "step": 26401
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2359747198270832e-06,
+      "loss": 0.7017,
+      "step": 26402
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2354756490741416e-06,
+      "loss": 0.9248,
+      "step": 26403
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2349766724665569e-06,
+      "loss": 0.8643,
+      "step": 26404
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2344777900096916e-06,
+      "loss": 0.9893,
+      "step": 26405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2339790017089004e-06,
+      "loss": 1.0625,
+      "step": 26406
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2334803075695479e-06,
+      "loss": 1.2979,
+      "step": 26407
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2329817075969851e-06,
+      "loss": 1.1748,
+      "step": 26408
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2324832017965682e-06,
+      "loss": 0.8955,
+      "step": 26409
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2319847901736537e-06,
+      "loss": 1.0264,
+      "step": 26410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2314864727335973e-06,
+      "loss": 1.1318,
+      "step": 26411
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2309882494817483e-06,
+      "loss": 0.8086,
+      "step": 26412
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2304901204234587e-06,
+      "loss": 1.0312,
+      "step": 26413
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2299920855640812e-06,
+      "loss": 0.8936,
+      "step": 26414
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2294941449089637e-06,
+      "loss": 0.7266,
+      "step": 26415
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2289962984634563e-06,
+      "loss": 1.0576,
+      "step": 26416
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2284985462329046e-06,
+      "loss": 0.957,
+      "step": 26417
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.228000888222659e-06,
+      "loss": 1.0674,
+      "step": 26418
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2275033244380618e-06,
+      "loss": 1.1504,
+      "step": 26419
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2270058548844577e-06,
+      "loss": 0.7231,
+      "step": 26420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.226508479567191e-06,
+      "loss": 0.8633,
+      "step": 26421
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2260111984916069e-06,
+      "loss": 0.957,
+      "step": 26422
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2255140116630437e-06,
+      "loss": 0.7441,
+      "step": 26423
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2250169190868423e-06,
+      "loss": 0.7344,
+      "step": 26424
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2245199207683422e-06,
+      "loss": 1.3467,
+      "step": 26425
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2240230167128863e-06,
+      "loss": 0.8936,
+      "step": 26426
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.223526206925807e-06,
+      "loss": 1.2178,
+      "step": 26427
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2230294914124419e-06,
+      "loss": 0.9932,
+      "step": 26428
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.222532870178128e-06,
+      "loss": 0.917,
+      "step": 26429
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2220363432281978e-06,
+      "loss": 0.9409,
+      "step": 26430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2215399105679881e-06,
+      "loss": 0.7871,
+      "step": 26431
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2210435722028268e-06,
+      "loss": 1.0884,
+      "step": 26432
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2205473281380509e-06,
+      "loss": 0.834,
+      "step": 26433
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.220051178378987e-06,
+      "loss": 1.1611,
+      "step": 26434
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2195551229309633e-06,
+      "loss": 1.3447,
+      "step": 26435
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2190591617993108e-06,
+      "loss": 0.8931,
+      "step": 26436
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.218563294989359e-06,
+      "loss": 0.8037,
+      "step": 26437
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2180675225064308e-06,
+      "loss": 0.8486,
+      "step": 26438
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2175718443558516e-06,
+      "loss": 1.0234,
+      "step": 26439
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2170762605429488e-06,
+      "loss": 0.9844,
+      "step": 26440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2165807710730415e-06,
+      "loss": 0.8799,
+      "step": 26441
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2160853759514567e-06,
+      "loss": 0.8164,
+      "step": 26442
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2155900751835115e-06,
+      "loss": 1.002,
+      "step": 26443
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2150948687745301e-06,
+      "loss": 1.0288,
+      "step": 26444
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2145997567298296e-06,
+      "loss": 1.0278,
+      "step": 26445
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2141047390547267e-06,
+      "loss": 0.7979,
+      "step": 26446
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2136098157545406e-06,
+      "loss": 0.6726,
+      "step": 26447
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2131149868345905e-06,
+      "loss": 0.8545,
+      "step": 26448
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2126202523001885e-06,
+      "loss": 0.8936,
+      "step": 26449
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.212125612156647e-06,
+      "loss": 1.1094,
+      "step": 26450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2116310664092812e-06,
+      "loss": 0.9463,
+      "step": 26451
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2111366150634062e-06,
+      "loss": 0.958,
+      "step": 26452
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2106422581243293e-06,
+      "loss": 1.1533,
+      "step": 26453
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2101479955973617e-06,
+      "loss": 0.8984,
+      "step": 26454
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2096538274878134e-06,
+      "loss": 1.0459,
+      "step": 26455
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2091597538009914e-06,
+      "loss": 0.752,
+      "step": 26456
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2086657745422047e-06,
+      "loss": 1.0146,
+      "step": 26457
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.208171889716756e-06,
+      "loss": 0.8833,
+      "step": 26458
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2076780993299553e-06,
+      "loss": 0.9961,
+      "step": 26459
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.207184403387104e-06,
+      "loss": 1.2314,
+      "step": 26460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2066908018935008e-06,
+      "loss": 0.9434,
+      "step": 26461
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2061972948544575e-06,
+      "loss": 0.8633,
+      "step": 26462
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2057038822752686e-06,
+      "loss": 0.8643,
+      "step": 26463
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2052105641612367e-06,
+      "loss": 1.2754,
+      "step": 26464
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.204717340517657e-06,
+      "loss": 1.04,
+      "step": 26465
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2042242113498304e-06,
+      "loss": 0.8901,
+      "step": 26466
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2037311766630555e-06,
+      "loss": 1.1738,
+      "step": 26467
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.203238236462625e-06,
+      "loss": 1.1621,
+      "step": 26468
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2027453907538367e-06,
+      "loss": 1.1533,
+      "step": 26469
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2022526395419841e-06,
+      "loss": 0.8394,
+      "step": 26470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.201759982832359e-06,
+      "loss": 0.7056,
+      "step": 26471
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.20126742063025e-06,
+      "loss": 1.0293,
+      "step": 26472
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2007749529409551e-06,
+      "loss": 1.1519,
+      "step": 26473
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2002825797697604e-06,
+      "loss": 0.8184,
+      "step": 26474
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1997903011219548e-06,
+      "loss": 1.1484,
+      "step": 26475
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1992981170028273e-06,
+      "loss": 0.8857,
+      "step": 26476
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1988060274176626e-06,
+      "loss": 1.25,
+      "step": 26477
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1983140323717502e-06,
+      "loss": 1.0215,
+      "step": 26478
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1978221318703697e-06,
+      "loss": 0.6616,
+      "step": 26479
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1973303259188107e-06,
+      "loss": 0.9102,
+      "step": 26480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1968386145223532e-06,
+      "loss": 0.8682,
+      "step": 26481
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1963469976862773e-06,
+      "loss": 0.8887,
+      "step": 26482
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1958554754158659e-06,
+      "loss": 0.8574,
+      "step": 26483
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1953640477163997e-06,
+      "loss": 0.7432,
+      "step": 26484
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1948727145931561e-06,
+      "loss": 1.0459,
+      "step": 26485
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1943814760514105e-06,
+      "loss": 1.0361,
+      "step": 26486
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1938903320964423e-06,
+      "loss": 0.8076,
+      "step": 26487
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1933992827335295e-06,
+      "loss": 1.1162,
+      "step": 26488
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1929083279679432e-06,
+      "loss": 1.083,
+      "step": 26489
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1924174678049561e-06,
+      "loss": 1.0293,
+      "step": 26490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.191926702249845e-06,
+      "loss": 1.1758,
+      "step": 26491
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1914360313078777e-06,
+      "loss": 0.9551,
+      "step": 26492
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1909454549843269e-06,
+      "loss": 1.0474,
+      "step": 26493
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1904549732844607e-06,
+      "loss": 1.1309,
+      "step": 26494
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.189964586213551e-06,
+      "loss": 0.8848,
+      "step": 26495
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1894742937768622e-06,
+      "loss": 1.1924,
+      "step": 26496
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1889840959796594e-06,
+      "loss": 1.4688,
+      "step": 26497
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.188493992827211e-06,
+      "loss": 1.4727,
+      "step": 26498
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1880039843247827e-06,
+      "loss": 1.3711,
+      "step": 26499
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1875140704776367e-06,
+      "loss": 0.5581,
+      "step": 26500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1870242512910325e-06,
+      "loss": 0.6792,
+      "step": 26501
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1865345267702355e-06,
+      "loss": 0.998,
+      "step": 26502
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1860448969205029e-06,
+      "loss": 0.874,
+      "step": 26503
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1855553617470982e-06,
+      "loss": 1.1855,
+      "step": 26504
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.185065921255275e-06,
+      "loss": 0.8271,
+      "step": 26505
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1845765754502957e-06,
+      "loss": 1.127,
+      "step": 26506
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.184087324337414e-06,
+      "loss": 0.7314,
+      "step": 26507
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1835981679218833e-06,
+      "loss": 0.8252,
+      "step": 26508
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1831091062089594e-06,
+      "loss": 1.1099,
+      "step": 26509
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1826201392038995e-06,
+      "loss": 1.3516,
+      "step": 26510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1821312669119523e-06,
+      "loss": 1.0459,
+      "step": 26511
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1816424893383671e-06,
+      "loss": 0.6338,
+      "step": 26512
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1811538064883966e-06,
+      "loss": 0.7812,
+      "step": 26513
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1806652183672917e-06,
+      "loss": 1.2227,
+      "step": 26514
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1801767249802998e-06,
+      "loss": 0.8013,
+      "step": 26515
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1796883263326643e-06,
+      "loss": 0.7534,
+      "step": 26516
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1792000224296373e-06,
+      "loss": 1.4102,
+      "step": 26517
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1787118132764586e-06,
+      "loss": 1.1489,
+      "step": 26518
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1782236988783768e-06,
+      "loss": 1.1274,
+      "step": 26519
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1777356792406303e-06,
+      "loss": 1.0918,
+      "step": 26520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1772477543684669e-06,
+      "loss": 1.3438,
+      "step": 26521
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1767599242671245e-06,
+      "loss": 1.1484,
+      "step": 26522
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1762721889418415e-06,
+      "loss": 1.4492,
+      "step": 26523
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.175784548397859e-06,
+      "loss": 0.9219,
+      "step": 26524
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1752970026404176e-06,
+      "loss": 1.0361,
+      "step": 26525
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1748095516747515e-06,
+      "loss": 1.1025,
+      "step": 26526
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1743221955060946e-06,
+      "loss": 0.9141,
+      "step": 26527
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1738349341396848e-06,
+      "loss": 0.9688,
+      "step": 26528
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.173347767580757e-06,
+      "loss": 0.9238,
+      "step": 26529
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1728606958345422e-06,
+      "loss": 0.8857,
+      "step": 26530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1723737189062723e-06,
+      "loss": 0.8862,
+      "step": 26531
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1718868368011804e-06,
+      "loss": 0.9424,
+      "step": 26532
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1714000495244949e-06,
+      "loss": 0.752,
+      "step": 26533
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1709133570814424e-06,
+      "loss": 1.3125,
+      "step": 26534
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1704267594772523e-06,
+      "loss": 0.8428,
+      "step": 26535
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1699402567171558e-06,
+      "loss": 0.896,
+      "step": 26536
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1694538488063733e-06,
+      "loss": 0.8872,
+      "step": 26537
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1689675357501307e-06,
+      "loss": 1.2666,
+      "step": 26538
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1684813175536525e-06,
+      "loss": 0.708,
+      "step": 26539
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1679951942221624e-06,
+      "loss": 0.7969,
+      "step": 26540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1675091657608829e-06,
+      "loss": 1.2031,
+      "step": 26541
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1670232321750296e-06,
+      "loss": 1.0088,
+      "step": 26542
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1665373934698288e-06,
+      "loss": 0.957,
+      "step": 26543
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1660516496504938e-06,
+      "loss": 0.7988,
+      "step": 26544
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1655660007222457e-06,
+      "loss": 1.0801,
+      "step": 26545
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1650804466902987e-06,
+      "loss": 0.7324,
+      "step": 26546
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1645949875598729e-06,
+      "loss": 1.2695,
+      "step": 26547
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1641096233361781e-06,
+      "loss": 1.1748,
+      "step": 26548
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1636243540244274e-06,
+      "loss": 0.7617,
+      "step": 26549
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1631391796298385e-06,
+      "loss": 0.8701,
+      "step": 26550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1626541001576208e-06,
+      "loss": 1.208,
+      "step": 26551
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1621691156129833e-06,
+      "loss": 1.2988,
+      "step": 26552
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1616842260011352e-06,
+      "loss": 0.8662,
+      "step": 26553
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1611994313272856e-06,
+      "loss": 0.6899,
+      "step": 26554
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1607147315966449e-06,
+      "loss": 0.8594,
+      "step": 26555
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1602301268144166e-06,
+      "loss": 0.5723,
+      "step": 26556
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1597456169858057e-06,
+      "loss": 0.998,
+      "step": 26557
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.15926120211602e-06,
+      "loss": 1.0254,
+      "step": 26558
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1587768822102573e-06,
+      "loss": 0.855,
+      "step": 26559
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1582926572737241e-06,
+      "loss": 0.916,
+      "step": 26560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1578085273116235e-06,
+      "loss": 0.7539,
+      "step": 26561
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1573244923291526e-06,
+      "loss": 1.0205,
+      "step": 26562
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1568405523315118e-06,
+      "loss": 0.9141,
+      "step": 26563
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.156356707323898e-06,
+      "loss": 0.623,
+      "step": 26564
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1558729573115079e-06,
+      "loss": 1.0396,
+      "step": 26565
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1553893022995432e-06,
+      "loss": 1.418,
+      "step": 26566
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.154905742293192e-06,
+      "loss": 1.0186,
+      "step": 26567
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1544222772976543e-06,
+      "loss": 1.2021,
+      "step": 26568
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1539389073181217e-06,
+      "loss": 1.2012,
+      "step": 26569
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1534556323597824e-06,
+      "loss": 1.0791,
+      "step": 26570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.152972452427832e-06,
+      "loss": 0.9883,
+      "step": 26571
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1524893675274606e-06,
+      "loss": 0.9766,
+      "step": 26572
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1520063776638556e-06,
+      "loss": 0.8169,
+      "step": 26573
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1515234828422074e-06,
+      "loss": 0.9072,
+      "step": 26574
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1510406830676957e-06,
+      "loss": 0.7686,
+      "step": 26575
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1505579783455167e-06,
+      "loss": 1.0693,
+      "step": 26576
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1500753686808519e-06,
+      "loss": 0.8564,
+      "step": 26577
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1495928540788815e-06,
+      "loss": 0.6689,
+      "step": 26578
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1491104345447935e-06,
+      "loss": 1.0352,
+      "step": 26579
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1486281100837648e-06,
+      "loss": 0.918,
+      "step": 26580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1481458807009827e-06,
+      "loss": 0.998,
+      "step": 26581
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1476637464016215e-06,
+      "loss": 0.7959,
+      "step": 26582
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1471817071908643e-06,
+      "loss": 0.9844,
+      "step": 26583
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1466997630738863e-06,
+      "loss": 1.042,
+      "step": 26584
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1462179140558638e-06,
+      "loss": 0.8867,
+      "step": 26585
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1457361601419737e-06,
+      "loss": 1.4424,
+      "step": 26586
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1452545013373928e-06,
+      "loss": 1.1611,
+      "step": 26587
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1447729376472937e-06,
+      "loss": 0.6501,
+      "step": 26588
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1442914690768459e-06,
+      "loss": 0.7646,
+      "step": 26589
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1438100956312249e-06,
+      "loss": 1.4102,
+      "step": 26590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.143328817315602e-06,
+      "loss": 0.8418,
+      "step": 26591
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1428476341351457e-06,
+      "loss": 0.9561,
+      "step": 26592
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1423665460950217e-06,
+      "loss": 0.709,
+      "step": 26593
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1418855532004025e-06,
+      "loss": 1.1055,
+      "step": 26594
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1414046554564528e-06,
+      "loss": 0.8516,
+      "step": 26595
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1409238528683363e-06,
+      "loss": 0.998,
+      "step": 26596
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.140443145441219e-06,
+      "loss": 1.1367,
+      "step": 26597
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1399625331802677e-06,
+      "loss": 0.8311,
+      "step": 26598
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1394820160906406e-06,
+      "loss": 1.0503,
+      "step": 26599
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1390015941775e-06,
+      "loss": 0.8828,
+      "step": 26600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1385212674460066e-06,
+      "loss": 1.3555,
+      "step": 26601
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1380410359013227e-06,
+      "loss": 1.2295,
+      "step": 26602
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1375608995486042e-06,
+      "loss": 0.8896,
+      "step": 26603
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.137080858393007e-06,
+      "loss": 0.8105,
+      "step": 26604
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1366009124396916e-06,
+      "loss": 0.9619,
+      "step": 26605
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.136121061693809e-06,
+      "loss": 0.8926,
+      "step": 26606
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1356413061605175e-06,
+      "loss": 1.0381,
+      "step": 26607
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1351616458449667e-06,
+      "loss": 1.1123,
+      "step": 26608
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.134682080752314e-06,
+      "loss": 0.8828,
+      "step": 26609
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1342026108877069e-06,
+      "loss": 0.8887,
+      "step": 26610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1337232362562944e-06,
+      "loss": 1.0527,
+      "step": 26611
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.133243956863228e-06,
+      "loss": 0.9883,
+      "step": 26612
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1327647727136582e-06,
+      "loss": 1.2031,
+      "step": 26613
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1322856838127294e-06,
+      "loss": 1.0723,
+      "step": 26614
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1318066901655867e-06,
+      "loss": 0.9116,
+      "step": 26615
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1313277917773757e-06,
+      "loss": 1.1045,
+      "step": 26616
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1308489886532448e-06,
+      "loss": 0.8252,
+      "step": 26617
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.130370280798334e-06,
+      "loss": 1.0967,
+      "step": 26618
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.129891668217783e-06,
+      "loss": 1.1763,
+      "step": 26619
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1294131509167372e-06,
+      "loss": 0.8438,
+      "step": 26620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.128934728900335e-06,
+      "loss": 1.0918,
+      "step": 26621
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1284564021737132e-06,
+      "loss": 0.9512,
+      "step": 26622
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1279781707420123e-06,
+      "loss": 0.9707,
+      "step": 26623
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1275000346103704e-06,
+      "loss": 0.8794,
+      "step": 26624
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1270219937839212e-06,
+      "loss": 1.04,
+      "step": 26625
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1265440482677993e-06,
+      "loss": 0.8496,
+      "step": 26626
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1260661980671383e-06,
+      "loss": 0.917,
+      "step": 26627
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1255884431870757e-06,
+      "loss": 0.9717,
+      "step": 26628
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.125110783632739e-06,
+      "loss": 0.7334,
+      "step": 26629
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1246332194092591e-06,
+      "loss": 0.958,
+      "step": 26630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1241557505217682e-06,
+      "loss": 0.8184,
+      "step": 26631
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1236783769753924e-06,
+      "loss": 1.0586,
+      "step": 26632
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1232010987752618e-06,
+      "loss": 1.0156,
+      "step": 26633
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1227239159265013e-06,
+      "loss": 0.4863,
+      "step": 26634
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1222468284342391e-06,
+      "loss": 0.8535,
+      "step": 26635
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1217698363035978e-06,
+      "loss": 0.7754,
+      "step": 26636
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1212929395396998e-06,
+      "loss": 0.8545,
+      "step": 26637
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.12081613814767e-06,
+      "loss": 1.1787,
+      "step": 26638
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1203394321326312e-06,
+      "loss": 0.9355,
+      "step": 26639
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1198628214997033e-06,
+      "loss": 0.6719,
+      "step": 26640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1193863062540012e-06,
+      "loss": 0.5159,
+      "step": 26641
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1189098864006488e-06,
+      "loss": 0.8374,
+      "step": 26642
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1184335619447628e-06,
+      "loss": 0.9648,
+      "step": 26643
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1179573328914596e-06,
+      "loss": 0.4297,
+      "step": 26644
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1174811992458524e-06,
+      "loss": 0.9248,
+      "step": 26645
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1170051610130583e-06,
+      "loss": 1.2632,
+      "step": 26646
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1165292181981878e-06,
+      "loss": 0.8818,
+      "step": 26647
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.116053370806356e-06,
+      "loss": 0.7246,
+      "step": 26648
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1155776188426748e-06,
+      "loss": 0.9473,
+      "step": 26649
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1151019623122539e-06,
+      "loss": 0.6719,
+      "step": 26650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1146264012202013e-06,
+      "loss": 1.2109,
+      "step": 26651
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.114150935571624e-06,
+      "loss": 1.0371,
+      "step": 26652
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1136755653716313e-06,
+      "loss": 0.876,
+      "step": 26653
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1132002906253314e-06,
+      "loss": 1.3418,
+      "step": 26654
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.112725111337828e-06,
+      "loss": 1.168,
+      "step": 26655
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1122500275142233e-06,
+      "loss": 0.75,
+      "step": 26656
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1117750391596228e-06,
+      "loss": 1.0596,
+      "step": 26657
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1113001462791261e-06,
+      "loss": 0.7373,
+      "step": 26658
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1108253488778364e-06,
+      "loss": 0.9276,
+      "step": 26659
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.110350646960856e-06,
+      "loss": 1.2666,
+      "step": 26660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.109876040533281e-06,
+      "loss": 0.9902,
+      "step": 26661
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1094015296002104e-06,
+      "loss": 1.1514,
+      "step": 26662
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1089271141667357e-06,
+      "loss": 0.9219,
+      "step": 26663
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.108452794237963e-06,
+      "loss": 1.2793,
+      "step": 26664
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1079785698189816e-06,
+      "loss": 1.2725,
+      "step": 26665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.107504440914885e-06,
+      "loss": 1.0869,
+      "step": 26666
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.107030407530768e-06,
+      "loss": 1.0908,
+      "step": 26667
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.10655646967172e-06,
+      "loss": 0.9707,
+      "step": 26668
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.106082627342836e-06,
+      "loss": 0.7588,
+      "step": 26669
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1056088805492016e-06,
+      "loss": 1.0698,
+      "step": 26670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1051352292959084e-06,
+      "loss": 1.1895,
+      "step": 26671
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1046616735880443e-06,
+      "loss": 0.8066,
+      "step": 26672
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1041882134306924e-06,
+      "loss": 1.0186,
+      "step": 26673
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1037148488289418e-06,
+      "loss": 0.8301,
+      "step": 26674
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.103241579787877e-06,
+      "loss": 1.0596,
+      "step": 26675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.102768406312582e-06,
+      "loss": 1.1299,
+      "step": 26676
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1022953284081373e-06,
+      "loss": 1.4844,
+      "step": 26677
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.101822346079625e-06,
+      "loss": 0.6787,
+      "step": 26678
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1013494593321283e-06,
+      "loss": 0.7432,
+      "step": 26679
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.100876668170725e-06,
+      "loss": 0.8867,
+      "step": 26680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1004039726004922e-06,
+      "loss": 1.3154,
+      "step": 26681
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0999313726265104e-06,
+      "loss": 1.0117,
+      "step": 26682
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0994588682538555e-06,
+      "loss": 1.3477,
+      "step": 26683
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0989864594875988e-06,
+      "loss": 0.7837,
+      "step": 26684
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0985141463328185e-06,
+      "loss": 1.0752,
+      "step": 26685
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.09804192879459e-06,
+      "loss": 0.9839,
+      "step": 26686
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0975698068779827e-06,
+      "loss": 1.2529,
+      "step": 26687
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0970977805880666e-06,
+      "loss": 1.0967,
+      "step": 26688
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0966258499299144e-06,
+      "loss": 1.2061,
+      "step": 26689
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0961540149085958e-06,
+      "loss": 1.2207,
+      "step": 26690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0956822755291785e-06,
+      "loss": 1.0923,
+      "step": 26691
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0952106317967281e-06,
+      "loss": 0.9785,
+      "step": 26692
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.094739083716313e-06,
+      "loss": 0.9082,
+      "step": 26693
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0942676312929968e-06,
+      "loss": 0.5952,
+      "step": 26694
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0937962745318465e-06,
+      "loss": 1.1011,
+      "step": 26695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0933250134379202e-06,
+      "loss": 0.9316,
+      "step": 26696
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0928538480162853e-06,
+      "loss": 0.9111,
+      "step": 26697
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0923827782720008e-06,
+      "loss": 0.9087,
+      "step": 26698
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.091911804210124e-06,
+      "loss": 0.7461,
+      "step": 26699
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0914409258357172e-06,
+      "loss": 0.7607,
+      "step": 26700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.09097014315384e-06,
+      "loss": 0.9834,
+      "step": 26701
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0904994561695458e-06,
+      "loss": 1.3066,
+      "step": 26702
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0900288648878909e-06,
+      "loss": 0.9619,
+      "step": 26703
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0895583693139311e-06,
+      "loss": 1.168,
+      "step": 26704
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0890879694527213e-06,
+      "loss": 0.9131,
+      "step": 26705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.088617665309315e-06,
+      "loss": 0.8145,
+      "step": 26706
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0881474568887595e-06,
+      "loss": 0.8052,
+      "step": 26707
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0876773441961097e-06,
+      "loss": 0.9082,
+      "step": 26708
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0872073272364136e-06,
+      "loss": 1.1104,
+      "step": 26709
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0867374060147217e-06,
+      "loss": 0.7686,
+      "step": 26710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.08626758053608e-06,
+      "loss": 0.8525,
+      "step": 26711
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0857978508055367e-06,
+      "loss": 0.874,
+      "step": 26712
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0853282168281366e-06,
+      "loss": 0.7002,
+      "step": 26713
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.084858678608922e-06,
+      "loss": 1.1309,
+      "step": 26714
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0843892361529395e-06,
+      "loss": 0.8994,
+      "step": 26715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0839198894652326e-06,
+      "loss": 0.752,
+      "step": 26716
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0834506385508414e-06,
+      "loss": 0.7344,
+      "step": 26717
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0829814834148044e-06,
+      "loss": 1.3516,
+      "step": 26718
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0825124240621644e-06,
+      "loss": 1.2285,
+      "step": 26719
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0820434604979568e-06,
+      "loss": 0.7219,
+      "step": 26720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0815745927272226e-06,
+      "loss": 0.9844,
+      "step": 26721
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0811058207549952e-06,
+      "loss": 0.6357,
+      "step": 26722
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0806371445863117e-06,
+      "loss": 1.1582,
+      "step": 26723
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.080168564226206e-06,
+      "loss": 0.9062,
+      "step": 26724
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0797000796797096e-06,
+      "loss": 1.1523,
+      "step": 26725
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0792316909518575e-06,
+      "loss": 0.6206,
+      "step": 26726
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0787633980476796e-06,
+      "loss": 0.8462,
+      "step": 26727
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0782952009722082e-06,
+      "loss": 1.0918,
+      "step": 26728
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0778270997304685e-06,
+      "loss": 1.3311,
+      "step": 26729
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0773590943274904e-06,
+      "loss": 1.001,
+      "step": 26730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.076891184768304e-06,
+      "loss": 0.8613,
+      "step": 26731
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.076423371057932e-06,
+      "loss": 1.1465,
+      "step": 26732
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0759556532013994e-06,
+      "loss": 1.1631,
+      "step": 26733
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.075488031203733e-06,
+      "loss": 0.6177,
+      "step": 26734
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0750205050699525e-06,
+      "loss": 1.3623,
+      "step": 26735
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0745530748050825e-06,
+      "loss": 1.125,
+      "step": 26736
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0740857404141425e-06,
+      "loss": 1.0303,
+      "step": 26737
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.073618501902154e-06,
+      "loss": 0.9341,
+      "step": 26738
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.073151359274135e-06,
+      "loss": 0.6633,
+      "step": 26739
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0726843125350982e-06,
+      "loss": 0.9077,
+      "step": 26740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.072217361690071e-06,
+      "loss": 1.0977,
+      "step": 26741
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0717505067440637e-06,
+      "loss": 0.9805,
+      "step": 26742
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.07128374770209e-06,
+      "loss": 0.8779,
+      "step": 26743
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0708170845691635e-06,
+      "loss": 0.7852,
+      "step": 26744
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0703505173502993e-06,
+      "loss": 1.0889,
+      "step": 26745
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0698840460505067e-06,
+      "loss": 1.0469,
+      "step": 26746
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.069417670674797e-06,
+      "loss": 0.7466,
+      "step": 26747
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0689513912281835e-06,
+      "loss": 0.9854,
+      "step": 26748
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0684852077156705e-06,
+      "loss": 1.0732,
+      "step": 26749
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0680191201422686e-06,
+      "loss": 1.3604,
+      "step": 26750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.067553128512977e-06,
+      "loss": 0.9287,
+      "step": 26751
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0670872328328108e-06,
+      "loss": 0.8389,
+      "step": 26752
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0666214331067715e-06,
+      "loss": 1.1709,
+      "step": 26753
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0661557293398606e-06,
+      "loss": 1.1748,
+      "step": 26754
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0656901215370795e-06,
+      "loss": 1.3232,
+      "step": 26755
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0652246097034325e-06,
+      "loss": 1.1387,
+      "step": 26756
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0647591938439195e-06,
+      "loss": 1.1504,
+      "step": 26757
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.064293873963539e-06,
+      "loss": 0.7412,
+      "step": 26758
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.06382865006729e-06,
+      "loss": 0.7847,
+      "step": 26759
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0633635221601702e-06,
+      "loss": 0.916,
+      "step": 26760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.062898490247174e-06,
+      "loss": 0.8125,
+      "step": 26761
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0624335543332964e-06,
+      "loss": 1.1543,
+      "step": 26762
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0619687144235347e-06,
+      "loss": 0.8369,
+      "step": 26763
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0615039705228801e-06,
+      "loss": 0.8169,
+      "step": 26764
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0610393226363236e-06,
+      "loss": 0.7539,
+      "step": 26765
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0605747707688573e-06,
+      "loss": 0.8906,
+      "step": 26766
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0601103149254732e-06,
+      "loss": 0.7349,
+      "step": 26767
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0596459551111582e-06,
+      "loss": 0.7817,
+      "step": 26768
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0591816913308982e-06,
+      "loss": 0.8838,
+      "step": 26769
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0587175235896862e-06,
+      "loss": 0.9604,
+      "step": 26770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0582534518925014e-06,
+      "loss": 1.1689,
+      "step": 26771
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.057789476244333e-06,
+      "loss": 0.8662,
+      "step": 26772
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0573255966501627e-06,
+      "loss": 0.811,
+      "step": 26773
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0568618131149765e-06,
+      "loss": 0.8652,
+      "step": 26774
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0563981256437527e-06,
+      "loss": 0.8325,
+      "step": 26775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0559345342414728e-06,
+      "loss": 0.9531,
+      "step": 26776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0554710389131162e-06,
+      "loss": 1.0234,
+      "step": 26777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0550076396636644e-06,
+      "loss": 0.9551,
+      "step": 26778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0545443364980934e-06,
+      "loss": 1.4443,
+      "step": 26779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0540811294213771e-06,
+      "loss": 0.9844,
+      "step": 26780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.053618018438496e-06,
+      "loss": 0.9277,
+      "step": 26781
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0531550035544202e-06,
+      "loss": 0.9248,
+      "step": 26782
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0526920847741263e-06,
+      "loss": 1.1172,
+      "step": 26783
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0522292621025853e-06,
+      "loss": 0.728,
+      "step": 26784
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0517665355447704e-06,
+      "loss": 0.752,
+      "step": 26785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0513039051056507e-06,
+      "loss": 0.8965,
+      "step": 26786
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0508413707901931e-06,
+      "loss": 0.6143,
+      "step": 26787
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0503789326033697e-06,
+      "loss": 1.3477,
+      "step": 26788
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0499165905501485e-06,
+      "loss": 1.1211,
+      "step": 26789
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0494543446354933e-06,
+      "loss": 0.8496,
+      "step": 26790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0489921948643688e-06,
+      "loss": 0.7905,
+      "step": 26791
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0485301412417404e-06,
+      "loss": 0.7656,
+      "step": 26792
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0480681837725727e-06,
+      "loss": 0.8999,
+      "step": 26793
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0476063224618271e-06,
+      "loss": 0.8672,
+      "step": 26794
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0471445573144623e-06,
+      "loss": 1.0417,
+      "step": 26795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0466828883354429e-06,
+      "loss": 0.6958,
+      "step": 26796
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.046221315529723e-06,
+      "loss": 1.0625,
+      "step": 26797
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.045759838902266e-06,
+      "loss": 0.8906,
+      "step": 26798
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0452984584580228e-06,
+      "loss": 0.9736,
+      "step": 26799
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0448371742019548e-06,
+      "loss": 0.8496,
+      "step": 26800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0443759861390157e-06,
+      "loss": 1.1035,
+      "step": 26801
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.043914894274156e-06,
+      "loss": 1.1475,
+      "step": 26802
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.043453898612331e-06,
+      "loss": 0.8564,
+      "step": 26803
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0429929991584942e-06,
+      "loss": 1.0371,
+      "step": 26804
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.042532195917595e-06,
+      "loss": 0.9111,
+      "step": 26805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0420714888945816e-06,
+      "loss": 0.8252,
+      "step": 26806
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0416108780944056e-06,
+      "loss": 0.728,
+      "step": 26807
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0411503635220121e-06,
+      "loss": 0.5815,
+      "step": 26808
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0406899451823504e-06,
+      "loss": 0.959,
+      "step": 26809
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.040229623080362e-06,
+      "loss": 0.9189,
+      "step": 26810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0397693972209977e-06,
+      "loss": 1.0771,
+      "step": 26811
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0393092676091965e-06,
+      "loss": 0.959,
+      "step": 26812
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0388492342499001e-06,
+      "loss": 0.7451,
+      "step": 26813
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0383892971480524e-06,
+      "loss": 1.0034,
+      "step": 26814
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.037929456308595e-06,
+      "loss": 1.1562,
+      "step": 26815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0374697117364663e-06,
+      "loss": 0.9834,
+      "step": 26816
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.037010063436602e-06,
+      "loss": 1.0449,
+      "step": 26817
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0365505114139417e-06,
+      "loss": 0.6963,
+      "step": 26818
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0360910556734238e-06,
+      "loss": 1.4902,
+      "step": 26819
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.035631696219982e-06,
+      "loss": 1.02,
+      "step": 26820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0351724330585466e-06,
+      "loss": 1.1387,
+      "step": 26821
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0347132661940572e-06,
+      "loss": 0.6211,
+      "step": 26822
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0342541956314423e-06,
+      "loss": 0.6489,
+      "step": 26823
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0337952213756342e-06,
+      "loss": 0.6523,
+      "step": 26824
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0333363434315613e-06,
+      "loss": 0.7085,
+      "step": 26825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0328775618041565e-06,
+      "loss": 1.0303,
+      "step": 26826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0324188764983456e-06,
+      "loss": 0.7874,
+      "step": 26827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0319602875190525e-06,
+      "loss": 0.8721,
+      "step": 26828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.031501794871207e-06,
+      "loss": 1.0684,
+      "step": 26829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0310433985597357e-06,
+      "loss": 1.0312,
+      "step": 26830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0305850985895594e-06,
+      "loss": 0.9629,
+      "step": 26831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0301268949656007e-06,
+      "loss": 1.0146,
+      "step": 26832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0296687876927813e-06,
+      "loss": 0.5566,
+      "step": 26833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0292107767760261e-06,
+      "loss": 0.9258,
+      "step": 26834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0287528622202513e-06,
+      "loss": 1.1992,
+      "step": 26835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0282950440303752e-06,
+      "loss": 1.2871,
+      "step": 26836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.027837322211318e-06,
+      "loss": 0.853,
+      "step": 26837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.027379696767995e-06,
+      "loss": 0.8335,
+      "step": 26838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0269221677053177e-06,
+      "loss": 1.0791,
+      "step": 26839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0264647350282097e-06,
+      "loss": 0.9766,
+      "step": 26840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0260073987415798e-06,
+      "loss": 0.7041,
+      "step": 26841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0255501588503391e-06,
+      "loss": 0.9297,
+      "step": 26842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0250930153593996e-06,
+      "loss": 1.1729,
+      "step": 26843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0246359682736729e-06,
+      "loss": 0.7837,
+      "step": 26844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0241790175980704e-06,
+      "loss": 0.6753,
+      "step": 26845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0237221633374982e-06,
+      "loss": 1.0654,
+      "step": 26846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0232654054968617e-06,
+      "loss": 1.2158,
+      "step": 26847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0228087440810707e-06,
+      "loss": 0.7432,
+      "step": 26848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0223521790950276e-06,
+      "loss": 0.9824,
+      "step": 26849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0218957105436378e-06,
+      "loss": 0.6543,
+      "step": 26850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0214393384318077e-06,
+      "loss": 1.2759,
+      "step": 26851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0209830627644369e-06,
+      "loss": 0.8232,
+      "step": 26852
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0205268835464254e-06,
+      "loss": 0.8965,
+      "step": 26853
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0200708007826698e-06,
+      "loss": 0.9272,
+      "step": 26854
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0196148144780793e-06,
+      "loss": 1.4434,
+      "step": 26855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0191589246375456e-06,
+      "loss": 0.999,
+      "step": 26856
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0187031312659634e-06,
+      "loss": 1.0039,
+      "step": 26857
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0182474343682346e-06,
+      "loss": 0.8579,
+      "step": 26858
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0177918339492488e-06,
+      "loss": 1.0684,
+      "step": 26859
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.017336330013905e-06,
+      "loss": 1.0322,
+      "step": 26860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0168809225670907e-06,
+      "loss": 0.8926,
+      "step": 26861
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0164256116137016e-06,
+      "loss": 0.6616,
+      "step": 26862
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0159703971586287e-06,
+      "loss": 0.6865,
+      "step": 26863
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0155152792067568e-06,
+      "loss": 1.1162,
+      "step": 26864
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0150602577629797e-06,
+      "loss": 1.3071,
+      "step": 26865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0146053328321847e-06,
+      "loss": 1.0684,
+      "step": 26866
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0141505044192567e-06,
+      "loss": 1.0117,
+      "step": 26867
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0136957725290808e-06,
+      "loss": 1.1826,
+      "step": 26868
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.013241137166543e-06,
+      "loss": 0.96,
+      "step": 26869
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.012786598336526e-06,
+      "loss": 0.7651,
+      "step": 26870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0123321560439136e-06,
+      "loss": 0.9634,
+      "step": 26871
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0118778102935856e-06,
+      "loss": 1.0508,
+      "step": 26872
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0114235610904244e-06,
+      "loss": 0.8105,
+      "step": 26873
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0109694084393085e-06,
+      "loss": 0.9971,
+      "step": 26874
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0105153523451129e-06,
+      "loss": 0.7129,
+      "step": 26875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.010061392812719e-06,
+      "loss": 0.7417,
+      "step": 26876
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0096075298470032e-06,
+      "loss": 1.125,
+      "step": 26877
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0091537634528402e-06,
+      "loss": 1.0615,
+      "step": 26878
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.008700093635101e-06,
+      "loss": 1.1074,
+      "step": 26879
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0082465203986613e-06,
+      "loss": 1.0703,
+      "step": 26880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0077930437483964e-06,
+      "loss": 0.9785,
+      "step": 26881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.007339663689173e-06,
+      "loss": 1.3604,
+      "step": 26882
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0068863802258598e-06,
+      "loss": 1.1934,
+      "step": 26883
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0064331933633308e-06,
+      "loss": 0.594,
+      "step": 26884
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0059801031064497e-06,
+      "loss": 0.7788,
+      "step": 26885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005527109460087e-06,
+      "loss": 1.0283,
+      "step": 26886
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0050742124291058e-06,
+      "loss": 0.7871,
+      "step": 26887
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0046214120183728e-06,
+      "loss": 0.8994,
+      "step": 26888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0041687082327512e-06,
+      "loss": 0.689,
+      "step": 26889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0037161010771013e-06,
+      "loss": 1.3037,
+      "step": 26890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003263590556288e-06,
+      "loss": 0.6172,
+      "step": 26891
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0028111766751726e-06,
+      "loss": 1.1162,
+      "step": 26892
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0023588594386124e-06,
+      "loss": 0.876,
+      "step": 26893
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0019066388514664e-06,
+      "loss": 1.0283,
+      "step": 26894
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0014545149185916e-06,
+      "loss": 0.9004,
+      "step": 26895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0010024876448488e-06,
+      "loss": 0.9683,
+      "step": 26896
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0005505570350893e-06,
+      "loss": 1.0918,
+      "step": 26897
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0000987230941673e-06,
+      "loss": 0.8408,
+      "step": 26898
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.996469858269386e-07,
+      "loss": 0.9258,
+      "step": 26899
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.991953452382552e-07,
+      "loss": 0.644,
+      "step": 26900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.987438013329664e-07,
+      "loss": 0.6582,
+      "step": 26901
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.98292354115924e-07,
+      "loss": 1.209,
+      "step": 26902
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.978410035919784e-07,
+      "loss": 0.8735,
+      "step": 26903
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.973897497659767e-07,
+      "loss": 0.9707,
+      "step": 26904
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.969385926427643e-07,
+      "loss": 1.457,
+      "step": 26905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.964875322271895e-07,
+      "loss": 1.0908,
+      "step": 26906
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.96036568524099e-07,
+      "loss": 1.25,
+      "step": 26907
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.95585701538334e-07,
+      "loss": 0.7476,
+      "step": 26908
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.951349312747383e-07,
+      "loss": 1.0205,
+      "step": 26909
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.946842577381544e-07,
+      "loss": 0.8281,
+      "step": 26910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.942336809334218e-07,
+      "loss": 1.0732,
+      "step": 26911
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.937832008653837e-07,
+      "loss": 1.1177,
+      "step": 26912
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.933328175388745e-07,
+      "loss": 1.1738,
+      "step": 26913
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.928825309587375e-07,
+      "loss": 0.9365,
+      "step": 26914
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.924323411298054e-07,
+      "loss": 0.9746,
+      "step": 26915
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.91982248056914e-07,
+      "loss": 1.2114,
+      "step": 26916
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.91532251744899e-07,
+      "loss": 1.1338,
+      "step": 26917
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.910823521985956e-07,
+      "loss": 1.1172,
+      "step": 26918
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.906325494228354e-07,
+      "loss": 0.6965,
+      "step": 26919
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.90182843422447e-07,
+      "loss": 1.2207,
+      "step": 26920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.897332342022648e-07,
+      "loss": 0.8203,
+      "step": 26921
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.89283721767118e-07,
+      "loss": 1.2051,
+      "step": 26922
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.888343061218352e-07,
+      "loss": 1.0117,
+      "step": 26923
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.883849872712413e-07,
+      "loss": 0.8115,
+      "step": 26924
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.87935765220166e-07,
+      "loss": 0.915,
+      "step": 26925
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.874866399734306e-07,
+      "loss": 1.0771,
+      "step": 26926
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.870376115358648e-07,
+      "loss": 0.959,
+      "step": 26927
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.86588679912287e-07,
+      "loss": 0.8203,
+      "step": 26928
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.861398451075221e-07,
+      "loss": 1.4512,
+      "step": 26929
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.85691107126392e-07,
+      "loss": 0.8535,
+      "step": 26930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.852424659737136e-07,
+      "loss": 1.1201,
+      "step": 26931
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.84793921654308e-07,
+      "loss": 0.8555,
+      "step": 26932
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.84345474172994e-07,
+      "loss": 1.0186,
+      "step": 26933
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.838971235345894e-07,
+      "loss": 1.0107,
+      "step": 26934
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.83448869743907e-07,
+      "loss": 1.2246,
+      "step": 26935
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.83000712805765e-07,
+      "loss": 1.4297,
+      "step": 26936
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.82552652724974e-07,
+      "loss": 1.0352,
+      "step": 26937
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.821046895063479e-07,
+      "loss": 1.1357,
+      "step": 26938
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.81656823154703e-07,
+      "loss": 1.1621,
+      "step": 26939
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.812090536748453e-07,
+      "loss": 1.0215,
+      "step": 26940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.807613810715865e-07,
+      "loss": 0.833,
+      "step": 26941
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.803138053497318e-07,
+      "loss": 1.0693,
+      "step": 26942
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.79866326514094e-07,
+      "loss": 1.1426,
+      "step": 26943
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.79418944569479e-07,
+      "loss": 1.1079,
+      "step": 26944
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.789716595206899e-07,
+      "loss": 1.2939,
+      "step": 26945
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.785244713725316e-07,
+      "loss": 0.915,
+      "step": 26946
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.78077380129807e-07,
+      "loss": 1.1104,
+      "step": 26947
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.776303857973235e-07,
+      "loss": 0.6816,
+      "step": 26948
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.771834883798758e-07,
+      "loss": 1.0283,
+      "step": 26949
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.767366878822705e-07,
+      "loss": 1.0107,
+      "step": 26950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.762899843093031e-07,
+      "loss": 1.0586,
+      "step": 26951
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.758433776657717e-07,
+      "loss": 0.876,
+      "step": 26952
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.75396867956474e-07,
+      "loss": 0.8823,
+      "step": 26953
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.749504551862087e-07,
+      "loss": 0.9058,
+      "step": 26954
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.74504139359771e-07,
+      "loss": 0.957,
+      "step": 26955
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.7405792048195e-07,
+      "loss": 0.9961,
+      "step": 26956
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.736117985575443e-07,
+      "loss": 1.0811,
+      "step": 26957
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.731657735913424e-07,
+      "loss": 0.8003,
+      "step": 26958
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.72719845588138e-07,
+      "loss": 0.8574,
+      "step": 26959
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.722740145527188e-07,
+      "loss": 0.7217,
+      "step": 26960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.71828280489876e-07,
+      "loss": 1.146,
+      "step": 26961
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.713826434043972e-07,
+      "loss": 0.999,
+      "step": 26962
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.70937103301065e-07,
+      "loss": 0.8003,
+      "step": 26963
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.704916601846704e-07,
+      "loss": 0.9551,
+      "step": 26964
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.70046314059998e-07,
+      "loss": 0.7871,
+      "step": 26965
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.696010649318288e-07,
+      "loss": 0.8379,
+      "step": 26966
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.691559128049466e-07,
+      "loss": 1.3145,
+      "step": 26967
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.68710857684132e-07,
+      "loss": 1.084,
+      "step": 26968
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.682658995741689e-07,
+      "loss": 0.8916,
+      "step": 26969
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.678210384798348e-07,
+      "loss": 0.6074,
+      "step": 26970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.673762744059068e-07,
+      "loss": 0.4431,
+      "step": 26971
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.669316073571654e-07,
+      "loss": 1.2109,
+      "step": 26972
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.664870373383827e-07,
+      "loss": 1.1689,
+      "step": 26973
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.660425643543403e-07,
+      "loss": 0.9717,
+      "step": 26974
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.655981884098065e-07,
+      "loss": 1.3945,
+      "step": 26975
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.651539095095586e-07,
+      "loss": 0.8672,
+      "step": 26976
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.647097276583683e-07,
+      "loss": 0.8296,
+      "step": 26977
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.642656428610042e-07,
+      "loss": 0.8135,
+      "step": 26978
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.63821655122238e-07,
+      "loss": 0.9033,
+      "step": 26979
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.633777644468412e-07,
+      "loss": 0.8418,
+      "step": 26980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.629339708395802e-07,
+      "loss": 1.082,
+      "step": 26981
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.6249027430522e-07,
+      "loss": 0.7207,
+      "step": 26982
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.62046674848528e-07,
+      "loss": 0.8438,
+      "step": 26983
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.616031724742725e-07,
+      "loss": 1.1038,
+      "step": 26984
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.611597671872143e-07,
+      "loss": 0.792,
+      "step": 26985
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.607164589921135e-07,
+      "loss": 1.1973,
+      "step": 26986
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.602732478937381e-07,
+      "loss": 0.9756,
+      "step": 26987
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.598301338968464e-07,
+      "loss": 1.3389,
+      "step": 26988
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.593871170061942e-07,
+      "loss": 0.7402,
+      "step": 26989
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.589441972265444e-07,
+      "loss": 1.209,
+      "step": 26990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.58501374562656e-07,
+      "loss": 0.9326,
+      "step": 26991
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.580586490192834e-07,
+      "loss": 1.2754,
+      "step": 26992
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.576160206011808e-07,
+      "loss": 0.8125,
+      "step": 26993
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.571734893131036e-07,
+      "loss": 0.9912,
+      "step": 26994
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.567310551598086e-07,
+      "loss": 1.144,
+      "step": 26995
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.562887181460456e-07,
+      "loss": 0.8293,
+      "step": 26996
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.558464782765642e-07,
+      "loss": 0.9033,
+      "step": 26997
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.554043355561182e-07,
+      "loss": 0.7261,
+      "step": 26998
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.54962289989454e-07,
+      "loss": 1.168,
+      "step": 26999
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.54520341581323e-07,
+      "loss": 0.958,
+      "step": 27000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.540784903364697e-07,
+      "loss": 1.0229,
+      "step": 27001
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.536367362596422e-07,
+      "loss": 1.0156,
+      "step": 27002
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.531950793555856e-07,
+      "loss": 0.8369,
+      "step": 27003
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.527535196290416e-07,
+      "loss": 0.9297,
+      "step": 27004
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.523120570847544e-07,
+      "loss": 0.8491,
+      "step": 27005
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.51870691727469e-07,
+      "loss": 1.2383,
+      "step": 27006
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.514294235619226e-07,
+      "loss": 0.8936,
+      "step": 27007
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.50988252592856e-07,
+      "loss": 0.96,
+      "step": 27008
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.505471788250087e-07,
+      "loss": 1.1123,
+      "step": 27009
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.501062022631203e-07,
+      "loss": 1.1123,
+      "step": 27010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.496653229119245e-07,
+      "loss": 0.7456,
+      "step": 27011
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.49224540776158e-07,
+      "loss": 0.8096,
+      "step": 27012
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.487838558605567e-07,
+      "loss": 0.7354,
+      "step": 27013
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.483432681698524e-07,
+      "loss": 0.585,
+      "step": 27014
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.479027777087801e-07,
+      "loss": 1.0273,
+      "step": 27015
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.474623844820685e-07,
+      "loss": 1.0327,
+      "step": 27016
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.470220884944525e-07,
+      "loss": 1.1333,
+      "step": 27017
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.465818897506573e-07,
+      "loss": 1.042,
+      "step": 27018
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.461417882554114e-07,
+      "loss": 1.1055,
+      "step": 27019
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.457017840134441e-07,
+      "loss": 0.875,
+      "step": 27020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.452618770294831e-07,
+      "loss": 0.9854,
+      "step": 27021
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.44822067308252e-07,
+      "loss": 0.8975,
+      "step": 27022
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.44382354854474e-07,
+      "loss": 0.8325,
+      "step": 27023
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.43942739672874e-07,
+      "loss": 0.9312,
+      "step": 27024
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.435032217681728e-07,
+      "loss": 1.0205,
+      "step": 27025
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.430638011450932e-07,
+      "loss": 0.9883,
+      "step": 27026
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.426244778083526e-07,
+      "loss": 1.2852,
+      "step": 27027
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.42185251762674e-07,
+      "loss": 0.8628,
+      "step": 27028
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.417461230127733e-07,
+      "loss": 0.7495,
+      "step": 27029
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.413070915633637e-07,
+      "loss": 1.0586,
+      "step": 27030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.408681574191691e-07,
+      "loss": 0.77,
+      "step": 27031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.404293205849003e-07,
+      "loss": 1.2827,
+      "step": 27032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.399905810652698e-07,
+      "loss": 0.918,
+      "step": 27033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.39551938864991e-07,
+      "loss": 0.9229,
+      "step": 27034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.391133939887754e-07,
+      "loss": 0.7241,
+      "step": 27035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.386749464413358e-07,
+      "loss": 1.2139,
+      "step": 27036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.382365962273798e-07,
+      "loss": 1.042,
+      "step": 27037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.377983433516181e-07,
+      "loss": 0.998,
+      "step": 27038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.373601878187566e-07,
+      "loss": 1.3301,
+      "step": 27039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.369221296335007e-07,
+      "loss": 1.002,
+      "step": 27040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.364841688005566e-07,
+      "loss": 0.8804,
+      "step": 27041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.360463053246316e-07,
+      "loss": 0.6553,
+      "step": 27042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.356085392104275e-07,
+      "loss": 0.8127,
+      "step": 27043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.35170870462645e-07,
+      "loss": 0.8335,
+      "step": 27044
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.347332990859837e-07,
+      "loss": 0.874,
+      "step": 27045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.342958250851486e-07,
+      "loss": 0.6934,
+      "step": 27046
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.338584484648383e-07,
+      "loss": 1.2285,
+      "step": 27047
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.334211692297468e-07,
+      "loss": 0.8906,
+      "step": 27048
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.329839873845759e-07,
+      "loss": 0.8633,
+      "step": 27049
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.325469029340195e-07,
+      "loss": 1.1846,
+      "step": 27050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.321099158827707e-07,
+      "loss": 0.5786,
+      "step": 27051
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.316730262355256e-07,
+      "loss": 1.1733,
+      "step": 27052
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.312362339969782e-07,
+      "loss": 1.2227,
+      "step": 27053
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.307995391718194e-07,
+      "loss": 0.7734,
+      "step": 27054
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.303629417647375e-07,
+      "loss": 0.9746,
+      "step": 27055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.299264417804254e-07,
+      "loss": 1.0215,
+      "step": 27056
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.294900392235728e-07,
+      "loss": 1.0791,
+      "step": 27057
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.290537340988647e-07,
+      "loss": 1.0186,
+      "step": 27058
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.286175264109865e-07,
+      "loss": 1.502,
+      "step": 27059
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.281814161646286e-07,
+      "loss": 0.9209,
+      "step": 27060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.277454033644706e-07,
+      "loss": 1.0654,
+      "step": 27061
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.27309488015199e-07,
+      "loss": 0.9365,
+      "step": 27062
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.268736701214942e-07,
+      "loss": 1.0137,
+      "step": 27063
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.264379496880415e-07,
+      "loss": 1.1514,
+      "step": 27064
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.260023267195173e-07,
+      "loss": 0.7969,
+      "step": 27065
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.25566801220601e-07,
+      "loss": 1.0645,
+      "step": 27066
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.251313731959711e-07,
+      "loss": 0.8066,
+      "step": 27067
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.246960426503082e-07,
+      "loss": 0.8408,
+      "step": 27068
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.242608095882865e-07,
+      "loss": 1.0562,
+      "step": 27069
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.238256740145779e-07,
+      "loss": 0.8896,
+      "step": 27070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.233906359338596e-07,
+      "loss": 1.0098,
+      "step": 27071
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.229556953508057e-07,
+      "loss": 1.1279,
+      "step": 27072
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.225208522700868e-07,
+      "loss": 0.7451,
+      "step": 27073
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.220861066963715e-07,
+      "loss": 0.8984,
+      "step": 27074
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.21651458634334e-07,
+      "loss": 0.9277,
+      "step": 27075
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.21216908088639e-07,
+      "loss": 0.7866,
+      "step": 27076
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.207824550639589e-07,
+      "loss": 1.0474,
+      "step": 27077
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.20348099564955e-07,
+      "loss": 1.7031,
+      "step": 27078
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.199138415962983e-07,
+      "loss": 0.8955,
+      "step": 27079
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.194796811626505e-07,
+      "loss": 0.8438,
+      "step": 27080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.190456182686736e-07,
+      "loss": 1.0625,
+      "step": 27081
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.186116529190325e-07,
+      "loss": 0.7393,
+      "step": 27082
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.181777851183904e-07,
+      "loss": 0.936,
+      "step": 27083
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.177440148714056e-07,
+      "loss": 0.998,
+      "step": 27084
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.173103421827356e-07,
+      "loss": 0.9463,
+      "step": 27085
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.168767670570433e-07,
+      "loss": 0.9746,
+      "step": 27086
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.164432894989806e-07,
+      "loss": 0.8721,
+      "step": 27087
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.160099095132092e-07,
+      "loss": 1.1699,
+      "step": 27088
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.155766271043798e-07,
+      "loss": 1.1436,
+      "step": 27089
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.151434422771499e-07,
+      "loss": 0.7397,
+      "step": 27090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.147103550361724e-07,
+      "loss": 1.0186,
+      "step": 27091
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.142773653860948e-07,
+      "loss": 1.0977,
+      "step": 27092
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.138444733315732e-07,
+      "loss": 0.8867,
+      "step": 27093
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.134116788772563e-07,
+      "loss": 0.7158,
+      "step": 27094
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.129789820277935e-07,
+      "loss": 0.9883,
+      "step": 27095
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.125463827878301e-07,
+      "loss": 1.0654,
+      "step": 27096
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.121138811620145e-07,
+      "loss": 0.9043,
+      "step": 27097
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.116814771549943e-07,
+      "loss": 1.0068,
+      "step": 27098
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.112491707714122e-07,
+      "loss": 0.7886,
+      "step": 27099
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.108169620159102e-07,
+      "loss": 1.2871,
+      "step": 27100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.103848508931357e-07,
+      "loss": 0.729,
+      "step": 27101
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.099528374077249e-07,
+      "loss": 0.8145,
+      "step": 27102
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.09520921564323e-07,
+      "loss": 0.9175,
+      "step": 27103
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.090891033675653e-07,
+      "loss": 0.7627,
+      "step": 27104
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.086573828220946e-07,
+      "loss": 1.124,
+      "step": 27105
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.082257599325461e-07,
+      "loss": 0.874,
+      "step": 27106
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.077942347035551e-07,
+      "loss": 1.0093,
+      "step": 27107
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.073628071397567e-07,
+      "loss": 1.0303,
+      "step": 27108
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.069314772457882e-07,
+      "loss": 1.1309,
+      "step": 27109
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.065002450262816e-07,
+      "loss": 1.1367,
+      "step": 27110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.060691104858655e-07,
+      "loss": 0.8691,
+      "step": 27111
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.05638073629177e-07,
+      "loss": 0.5098,
+      "step": 27112
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.052071344608404e-07,
+      "loss": 1.0342,
+      "step": 27113
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.047762929854886e-07,
+      "loss": 1.1416,
+      "step": 27114
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.043455492077468e-07,
+      "loss": 0.8904,
+      "step": 27115
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.039149031322459e-07,
+      "loss": 1.1934,
+      "step": 27116
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.034843547636085e-07,
+      "loss": 0.8115,
+      "step": 27117
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.030539041064567e-07,
+      "loss": 0.7144,
+      "step": 27118
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.026235511654202e-07,
+      "loss": 0.6807,
+      "step": 27119
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.021932959451207e-07,
+      "loss": 0.9102,
+      "step": 27120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.017631384501768e-07,
+      "loss": 1.0317,
+      "step": 27121
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.013330786852093e-07,
+      "loss": 0.7979,
+      "step": 27122
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.009031166548399e-07,
+      "loss": 1.1738,
+      "step": 27123
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.004732523636872e-07,
+      "loss": 0.8965,
+      "step": 27124
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.000434858163675e-07,
+      "loss": 0.7354,
+      "step": 27125
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.99613817017495e-07,
+      "loss": 0.8936,
+      "step": 27126
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.991842459716904e-07,
+      "loss": 1.0137,
+      "step": 27127
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.987547726835622e-07,
+      "loss": 1.0205,
+      "step": 27128
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.983253971577278e-07,
+      "loss": 0.8438,
+      "step": 27129
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.978961193987989e-07,
+      "loss": 1.0615,
+      "step": 27130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.974669394113855e-07,
+      "loss": 1.0632,
+      "step": 27131
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.970378572000981e-07,
+      "loss": 0.75,
+      "step": 27132
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.966088727695422e-07,
+      "loss": 0.8496,
+      "step": 27133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.961799861243325e-07,
+      "loss": 1.103,
+      "step": 27134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.957511972690735e-07,
+      "loss": 1.2051,
+      "step": 27135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.953225062083671e-07,
+      "loss": 0.7891,
+      "step": 27136
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.948939129468237e-07,
+      "loss": 0.6729,
+      "step": 27137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.944654174890432e-07,
+      "loss": 1.1113,
+      "step": 27138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.94037019839632e-07,
+      "loss": 0.9502,
+      "step": 27139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.936087200031874e-07,
+      "loss": 0.8379,
+      "step": 27140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.931805179843134e-07,
+      "loss": 0.7959,
+      "step": 27141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.927524137876098e-07,
+      "loss": 1.0654,
+      "step": 27142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.923244074176729e-07,
+      "loss": 0.7285,
+      "step": 27143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.918964988790968e-07,
+      "loss": 1.1143,
+      "step": 27144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.914686881764867e-07,
+      "loss": 1.0547,
+      "step": 27145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.910409753144344e-07,
+      "loss": 1.2061,
+      "step": 27146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.906133602975297e-07,
+      "loss": 1.3516,
+      "step": 27147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.901858431303723e-07,
+      "loss": 0.9434,
+      "step": 27148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.897584238175505e-07,
+      "loss": 0.9756,
+      "step": 27149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.893311023636575e-07,
+      "loss": 1.1152,
+      "step": 27150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.889038787732807e-07,
+      "loss": 0.8599,
+      "step": 27151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.884767530510141e-07,
+      "loss": 0.9385,
+      "step": 27152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.88049725201442e-07,
+      "loss": 0.8887,
+      "step": 27153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.876227952291494e-07,
+      "loss": 0.667,
+      "step": 27154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.871959631387261e-07,
+      "loss": 0.9009,
+      "step": 27155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.867692289347574e-07,
+      "loss": 1.0156,
+      "step": 27156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.86342592621825e-07,
+      "loss": 0.9243,
+      "step": 27157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.859160542045109e-07,
+      "loss": 1.0811,
+      "step": 27158
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.854896136873969e-07,
+      "loss": 1.0938,
+      "step": 27159
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.850632710750684e-07,
+      "loss": 1.1367,
+      "step": 27160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.846370263721004e-07,
+      "loss": 0.8496,
+      "step": 27161
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.842108795830718e-07,
+      "loss": 1.1016,
+      "step": 27162
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.837848307125618e-07,
+      "loss": 1.0488,
+      "step": 27163
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.833588797651438e-07,
+      "loss": 1.0928,
+      "step": 27164
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.829330267453972e-07,
+      "loss": 1.1582,
+      "step": 27165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.825072716578942e-07,
+      "loss": 1.123,
+      "step": 27166
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.820816145072086e-07,
+      "loss": 1.2539,
+      "step": 27167
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.816560552979137e-07,
+      "loss": 1.0366,
+      "step": 27168
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.812305940345767e-07,
+      "loss": 1.0137,
+      "step": 27169
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.808052307217707e-07,
+      "loss": 0.8447,
+      "step": 27170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.803799653640665e-07,
+      "loss": 0.8062,
+      "step": 27171
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.799547979660306e-07,
+      "loss": 0.9424,
+      "step": 27172
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.795297285322279e-07,
+      "loss": 0.7588,
+      "step": 27173
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.791047570672273e-07,
+      "loss": 0.8999,
+      "step": 27174
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.786798835755916e-07,
+      "loss": 0.6699,
+      "step": 27175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.782551080618862e-07,
+      "loss": 0.729,
+      "step": 27176
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.778304305306729e-07,
+      "loss": 1.1494,
+      "step": 27177
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.774058509865146e-07,
+      "loss": 1.0752,
+      "step": 27178
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.769813694339713e-07,
+      "loss": 0.8872,
+      "step": 27179
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.765569858776002e-07,
+      "loss": 1.0361,
+      "step": 27180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.761327003219622e-07,
+      "loss": 1.1924,
+      "step": 27181
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.757085127716169e-07,
+      "loss": 1.0586,
+      "step": 27182
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.752844232311186e-07,
+      "loss": 1.0713,
+      "step": 27183
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.748604317050213e-07,
+      "loss": 0.8359,
+      "step": 27184
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.744365381978804e-07,
+      "loss": 1.1816,
+      "step": 27185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.740127427142519e-07,
+      "loss": 0.6382,
+      "step": 27186
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.735890452586848e-07,
+      "loss": 0.8521,
+      "step": 27187
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.731654458357309e-07,
+      "loss": 0.7573,
+      "step": 27188
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.727419444499419e-07,
+      "loss": 0.9912,
+      "step": 27189
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.723185411058633e-07,
+      "loss": 0.9316,
+      "step": 27190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.718952358080479e-07,
+      "loss": 0.9238,
+      "step": 27191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.71472028561039e-07,
+      "loss": 1.1211,
+      "step": 27192
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.710489193693849e-07,
+      "loss": 0.7109,
+      "step": 27193
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.7062590823763e-07,
+      "loss": 1.0352,
+      "step": 27194
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.702029951703161e-07,
+      "loss": 0.9785,
+      "step": 27195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.697801801719874e-07,
+      "loss": 0.6748,
+      "step": 27196
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.693574632471868e-07,
+      "loss": 1.0332,
+      "step": 27197
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.689348444004541e-07,
+      "loss": 1.166,
+      "step": 27198
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.685123236363279e-07,
+      "loss": 0.873,
+      "step": 27199
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.680899009593469e-07,
+      "loss": 1.001,
+      "step": 27200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.676675763740505e-07,
+      "loss": 0.5815,
+      "step": 27201
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.672453498849754e-07,
+      "loss": 1.0669,
+      "step": 27202
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.668232214966532e-07,
+      "loss": 0.9258,
+      "step": 27203
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.664011912136217e-07,
+      "loss": 0.7617,
+      "step": 27204
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.659792590404137e-07,
+      "loss": 0.79,
+      "step": 27205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.65557424981559e-07,
+      "loss": 1.0332,
+      "step": 27206
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.651356890415908e-07,
+      "loss": 1.0718,
+      "step": 27207
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.647140512250407e-07,
+      "loss": 1.1836,
+      "step": 27208
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.642925115364364e-07,
+      "loss": 0.8506,
+      "step": 27209
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.638710699803043e-07,
+      "loss": 0.9971,
+      "step": 27210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.634497265611719e-07,
+      "loss": 0.9766,
+      "step": 27211
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.630284812835677e-07,
+      "loss": 0.8442,
+      "step": 27212
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.626073341520158e-07,
+      "loss": 0.916,
+      "step": 27213
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.621862851710372e-07,
+      "loss": 0.8037,
+      "step": 27214
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.617653343451571e-07,
+      "loss": 1.4668,
+      "step": 27215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.613444816788963e-07,
+      "loss": 0.8184,
+      "step": 27216
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.609237271767767e-07,
+      "loss": 0.9028,
+      "step": 27217
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.605030708433149e-07,
+      "loss": 0.9844,
+      "step": 27218
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.600825126830336e-07,
+      "loss": 1.0234,
+      "step": 27219
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.596620527004485e-07,
+      "loss": 0.7134,
+      "step": 27220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.592416909000722e-07,
+      "loss": 1.1602,
+      "step": 27221
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.58821427286427e-07,
+      "loss": 1.001,
+      "step": 27222
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.584012618640236e-07,
+      "loss": 1.0967,
+      "step": 27223
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.579811946373761e-07,
+      "loss": 0.8711,
+      "step": 27224
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.575612256109934e-07,
+      "loss": 0.6689,
+      "step": 27225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.571413547893903e-07,
+      "loss": 0.7627,
+      "step": 27226
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.567215821770769e-07,
+      "loss": 1.1152,
+      "step": 27227
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.563019077785606e-07,
+      "loss": 1.0044,
+      "step": 27228
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.558823315983511e-07,
+      "loss": 0.9111,
+      "step": 27229
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.554628536409537e-07,
+      "loss": 1.123,
+      "step": 27230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.550434739108737e-07,
+      "loss": 1.0396,
+      "step": 27231
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.546241924126175e-07,
+      "loss": 0.9473,
+      "step": 27232
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.542050091506904e-07,
+      "loss": 0.9014,
+      "step": 27233
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.537859241295921e-07,
+      "loss": 1.1084,
+      "step": 27234
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.533669373538245e-07,
+      "loss": 0.8066,
+      "step": 27235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.529480488278908e-07,
+      "loss": 0.7915,
+      "step": 27236
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.525292585562872e-07,
+      "loss": 1.2793,
+      "step": 27237
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.521105665435147e-07,
+      "loss": 1.001,
+      "step": 27238
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.516919727940686e-07,
+      "loss": 0.7542,
+      "step": 27239
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.512734773124487e-07,
+      "loss": 0.9033,
+      "step": 27240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.508550801031478e-07,
+      "loss": 0.8228,
+      "step": 27241
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.504367811706582e-07,
+      "loss": 0.9062,
+      "step": 27242
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.500185805194761e-07,
+      "loss": 0.9805,
+      "step": 27243
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.496004781540956e-07,
+      "loss": 0.9326,
+      "step": 27244
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.491824740790045e-07,
+      "loss": 0.6699,
+      "step": 27245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.487645682986912e-07,
+      "loss": 1.375,
+      "step": 27246
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.483467608176477e-07,
+      "loss": 1.0723,
+      "step": 27247
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.479290516403626e-07,
+      "loss": 1.1436,
+      "step": 27248
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.475114407713225e-07,
+      "loss": 0.5469,
+      "step": 27249
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.470939282150093e-07,
+      "loss": 0.7314,
+      "step": 27250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.466765139759125e-07,
+      "loss": 0.7983,
+      "step": 27251
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.462591980585133e-07,
+      "loss": 0.96,
+      "step": 27252
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.458419804672957e-07,
+      "loss": 1.2422,
+      "step": 27253
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.454248612067395e-07,
+      "loss": 1.124,
+      "step": 27254
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.450078402813277e-07,
+      "loss": 0.6296,
+      "step": 27255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.445909176955391e-07,
+      "loss": 0.813,
+      "step": 27256
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.44174093453849e-07,
+      "loss": 1.0322,
+      "step": 27257
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.43757367560738e-07,
+      "loss": 0.9326,
+      "step": 27258
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.433407400206828e-07,
+      "loss": 1.291,
+      "step": 27259
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.429242108381585e-07,
+      "loss": 0.9062,
+      "step": 27260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.425077800176362e-07,
+      "loss": 0.9014,
+      "step": 27261
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.420914475635911e-07,
+      "loss": 0.9785,
+      "step": 27262
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.416752134804962e-07,
+      "loss": 0.918,
+      "step": 27263
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.412590777728235e-07,
+      "loss": 0.7607,
+      "step": 27264
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.408430404450385e-07,
+      "loss": 0.9255,
+      "step": 27265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.404271015016152e-07,
+      "loss": 0.7686,
+      "step": 27266
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.400112609470179e-07,
+      "loss": 1.0996,
+      "step": 27267
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.395955187857141e-07,
+      "loss": 0.6328,
+      "step": 27268
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.391798750221702e-07,
+      "loss": 1.1519,
+      "step": 27269
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.387643296608516e-07,
+      "loss": 1.2949,
+      "step": 27270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.383488827062214e-07,
+      "loss": 1.2891,
+      "step": 27271
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.379335341627404e-07,
+      "loss": 1.4883,
+      "step": 27272
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.375182840348717e-07,
+      "loss": 1.2256,
+      "step": 27273
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.371031323270773e-07,
+      "loss": 0.6963,
+      "step": 27274
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.366880790438148e-07,
+      "loss": 0.9067,
+      "step": 27275
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.362731241895417e-07,
+      "loss": 0.8604,
+      "step": 27276
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.358582677687166e-07,
+      "loss": 0.9438,
+      "step": 27277
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.354435097857949e-07,
+      "loss": 1.3223,
+      "step": 27278
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.350288502452341e-07,
+      "loss": 0.8726,
+      "step": 27279
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.346142891514853e-07,
+      "loss": 1.0586,
+      "step": 27280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.341998265090035e-07,
+      "loss": 1.0762,
+      "step": 27281
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.337854623222408e-07,
+      "loss": 0.8594,
+      "step": 27282
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.333711965956448e-07,
+      "loss": 0.7734,
+      "step": 27283
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.329570293336686e-07,
+      "loss": 1.2051,
+      "step": 27284
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.325429605407632e-07,
+      "loss": 1.0312,
+      "step": 27285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.321289902213726e-07,
+      "loss": 1.0029,
+      "step": 27286
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.317151183799421e-07,
+      "loss": 0.9238,
+      "step": 27287
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.313013450209206e-07,
+      "loss": 1.0254,
+      "step": 27288
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.308876701487534e-07,
+      "loss": 1.0486,
+      "step": 27289
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.304740937678835e-07,
+      "loss": 0.9932,
+      "step": 27290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.300606158827495e-07,
+      "loss": 0.7446,
+      "step": 27291
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.296472364977981e-07,
+      "loss": 0.8721,
+      "step": 27292
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.292339556174678e-07,
+      "loss": 0.7451,
+      "step": 27293
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.288207732461951e-07,
+      "loss": 1.1602,
+      "step": 27294
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.284076893884197e-07,
+      "loss": 0.875,
+      "step": 27295
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.279947040485825e-07,
+      "loss": 0.7505,
+      "step": 27296
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.275818172311168e-07,
+      "loss": 0.9839,
+      "step": 27297
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.271690289404565e-07,
+      "loss": 1.6387,
+      "step": 27298
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.267563391810362e-07,
+      "loss": 0.9404,
+      "step": 27299
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.263437479572911e-07,
+      "loss": 0.6304,
+      "step": 27300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.259312552736521e-07,
+      "loss": 1.0459,
+      "step": 27301
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.255188611345477e-07,
+      "loss": 1.3809,
+      "step": 27302
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.251065655444101e-07,
+      "loss": 1.0596,
+      "step": 27303
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.246943685076669e-07,
+      "loss": 0.7998,
+      "step": 27304
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.242822700287478e-07,
+      "loss": 0.7837,
+      "step": 27305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.23870270112076e-07,
+      "loss": 1.0273,
+      "step": 27306
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.234583687620801e-07,
+      "loss": 0.9004,
+      "step": 27307
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.230465659831843e-07,
+      "loss": 1.0547,
+      "step": 27308
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.226348617798074e-07,
+      "loss": 0.9209,
+      "step": 27309
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.222232561563792e-07,
+      "loss": 0.8076,
+      "step": 27310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.218117491173172e-07,
+      "loss": 0.9873,
+      "step": 27311
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.214003406670413e-07,
+      "loss": 1.1699,
+      "step": 27312
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.209890308099699e-07,
+      "loss": 0.8037,
+      "step": 27313
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.205778195505221e-07,
+      "loss": 1.0391,
+      "step": 27314
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.201667068931174e-07,
+      "loss": 0.8584,
+      "step": 27315
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.197556928421702e-07,
+      "loss": 0.8828,
+      "step": 27316
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.193447774020935e-07,
+      "loss": 1.083,
+      "step": 27317
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.189339605773039e-07,
+      "loss": 0.667,
+      "step": 27318
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.185232423722112e-07,
+      "loss": 1.1611,
+      "step": 27319
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.181126227912295e-07,
+      "loss": 1.1094,
+      "step": 27320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.177021018387698e-07,
+      "loss": 1.042,
+      "step": 27321
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.172916795192421e-07,
+      "loss": 0.8032,
+      "step": 27322
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.168813558370536e-07,
+      "loss": 1.0059,
+      "step": 27323
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.164711307966111e-07,
+      "loss": 1.0234,
+      "step": 27324
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.16061004402322e-07,
+      "loss": 0.9463,
+      "step": 27325
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.156509766585929e-07,
+      "loss": 0.9941,
+      "step": 27326
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.152410475698258e-07,
+      "loss": 1.2471,
+      "step": 27327
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.148312171404282e-07,
+      "loss": 0.9619,
+      "step": 27328
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.144214853747989e-07,
+      "loss": 1.0566,
+      "step": 27329
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.140118522773377e-07,
+      "loss": 0.7964,
+      "step": 27330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.136023178524466e-07,
+      "loss": 1.0137,
+      "step": 27331
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.131928821045276e-07,
+      "loss": 1.0459,
+      "step": 27332
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.12783545037975e-07,
+      "loss": 0.9824,
+      "step": 27333
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.123743066571855e-07,
+      "loss": 1.4121,
+      "step": 27334
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.119651669665563e-07,
+      "loss": 0.8989,
+      "step": 27335
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.11556125970484e-07,
+      "loss": 1.5098,
+      "step": 27336
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.111471836733598e-07,
+      "loss": 1.3164,
+      "step": 27337
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.107383400795754e-07,
+      "loss": 0.7031,
+      "step": 27338
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.103295951935264e-07,
+      "loss": 1.2676,
+      "step": 27339
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.099209490196003e-07,
+      "loss": 1.0,
+      "step": 27340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.09512401562188e-07,
+      "loss": 0.9873,
+      "step": 27341
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.091039528256761e-07,
+      "loss": 0.7366,
+      "step": 27342
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.086956028144554e-07,
+      "loss": 0.917,
+      "step": 27343
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.082873515329093e-07,
+      "loss": 1.0693,
+      "step": 27344
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.078791989854228e-07,
+      "loss": 1.1436,
+      "step": 27345
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.074711451763817e-07,
+      "loss": 1.1045,
+      "step": 27346
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.070631901101711e-07,
+      "loss": 0.9805,
+      "step": 27347
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.066553337911698e-07,
+      "loss": 0.8115,
+      "step": 27348
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.062475762237587e-07,
+      "loss": 0.7427,
+      "step": 27349
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.058399174123177e-07,
+      "loss": 0.8066,
+      "step": 27350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.054323573612299e-07,
+      "loss": 0.8711,
+      "step": 27351
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.050248960748707e-07,
+      "loss": 1.043,
+      "step": 27352
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.046175335576134e-07,
+      "loss": 0.9409,
+      "step": 27353
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.042102698138399e-07,
+      "loss": 1.1836,
+      "step": 27354
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.0380310484792e-07,
+      "loss": 0.7734,
+      "step": 27355
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.033960386642292e-07,
+      "loss": 1.2598,
+      "step": 27356
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.029890712671385e-07,
+      "loss": 0.9155,
+      "step": 27357
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.02582202661023e-07,
+      "loss": 1.3389,
+      "step": 27358
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.021754328502518e-07,
+      "loss": 0.9629,
+      "step": 27359
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.017687618391912e-07,
+      "loss": 0.8516,
+      "step": 27360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.01362189632211e-07,
+      "loss": 1.1123,
+      "step": 27361
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.009557162336823e-07,
+      "loss": 1.0234,
+      "step": 27362
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.00549341647967e-07,
+      "loss": 0.5698,
+      "step": 27363
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.001430658794307e-07,
+      "loss": 1.1299,
+      "step": 27364
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.997368889324386e-07,
+      "loss": 0.8652,
+      "step": 27365
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.993308108113529e-07,
+      "loss": 1.269,
+      "step": 27366
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.989248315205366e-07,
+      "loss": 0.7539,
+      "step": 27367
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.985189510643488e-07,
+      "loss": 0.5439,
+      "step": 27368
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.981131694471522e-07,
+      "loss": 1.0303,
+      "step": 27369
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.977074866733037e-07,
+      "loss": 0.814,
+      "step": 27370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.973019027471596e-07,
+      "loss": 0.915,
+      "step": 27371
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.968964176730787e-07,
+      "loss": 1.0605,
+      "step": 27372
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.964910314554164e-07,
+      "loss": 0.9531,
+      "step": 27373
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.96085744098527e-07,
+      "loss": 0.9185,
+      "step": 27374
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.956805556067626e-07,
+      "loss": 0.7705,
+      "step": 27375
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.952754659844764e-07,
+      "loss": 1.2656,
+      "step": 27376
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.948704752360214e-07,
+      "loss": 1.1387,
+      "step": 27377
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.944655833657467e-07,
+      "loss": 0.9814,
+      "step": 27378
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.940607903779996e-07,
+      "loss": 0.918,
+      "step": 27379
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.936560962771312e-07,
+      "loss": 1.2637,
+      "step": 27380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.932515010674857e-07,
+      "loss": 0.9912,
+      "step": 27381
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.928470047534121e-07,
+      "loss": 0.8433,
+      "step": 27382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.924426073392533e-07,
+      "loss": 0.7275,
+      "step": 27383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.920383088293549e-07,
+      "loss": 0.9043,
+      "step": 27384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.916341092280588e-07,
+      "loss": 1.0205,
+      "step": 27385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.912300085397051e-07,
+      "loss": 1.249,
+      "step": 27386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.908260067686357e-07,
+      "loss": 1.1357,
+      "step": 27387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.904221039191929e-07,
+      "loss": 1.0127,
+      "step": 27388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.900182999957117e-07,
+      "loss": 0.812,
+      "step": 27389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.896145950025302e-07,
+      "loss": 0.8369,
+      "step": 27390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.892109889439869e-07,
+      "loss": 0.73,
+      "step": 27391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.88807481824414e-07,
+      "loss": 0.7383,
+      "step": 27392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.884040736481502e-07,
+      "loss": 0.7197,
+      "step": 27393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.880007644195242e-07,
+      "loss": 0.7563,
+      "step": 27394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.875975541428715e-07,
+      "loss": 0.9546,
+      "step": 27395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.87194442822522e-07,
+      "loss": 0.8364,
+      "step": 27396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.867914304628044e-07,
+      "loss": 0.772,
+      "step": 27397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.863885170680486e-07,
+      "loss": 0.6758,
+      "step": 27398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.859857026425844e-07,
+      "loss": 1.3271,
+      "step": 27399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.855829871907383e-07,
+      "loss": 1.0752,
+      "step": 27400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.851803707168326e-07,
+      "loss": 0.7405,
+      "step": 27401
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.847778532251948e-07,
+      "loss": 1.1699,
+      "step": 27402
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.843754347201505e-07,
+      "loss": 1.1489,
+      "step": 27403
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.839731152060193e-07,
+      "loss": 0.8965,
+      "step": 27404
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.835708946871235e-07,
+      "loss": 0.7395,
+      "step": 27405
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.83168773167784e-07,
+      "loss": 0.7305,
+      "step": 27406
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.827667506523184e-07,
+      "loss": 1.1494,
+      "step": 27407
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.823648271450479e-07,
+      "loss": 0.8086,
+      "step": 27408
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.819630026502889e-07,
+      "loss": 1.2031,
+      "step": 27409
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.815612771723591e-07,
+      "loss": 1.1289,
+      "step": 27410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.811596507155706e-07,
+      "loss": 1.1904,
+      "step": 27411
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.807581232842354e-07,
+      "loss": 0.8467,
+      "step": 27412
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.803566948826746e-07,
+      "loss": 1.0615,
+      "step": 27413
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.799553655151948e-07,
+      "loss": 1.2646,
+      "step": 27414
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.79554135186108e-07,
+      "loss": 0.9292,
+      "step": 27415
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.79153003899722e-07,
+      "loss": 0.9678,
+      "step": 27416
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.787519716603487e-07,
+      "loss": 0.8857,
+      "step": 27417
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.783510384722937e-07,
+      "loss": 1.0195,
+      "step": 27418
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.779502043398646e-07,
+      "loss": 1.1396,
+      "step": 27419
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.775494692673679e-07,
+      "loss": 1.2578,
+      "step": 27420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.771488332591082e-07,
+      "loss": 0.7153,
+      "step": 27421
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.767482963193885e-07,
+      "loss": 1.1689,
+      "step": 27422
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.763478584525063e-07,
+      "loss": 0.7725,
+      "step": 27423
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.759475196627708e-07,
+      "loss": 1.0518,
+      "step": 27424
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.755472799544794e-07,
+      "loss": 0.9971,
+      "step": 27425
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.751471393319287e-07,
+      "loss": 1.0469,
+      "step": 27426
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.747470977994209e-07,
+      "loss": 1.2041,
+      "step": 27427
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.743471553612502e-07,
+      "loss": 0.6704,
+      "step": 27428
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.739473120217144e-07,
+      "loss": 0.9268,
+      "step": 27429
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.735475677851067e-07,
+      "loss": 0.96,
+      "step": 27430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.731479226557237e-07,
+      "loss": 0.9375,
+      "step": 27431
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.727483766378563e-07,
+      "loss": 0.9316,
+      "step": 27432
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.723489297357955e-07,
+      "loss": 1.0049,
+      "step": 27433
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.719495819538325e-07,
+      "loss": 0.7632,
+      "step": 27434
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.715503332962593e-07,
+      "loss": 1.2412,
+      "step": 27435
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.711511837673624e-07,
+      "loss": 0.9648,
+      "step": 27436
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.707521333714285e-07,
+      "loss": 0.7656,
+      "step": 27437
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.703531821127442e-07,
+      "loss": 0.8789,
+      "step": 27438
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.699543299955992e-07,
+      "loss": 0.8857,
+      "step": 27439
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.695555770242735e-07,
+      "loss": 0.8423,
+      "step": 27440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.691569232030494e-07,
+      "loss": 0.9541,
+      "step": 27441
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.687583685362132e-07,
+      "loss": 0.9028,
+      "step": 27442
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.683599130280428e-07,
+      "loss": 1.418,
+      "step": 27443
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.679615566828202e-07,
+      "loss": 0.8418,
+      "step": 27444
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.675632995048221e-07,
+      "loss": 1.6328,
+      "step": 27445
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.671651414983294e-07,
+      "loss": 0.9414,
+      "step": 27446
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.667670826676176e-07,
+      "loss": 1.0264,
+      "step": 27447
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.663691230169612e-07,
+      "loss": 0.7959,
+      "step": 27448
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.659712625506355e-07,
+      "loss": 1.1123,
+      "step": 27449
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.65573501272916e-07,
+      "loss": 0.9961,
+      "step": 27450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.651758391880748e-07,
+      "loss": 0.71,
+      "step": 27451
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.647782763003808e-07,
+      "loss": 0.7241,
+      "step": 27452
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.643808126141083e-07,
+      "loss": 0.7485,
+      "step": 27453
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.639834481335229e-07,
+      "loss": 1.2217,
+      "step": 27454
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.635861828628976e-07,
+      "loss": 0.7852,
+      "step": 27455
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.631890168064938e-07,
+      "loss": 0.7324,
+      "step": 27456
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.627919499685832e-07,
+      "loss": 0.7268,
+      "step": 27457
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.623949823534282e-07,
+      "loss": 1.04,
+      "step": 27458
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.619981139652921e-07,
+      "loss": 1.0557,
+      "step": 27459
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.616013448084392e-07,
+      "loss": 0.9287,
+      "step": 27460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.612046748871327e-07,
+      "loss": 0.8433,
+      "step": 27461
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.608081042056325e-07,
+      "loss": 1.166,
+      "step": 27462
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.604116327681965e-07,
+      "loss": 0.8486,
+      "step": 27463
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.600152605790845e-07,
+      "loss": 1.1577,
+      "step": 27464
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.596189876425564e-07,
+      "loss": 1.1445,
+      "step": 27465
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.592228139628677e-07,
+      "loss": 0.8447,
+      "step": 27466
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.588267395442706e-07,
+      "loss": 1.7441,
+      "step": 27467
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.584307643910249e-07,
+      "loss": 1.1826,
+      "step": 27468
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.580348885073796e-07,
+      "loss": 0.9746,
+      "step": 27469
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.576391118975912e-07,
+      "loss": 1.1992,
+      "step": 27470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.572434345659064e-07,
+      "loss": 1.0674,
+      "step": 27471
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.568478565165793e-07,
+      "loss": 1.1914,
+      "step": 27472
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.56452377753858e-07,
+      "loss": 0.9678,
+      "step": 27473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.560569982819877e-07,
+      "loss": 0.9102,
+      "step": 27474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.556617181052184e-07,
+      "loss": 0.96,
+      "step": 27475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.552665372277978e-07,
+      "loss": 0.696,
+      "step": 27476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.548714556539671e-07,
+      "loss": 1.5156,
+      "step": 27477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.544764733879706e-07,
+      "loss": 0.9209,
+      "step": 27478
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.540815904340537e-07,
+      "loss": 1.042,
+      "step": 27479
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.536868067964553e-07,
+      "loss": 1.248,
+      "step": 27480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.532921224794176e-07,
+      "loss": 1.1348,
+      "step": 27481
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.528975374871783e-07,
+      "loss": 1.1299,
+      "step": 27482
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.525030518239795e-07,
+      "loss": 1.0449,
+      "step": 27483
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.521086654940557e-07,
+      "loss": 0.9141,
+      "step": 27484
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.517143785016423e-07,
+      "loss": 0.8215,
+      "step": 27485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.51320190850976e-07,
+      "loss": 1.0928,
+      "step": 27486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.509261025462933e-07,
+      "loss": 0.9106,
+      "step": 27487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.505321135918242e-07,
+      "loss": 0.6001,
+      "step": 27488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.501382239918021e-07,
+      "loss": 1.2832,
+      "step": 27489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.497444337504556e-07,
+      "loss": 0.8521,
+      "step": 27490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.493507428720203e-07,
+      "loss": 0.8735,
+      "step": 27491
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.489571513607208e-07,
+      "loss": 1.5723,
+      "step": 27492
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.485636592207845e-07,
+      "loss": 0.9292,
+      "step": 27493
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.481702664564406e-07,
+      "loss": 0.9854,
+      "step": 27494
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.47776973071912e-07,
+      "loss": 0.8867,
+      "step": 27495
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.473837790714266e-07,
+      "loss": 1.1787,
+      "step": 27496
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.469906844592045e-07,
+      "loss": 0.7539,
+      "step": 27497
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.465976892394711e-07,
+      "loss": 0.9326,
+      "step": 27498
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.462047934164473e-07,
+      "loss": 0.8965,
+      "step": 27499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.458119969943489e-07,
+      "loss": 1.1455,
+      "step": 27500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.454192999774035e-07,
+      "loss": 0.7979,
+      "step": 27501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.450267023698232e-07,
+      "loss": 1.0205,
+      "step": 27502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.446342041758281e-07,
+      "loss": 1.1631,
+      "step": 27503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.442418053996303e-07,
+      "loss": 1.3789,
+      "step": 27504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.438495060454487e-07,
+      "loss": 0.8623,
+      "step": 27505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.434573061174966e-07,
+      "loss": 0.8809,
+      "step": 27506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.430652056199849e-07,
+      "loss": 1.1768,
+      "step": 27507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.426732045571283e-07,
+      "loss": 1.0322,
+      "step": 27508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.422813029331366e-07,
+      "loss": 1.0508,
+      "step": 27509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.418895007522187e-07,
+      "loss": 0.9248,
+      "step": 27510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.414977980185801e-07,
+      "loss": 1.0127,
+      "step": 27511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.41106194736434e-07,
+      "loss": 0.998,
+      "step": 27512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.40714690909986e-07,
+      "loss": 0.8184,
+      "step": 27513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.403232865434385e-07,
+      "loss": 1.1338,
+      "step": 27514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.399319816409967e-07,
+      "loss": 0.9746,
+      "step": 27515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.395407762068641e-07,
+      "loss": 1.1631,
+      "step": 27516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.39149670245245e-07,
+      "loss": 0.6987,
+      "step": 27517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.387586637603362e-07,
+      "loss": 1.0977,
+      "step": 27518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.38367756756343e-07,
+      "loss": 1.3574,
+      "step": 27519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.379769492374611e-07,
+      "loss": 0.793,
+      "step": 27520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.375862412078882e-07,
+      "loss": 1.1777,
+      "step": 27521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.371956326718211e-07,
+      "loss": 0.9844,
+      "step": 27522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.368051236334584e-07,
+      "loss": 0.8281,
+      "step": 27523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.364147140969924e-07,
+      "loss": 0.9951,
+      "step": 27524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.360244040666154e-07,
+      "loss": 1.2695,
+      "step": 27525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.356341935465228e-07,
+      "loss": 1.0293,
+      "step": 27526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.352440825409057e-07,
+      "loss": 0.9414,
+      "step": 27527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.348540710539542e-07,
+      "loss": 0.8291,
+      "step": 27528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.34464159089856e-07,
+      "loss": 1.3281,
+      "step": 27529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.340743466528021e-07,
+      "loss": 1.1465,
+      "step": 27530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.33684633746976e-07,
+      "loss": 1.1641,
+      "step": 27531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.332950203765688e-07,
+      "loss": 1.0234,
+      "step": 27532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.329055065457613e-07,
+      "loss": 1.3105,
+      "step": 27533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.325160922587404e-07,
+      "loss": 0.7969,
+      "step": 27534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.321267775196883e-07,
+      "loss": 1.2793,
+      "step": 27535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.317375623327838e-07,
+      "loss": 1.2842,
+      "step": 27536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.313484467022114e-07,
+      "loss": 0.8628,
+      "step": 27537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.30959430632151e-07,
+      "loss": 1.0952,
+      "step": 27538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.305705141267805e-07,
+      "loss": 1.1992,
+      "step": 27539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.301816971902754e-07,
+      "loss": 0.7512,
+      "step": 27540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.297929798268144e-07,
+      "loss": 0.7861,
+      "step": 27541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.294043620405721e-07,
+      "loss": 0.8975,
+      "step": 27542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.290158438357242e-07,
+      "loss": 1.3135,
+      "step": 27543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.286274252164415e-07,
+      "loss": 0.8198,
+      "step": 27544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.282391061868999e-07,
+      "loss": 0.9648,
+      "step": 27545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.278508867512679e-07,
+      "loss": 0.9648,
+      "step": 27546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.274627669137146e-07,
+      "loss": 1.0928,
+      "step": 27547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.270747466784112e-07,
+      "loss": 0.9102,
+      "step": 27548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.266868260495264e-07,
+      "loss": 0.7007,
+      "step": 27549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.262990050312258e-07,
+      "loss": 0.8623,
+      "step": 27550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.259112836276728e-07,
+      "loss": 0.9487,
+      "step": 27551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.255236618430362e-07,
+      "loss": 0.8772,
+      "step": 27552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.251361396814782e-07,
+      "loss": 0.6831,
+      "step": 27553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.247487171471624e-07,
+      "loss": 0.6953,
+      "step": 27554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.243613942442474e-07,
+      "loss": 1.2471,
+      "step": 27555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.239741709768977e-07,
+      "loss": 0.9121,
+      "step": 27556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.235870473492679e-07,
+      "loss": 1.1475,
+      "step": 27557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.232000233655212e-07,
+      "loss": 0.9424,
+      "step": 27558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.22813099029811e-07,
+      "loss": 0.9609,
+      "step": 27559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.224262743462973e-07,
+      "loss": 0.791,
+      "step": 27560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.220395493191323e-07,
+      "loss": 1.062,
+      "step": 27561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.216529239524706e-07,
+      "loss": 1.1113,
+      "step": 27562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.212663982504642e-07,
+      "loss": 1.1934,
+      "step": 27563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.208799722172677e-07,
+      "loss": 0.7529,
+      "step": 27564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.204936458570322e-07,
+      "loss": 1.1094,
+      "step": 27565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.201074191739021e-07,
+      "loss": 0.9297,
+      "step": 27566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.19721292172032e-07,
+      "loss": 1.1816,
+      "step": 27567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.193352648555674e-07,
+      "loss": 0.8848,
+      "step": 27568
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.189493372286549e-07,
+      "loss": 0.873,
+      "step": 27569
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.18563509295439e-07,
+      "loss": 1.1533,
+      "step": 27570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.181777810600665e-07,
+      "loss": 0.9336,
+      "step": 27571
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.177921525266795e-07,
+      "loss": 0.9565,
+      "step": 27572
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.174066236994182e-07,
+      "loss": 1.1602,
+      "step": 27573
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.170211945824257e-07,
+      "loss": 0.6907,
+      "step": 27574
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.166358651798444e-07,
+      "loss": 1.3594,
+      "step": 27575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.162506354958099e-07,
+      "loss": 1.1162,
+      "step": 27576
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.15865505534461e-07,
+      "loss": 0.9648,
+      "step": 27577
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.154804752999344e-07,
+      "loss": 0.7646,
+      "step": 27578
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.150955447963681e-07,
+      "loss": 0.958,
+      "step": 27579
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.147107140278964e-07,
+      "loss": 0.7476,
+      "step": 27580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.143259829986493e-07,
+      "loss": 0.5195,
+      "step": 27581
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.139413517127635e-07,
+      "loss": 1.3613,
+      "step": 27582
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.135568201743681e-07,
+      "loss": 0.8114,
+      "step": 27583
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.131723883875974e-07,
+      "loss": 0.7021,
+      "step": 27584
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.127880563565748e-07,
+      "loss": 0.7227,
+      "step": 27585
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.124038240854348e-07,
+      "loss": 1.2549,
+      "step": 27586
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.120196915783018e-07,
+      "loss": 1.3125,
+      "step": 27587
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.116356588392992e-07,
+      "loss": 1.1196,
+      "step": 27588
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.112517258725559e-07,
+      "loss": 1.0771,
+      "step": 27589
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.108678926821966e-07,
+      "loss": 1.0459,
+      "step": 27590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.104841592723433e-07,
+      "loss": 0.9224,
+      "step": 27591
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.101005256471161e-07,
+      "loss": 0.8896,
+      "step": 27592
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.097169918106361e-07,
+      "loss": 0.9053,
+      "step": 27593
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.093335577670257e-07,
+      "loss": 1.2734,
+      "step": 27594
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.089502235204026e-07,
+      "loss": 0.7776,
+      "step": 27595
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.085669890748826e-07,
+      "loss": 0.7319,
+      "step": 27596
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.081838544345853e-07,
+      "loss": 1.4121,
+      "step": 27597
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.078008196036224e-07,
+      "loss": 0.9385,
+      "step": 27598
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.074178845861101e-07,
+      "loss": 1.0068,
+      "step": 27599
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.070350493861644e-07,
+      "loss": 0.897,
+      "step": 27600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.06652314007894e-07,
+      "loss": 0.8989,
+      "step": 27601
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.062696784554124e-07,
+      "loss": 0.7222,
+      "step": 27602
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.058871427328273e-07,
+      "loss": 0.9922,
+      "step": 27603
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.055047068442478e-07,
+      "loss": 0.8271,
+      "step": 27604
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.05122370793786e-07,
+      "loss": 1.1055,
+      "step": 27605
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.047401345855442e-07,
+      "loss": 0.769,
+      "step": 27606
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.043579982236315e-07,
+      "loss": 0.7227,
+      "step": 27607
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.039759617121511e-07,
+      "loss": 1.0635,
+      "step": 27608
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.035940250552053e-07,
+      "loss": 1.3486,
+      "step": 27609
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.032121882568976e-07,
+      "loss": 1.0332,
+      "step": 27610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.028304513213336e-07,
+      "loss": 0.8447,
+      "step": 27611
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.024488142526087e-07,
+      "loss": 0.6777,
+      "step": 27612
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.020672770548254e-07,
+      "loss": 1.0249,
+      "step": 27613
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.016858397320758e-07,
+      "loss": 1.0508,
+      "step": 27614
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.013045022884668e-07,
+      "loss": 0.4736,
+      "step": 27615
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.009232647280884e-07,
+      "loss": 0.96,
+      "step": 27616
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.005421270550361e-07,
+      "loss": 0.7812,
+      "step": 27617
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.001610892734067e-07,
+      "loss": 1.2998,
+      "step": 27618
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.997801513872904e-07,
+      "loss": 0.9199,
+      "step": 27619
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.993993134007804e-07,
+      "loss": 1.1299,
+      "step": 27620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.990185753179668e-07,
+      "loss": 0.8447,
+      "step": 27621
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.986379371429408e-07,
+      "loss": 1.2207,
+      "step": 27622
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.982573988797902e-07,
+      "loss": 0.7334,
+      "step": 27623
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.978769605326007e-07,
+      "loss": 1.2686,
+      "step": 27624
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.974966221054602e-07,
+      "loss": 1.1143,
+      "step": 27625
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.971163836024575e-07,
+      "loss": 1.0654,
+      "step": 27626
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.967362450276727e-07,
+      "loss": 1.0098,
+      "step": 27627
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.963562063851892e-07,
+      "loss": 0.6392,
+      "step": 27628
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.959762676790916e-07,
+      "loss": 0.6016,
+      "step": 27629
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.955964289134598e-07,
+      "loss": 0.614,
+      "step": 27630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.952166900923751e-07,
+      "loss": 0.8242,
+      "step": 27631
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.948370512199143e-07,
+      "loss": 1.2256,
+      "step": 27632
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.944575123001574e-07,
+      "loss": 0.7383,
+      "step": 27633
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.940780733371799e-07,
+      "loss": 0.9541,
+      "step": 27634
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.936987343350576e-07,
+      "loss": 0.873,
+      "step": 27635
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.93319495297865e-07,
+      "loss": 0.7754,
+      "step": 27636
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.929403562296777e-07,
+      "loss": 0.8716,
+      "step": 27637
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.925613171345679e-07,
+      "loss": 0.6719,
+      "step": 27638
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.921823780166049e-07,
+      "loss": 1.0254,
+      "step": 27639
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.918035388798594e-07,
+      "loss": 0.8857,
+      "step": 27640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.91424799728404e-07,
+      "loss": 0.9121,
+      "step": 27641
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.910461605663044e-07,
+      "loss": 1.1807,
+      "step": 27642
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.906676213976271e-07,
+      "loss": 0.9805,
+      "step": 27643
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.902891822264402e-07,
+      "loss": 1.1055,
+      "step": 27644
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.899108430568069e-07,
+      "loss": 0.9004,
+      "step": 27645
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.89532603892794e-07,
+      "loss": 0.7881,
+      "step": 27646
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.891544647384607e-07,
+      "loss": 1.2451,
+      "step": 27647
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.887764255978713e-07,
+      "loss": 0.9707,
+      "step": 27648
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.88398486475087e-07,
+      "loss": 0.8076,
+      "step": 27649
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.880206473741646e-07,
+      "loss": 1.1221,
+      "step": 27650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.876429082991654e-07,
+      "loss": 0.7666,
+      "step": 27651
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.87265269254147e-07,
+      "loss": 1.0635,
+      "step": 27652
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.868877302431642e-07,
+      "loss": 0.9106,
+      "step": 27653
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.865102912702726e-07,
+      "loss": 0.7705,
+      "step": 27654
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.861329523395266e-07,
+      "loss": 0.6714,
+      "step": 27655
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.85755713454982e-07,
+      "loss": 0.8057,
+      "step": 27656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.853785746206887e-07,
+      "loss": 0.918,
+      "step": 27657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.850015358406958e-07,
+      "loss": 1.0996,
+      "step": 27658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.846245971190579e-07,
+      "loss": 1.0024,
+      "step": 27659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.842477584598206e-07,
+      "loss": 0.9175,
+      "step": 27660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.838710198670318e-07,
+      "loss": 0.9102,
+      "step": 27661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.834943813447392e-07,
+      "loss": 0.9805,
+      "step": 27662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.831178428969898e-07,
+      "loss": 0.8311,
+      "step": 27663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.827414045278269e-07,
+      "loss": 0.7012,
+      "step": 27664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.823650662412928e-07,
+      "loss": 0.7959,
+      "step": 27665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.81988828041431e-07,
+      "loss": 1.0098,
+      "step": 27666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.816126899322851e-07,
+      "loss": 1.0039,
+      "step": 27667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.812366519178926e-07,
+      "loss": 0.9082,
+      "step": 27668
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.808607140022927e-07,
+      "loss": 1.1348,
+      "step": 27669
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.804848761895267e-07,
+      "loss": 0.7725,
+      "step": 27670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.801091384836267e-07,
+      "loss": 0.877,
+      "step": 27671
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.797335008886341e-07,
+      "loss": 0.9634,
+      "step": 27672
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.793579634085801e-07,
+      "loss": 0.7695,
+      "step": 27673
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.789825260475014e-07,
+      "loss": 0.6123,
+      "step": 27674
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.786071888094292e-07,
+      "loss": 0.8418,
+      "step": 27675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.782319516983937e-07,
+      "loss": 0.958,
+      "step": 27676
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.778568147184261e-07,
+      "loss": 0.8096,
+      "step": 27677
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.774817778735598e-07,
+      "loss": 0.728,
+      "step": 27678
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.771068411678205e-07,
+      "loss": 0.9766,
+      "step": 27679
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.767320046052351e-07,
+      "loss": 0.7949,
+      "step": 27680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.76357268189829e-07,
+      "loss": 0.6655,
+      "step": 27681
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.759826319256313e-07,
+      "loss": 0.6089,
+      "step": 27682
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.756080958166645e-07,
+      "loss": 0.6035,
+      "step": 27683
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.752336598669484e-07,
+      "loss": 1.168,
+      "step": 27684
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.748593240805101e-07,
+      "loss": 0.6626,
+      "step": 27685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.744850884613674e-07,
+      "loss": 0.9092,
+      "step": 27686
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.741109530135415e-07,
+      "loss": 1.29,
+      "step": 27687
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.737369177410503e-07,
+      "loss": 0.6538,
+      "step": 27688
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.733629826479138e-07,
+      "loss": 1.084,
+      "step": 27689
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.729891477381467e-07,
+      "loss": 0.959,
+      "step": 27690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.726154130157625e-07,
+      "loss": 0.918,
+      "step": 27691
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.72241778484779e-07,
+      "loss": 1.1729,
+      "step": 27692
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.718682441492109e-07,
+      "loss": 1.2109,
+      "step": 27693
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.714948100130669e-07,
+      "loss": 0.9873,
+      "step": 27694
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.711214760803597e-07,
+      "loss": 1.1934,
+      "step": 27695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.707482423551004e-07,
+      "loss": 0.9961,
+      "step": 27696
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.703751088412958e-07,
+      "loss": 1.0771,
+      "step": 27697
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.700020755429559e-07,
+      "loss": 0.8115,
+      "step": 27698
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.696291424640889e-07,
+      "loss": 1.2188,
+      "step": 27699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.692563096087001e-07,
+      "loss": 0.9189,
+      "step": 27700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.688835769807923e-07,
+      "loss": 1.0635,
+      "step": 27701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.685109445843674e-07,
+      "loss": 0.6704,
+      "step": 27702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.681384124234336e-07,
+      "loss": 0.8623,
+      "step": 27703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.677659805019909e-07,
+      "loss": 0.6758,
+      "step": 27704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.673936488240374e-07,
+      "loss": 1.0479,
+      "step": 27705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.670214173935751e-07,
+      "loss": 0.9707,
+      "step": 27706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.666492862146001e-07,
+      "loss": 0.9951,
+      "step": 27707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.662772552911124e-07,
+      "loss": 0.8594,
+      "step": 27708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.659053246271041e-07,
+      "loss": 0.7402,
+      "step": 27709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.655334942265757e-07,
+      "loss": 0.8701,
+      "step": 27710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.651617640935182e-07,
+      "loss": 0.7412,
+      "step": 27711
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.64790134231924e-07,
+      "loss": 0.6401,
+      "step": 27712
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.644186046457856e-07,
+      "loss": 0.7852,
+      "step": 27713
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.640471753390954e-07,
+      "loss": 0.9971,
+      "step": 27714
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.636758463158421e-07,
+      "loss": 1.126,
+      "step": 27715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.633046175800129e-07,
+      "loss": 0.9873,
+      "step": 27716
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.629334891355965e-07,
+      "loss": 0.9473,
+      "step": 27717
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.625624609865811e-07,
+      "loss": 0.9395,
+      "step": 27718
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.62191533136951e-07,
+      "loss": 1.0742,
+      "step": 27719
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.618207055906889e-07,
+      "loss": 0.8926,
+      "step": 27720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.614499783517814e-07,
+      "loss": 0.9932,
+      "step": 27721
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.610793514242075e-07,
+      "loss": 0.6621,
+      "step": 27722
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.607088248119486e-07,
+      "loss": 0.876,
+      "step": 27723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.60338398518986e-07,
+      "loss": 0.7026,
+      "step": 27724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.599680725493007e-07,
+      "loss": 0.7588,
+      "step": 27725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.595978469068676e-07,
+      "loss": 1.04,
+      "step": 27726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.592277215956633e-07,
+      "loss": 0.9541,
+      "step": 27727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.588576966196636e-07,
+      "loss": 1.2939,
+      "step": 27728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.584877719828465e-07,
+      "loss": 0.9775,
+      "step": 27729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.58117947689183e-07,
+      "loss": 0.9751,
+      "step": 27730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.577482237426447e-07,
+      "loss": 1.2109,
+      "step": 27731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.57378600147206e-07,
+      "loss": 0.9717,
+      "step": 27732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.570090769068326e-07,
+      "loss": 1.0503,
+      "step": 27733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.566396540254993e-07,
+      "loss": 1.3037,
+      "step": 27734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.562703315071695e-07,
+      "loss": 0.7988,
+      "step": 27735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.559011093558154e-07,
+      "loss": 0.8281,
+      "step": 27736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.555319875753985e-07,
+      "loss": 0.9639,
+      "step": 27737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.551629661698844e-07,
+      "loss": 1.0732,
+      "step": 27738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.547940451432388e-07,
+      "loss": 0.9399,
+      "step": 27739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.544252244994242e-07,
+      "loss": 1.0254,
+      "step": 27740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.540565042424029e-07,
+      "loss": 0.8296,
+      "step": 27741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.536878843761318e-07,
+      "loss": 1.2217,
+      "step": 27742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.533193649045743e-07,
+      "loss": 0.9258,
+      "step": 27743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.529509458316885e-07,
+      "loss": 0.9443,
+      "step": 27744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.525826271614322e-07,
+      "loss": 0.8438,
+      "step": 27745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.522144088977578e-07,
+      "loss": 0.8262,
+      "step": 27746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.518462910446266e-07,
+      "loss": 0.9961,
+      "step": 27747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.514782736059866e-07,
+      "loss": 0.7437,
+      "step": 27748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.511103565857957e-07,
+      "loss": 1.1592,
+      "step": 27749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.507425399880041e-07,
+      "loss": 0.959,
+      "step": 27750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.50374823816563e-07,
+      "loss": 0.9727,
+      "step": 27751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.500072080754227e-07,
+      "loss": 0.8232,
+      "step": 27752
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.496396927685289e-07,
+      "loss": 0.9272,
+      "step": 27753
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.492722778998328e-07,
+      "loss": 1.252,
+      "step": 27754
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.489049634732814e-07,
+      "loss": 0.9111,
+      "step": 27755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.485377494928191e-07,
+      "loss": 1.2051,
+      "step": 27756
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.481706359623884e-07,
+      "loss": 1.2588,
+      "step": 27757
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.478036228859363e-07,
+      "loss": 0.6353,
+      "step": 27758
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.474367102674017e-07,
+      "loss": 0.5842,
+      "step": 27759
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.470698981107293e-07,
+      "loss": 0.9185,
+      "step": 27760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.467031864198548e-07,
+      "loss": 1.0,
+      "step": 27761
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.463365751987227e-07,
+      "loss": 0.8916,
+      "step": 27762
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.459700644512679e-07,
+      "loss": 1.1924,
+      "step": 27763
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.456036541814259e-07,
+      "loss": 1.0459,
+      "step": 27764
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.452373443931336e-07,
+      "loss": 1.2871,
+      "step": 27765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.448711350903291e-07,
+      "loss": 1.1328,
+      "step": 27766
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.445050262769436e-07,
+      "loss": 0.9302,
+      "step": 27767
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.441390179569074e-07,
+      "loss": 1.4072,
+      "step": 27768
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.437731101341538e-07,
+      "loss": 1.1934,
+      "step": 27769
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.434073028126164e-07,
+      "loss": 0.9893,
+      "step": 27770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.430415959962211e-07,
+      "loss": 1.0303,
+      "step": 27771
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.426759896888956e-07,
+      "loss": 1.1484,
+      "step": 27772
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.423104838945693e-07,
+      "loss": 1.0645,
+      "step": 27773
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.419450786171666e-07,
+      "loss": 1.0762,
+      "step": 27774
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.415797738606144e-07,
+      "loss": 0.8838,
+      "step": 27775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.412145696288353e-07,
+      "loss": 0.9463,
+      "step": 27776
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.408494659257525e-07,
+      "loss": 0.7891,
+      "step": 27777
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.404844627552886e-07,
+      "loss": 0.877,
+      "step": 27778
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.401195601213606e-07,
+      "loss": 0.646,
+      "step": 27779
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.397547580278951e-07,
+      "loss": 0.9385,
+      "step": 27780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.393900564788058e-07,
+      "loss": 0.8027,
+      "step": 27781
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.390254554780106e-07,
+      "loss": 1.0967,
+      "step": 27782
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.386609550294253e-07,
+      "loss": 1.0527,
+      "step": 27783
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.38296555136968e-07,
+      "loss": 1.0205,
+      "step": 27784
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.379322558045498e-07,
+      "loss": 1.0859,
+      "step": 27785
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.375680570360876e-07,
+      "loss": 1.1133,
+      "step": 27786
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.372039588354894e-07,
+      "loss": 0.9932,
+      "step": 27787
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.368399612066689e-07,
+      "loss": 1.0918,
+      "step": 27788
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.364760641535361e-07,
+      "loss": 0.9248,
+      "step": 27789
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.361122676799958e-07,
+      "loss": 1.1133,
+      "step": 27790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.357485717899614e-07,
+      "loss": 1.1543,
+      "step": 27791
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.353849764873377e-07,
+      "loss": 0.9727,
+      "step": 27792
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.350214817760292e-07,
+      "loss": 0.5479,
+      "step": 27793
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.346580876599395e-07,
+      "loss": 0.5181,
+      "step": 27794
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.342947941429745e-07,
+      "loss": 0.9307,
+      "step": 27795
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.339316012290364e-07,
+      "loss": 0.9453,
+      "step": 27796
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.335685089220234e-07,
+      "loss": 0.9121,
+      "step": 27797
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.332055172258411e-07,
+      "loss": 1.2393,
+      "step": 27798
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.328426261443843e-07,
+      "loss": 0.8906,
+      "step": 27799
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.324798356815509e-07,
+      "loss": 1.0088,
+      "step": 27800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.321171458412401e-07,
+      "loss": 0.7471,
+      "step": 27801
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.317545566273475e-07,
+      "loss": 1.0791,
+      "step": 27802
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.313920680437679e-07,
+      "loss": 1.1328,
+      "step": 27803
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.310296800943938e-07,
+      "loss": 0.9307,
+      "step": 27804
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.306673927831164e-07,
+      "loss": 0.9785,
+      "step": 27805
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.303052061138316e-07,
+      "loss": 0.8872,
+      "step": 27806
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.299431200904282e-07,
+      "loss": 0.707,
+      "step": 27807
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.295811347167924e-07,
+      "loss": 0.5205,
+      "step": 27808
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.292192499968175e-07,
+      "loss": 1.0791,
+      "step": 27809
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.288574659343871e-07,
+      "loss": 1.0127,
+      "step": 27810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.284957825333893e-07,
+      "loss": 0.9854,
+      "step": 27811
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.281341997977075e-07,
+      "loss": 0.7856,
+      "step": 27812
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.277727177312276e-07,
+      "loss": 0.9443,
+      "step": 27813
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.274113363378321e-07,
+      "loss": 0.9805,
+      "step": 27814
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.270500556214009e-07,
+      "loss": 0.8203,
+      "step": 27815
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.266888755858158e-07,
+      "loss": 0.6367,
+      "step": 27816
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.26327796234959e-07,
+      "loss": 0.7969,
+      "step": 27817
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.259668175727063e-07,
+      "loss": 0.8584,
+      "step": 27818
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.256059396029346e-07,
+      "loss": 1.1377,
+      "step": 27819
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.25245162329523e-07,
+      "loss": 0.9844,
+      "step": 27820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.248844857563441e-07,
+      "loss": 0.7017,
+      "step": 27821
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.245239098872747e-07,
+      "loss": 1.1895,
+      "step": 27822
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.241634347261849e-07,
+      "loss": 0.7354,
+      "step": 27823
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.238030602769507e-07,
+      "loss": 0.6265,
+      "step": 27824
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.234427865434401e-07,
+      "loss": 0.8713,
+      "step": 27825
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.230826135295231e-07,
+      "loss": 0.8145,
+      "step": 27826
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.227225412390691e-07,
+      "loss": 0.8887,
+      "step": 27827
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.223625696759483e-07,
+      "loss": 0.8916,
+      "step": 27828
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.220026988440242e-07,
+      "loss": 0.8813,
+      "step": 27829
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.216429287471626e-07,
+      "loss": 1.0361,
+      "step": 27830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.212832593892292e-07,
+      "loss": 1.1055,
+      "step": 27831
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.209236907740879e-07,
+      "loss": 0.9473,
+      "step": 27832
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.205642229056008e-07,
+      "loss": 0.6621,
+      "step": 27833
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.202048557876273e-07,
+      "loss": 1.1445,
+      "step": 27834
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.198455894240296e-07,
+      "loss": 0.8535,
+      "step": 27835
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.194864238186649e-07,
+      "loss": 0.7012,
+      "step": 27836
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.191273589753943e-07,
+      "loss": 1.0244,
+      "step": 27837
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.187683948980716e-07,
+      "loss": 1.123,
+      "step": 27838
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.184095315905559e-07,
+      "loss": 0.8213,
+      "step": 27839
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.180507690566984e-07,
+      "loss": 1.0322,
+      "step": 27840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.176921073003538e-07,
+      "loss": 1.2686,
+      "step": 27841
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.17333546325376e-07,
+      "loss": 0.7051,
+      "step": 27842
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.169750861356172e-07,
+      "loss": 0.9912,
+      "step": 27843
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.166167267349254e-07,
+      "loss": 0.8579,
+      "step": 27844
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.16258468127151e-07,
+      "loss": 1.3154,
+      "step": 27845
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.159003103161432e-07,
+      "loss": 0.6211,
+      "step": 27846
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.155422533057464e-07,
+      "loss": 0.9258,
+      "step": 27847
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.151842970998112e-07,
+      "loss": 1.0693,
+      "step": 27848
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.148264417021788e-07,
+      "loss": 0.6934,
+      "step": 27849
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.144686871166949e-07,
+      "loss": 1.1611,
+      "step": 27850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.141110333472033e-07,
+      "loss": 1.0874,
+      "step": 27851
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.13753480397542e-07,
+      "loss": 0.8989,
+      "step": 27852
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.133960282715545e-07,
+      "loss": 0.9375,
+      "step": 27853
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.130386769730812e-07,
+      "loss": 0.8857,
+      "step": 27854
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.126814265059599e-07,
+      "loss": 0.8516,
+      "step": 27855
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.123242768740267e-07,
+      "loss": 0.8271,
+      "step": 27856
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.119672280811184e-07,
+      "loss": 1.1055,
+      "step": 27857
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.116102801310719e-07,
+      "loss": 0.998,
+      "step": 27858
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.112534330277209e-07,
+      "loss": 0.8125,
+      "step": 27859
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.108966867748967e-07,
+      "loss": 1.2734,
+      "step": 27860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.105400413764339e-07,
+      "loss": 1.0889,
+      "step": 27861
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.101834968361608e-07,
+      "loss": 0.9268,
+      "step": 27862
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.098270531579098e-07,
+      "loss": 0.8647,
+      "step": 27863
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.094707103455077e-07,
+      "loss": 0.707,
+      "step": 27864
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.091144684027838e-07,
+      "loss": 1.1797,
+      "step": 27865
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.08758327333564e-07,
+      "loss": 0.9541,
+      "step": 27866
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.084022871416728e-07,
+      "loss": 1.2627,
+      "step": 27867
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.080463478309351e-07,
+      "loss": 0.9873,
+      "step": 27868
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.076905094051766e-07,
+      "loss": 1.4766,
+      "step": 27869
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.073347718682177e-07,
+      "loss": 0.8066,
+      "step": 27870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.069791352238785e-07,
+      "loss": 1.1631,
+      "step": 27871
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.066235994759806e-07,
+      "loss": 0.7422,
+      "step": 27872
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.062681646283441e-07,
+      "loss": 1.1855,
+      "step": 27873
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.05912830684785e-07,
+      "loss": 0.7441,
+      "step": 27874
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.055575976491201e-07,
+      "loss": 0.7798,
+      "step": 27875
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.052024655251676e-07,
+      "loss": 1.002,
+      "step": 27876
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.048474343167398e-07,
+      "loss": 0.7817,
+      "step": 27877
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.044925040276483e-07,
+      "loss": 0.8296,
+      "step": 27878
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.041376746617112e-07,
+      "loss": 1.0283,
+      "step": 27879
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.037829462227374e-07,
+      "loss": 1.0957,
+      "step": 27880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.034283187145362e-07,
+      "loss": 0.9219,
+      "step": 27881
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.030737921409169e-07,
+      "loss": 0.9419,
+      "step": 27882
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.027193665056885e-07,
+      "loss": 1.0649,
+      "step": 27883
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.023650418126592e-07,
+      "loss": 0.9482,
+      "step": 27884
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.020108180656337e-07,
+      "loss": 1.1533,
+      "step": 27885
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.016566952684156e-07,
+      "loss": 0.833,
+      "step": 27886
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.013026734248117e-07,
+      "loss": 0.7705,
+      "step": 27887
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.009487525386226e-07,
+      "loss": 1.04,
+      "step": 27888
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.005949326136507e-07,
+      "loss": 1.1279,
+      "step": 27889
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.002412136536972e-07,
+      "loss": 1.0264,
+      "step": 27890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.998875956625616e-07,
+      "loss": 0.8389,
+      "step": 27891
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.995340786440418e-07,
+      "loss": 0.7246,
+      "step": 27892
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.991806626019325e-07,
+      "loss": 0.7017,
+      "step": 27893
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.988273475400353e-07,
+      "loss": 1.0752,
+      "step": 27894
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.984741334621425e-07,
+      "loss": 0.6196,
+      "step": 27895
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.981210203720467e-07,
+      "loss": 0.9395,
+      "step": 27896
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.977680082735449e-07,
+      "loss": 1.0088,
+      "step": 27897
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.974150971704252e-07,
+      "loss": 0.7812,
+      "step": 27898
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.97062287066481e-07,
+      "loss": 0.9097,
+      "step": 27899
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.967095779654997e-07,
+      "loss": 1.2617,
+      "step": 27900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.963569698712724e-07,
+      "loss": 1.0264,
+      "step": 27901
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.96004462787586e-07,
+      "loss": 0.8242,
+      "step": 27902
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.956520567182256e-07,
+      "loss": 1.123,
+      "step": 27903
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.952997516669779e-07,
+      "loss": 0.79,
+      "step": 27904
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.949475476376277e-07,
+      "loss": 1.1523,
+      "step": 27905
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.945954446339574e-07,
+      "loss": 1.0742,
+      "step": 27906
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.942434426597476e-07,
+      "loss": 0.9197,
+      "step": 27907
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.938915417187829e-07,
+      "loss": 0.9414,
+      "step": 27908
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.935397418148403e-07,
+      "loss": 0.7251,
+      "step": 27909
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.931880429517012e-07,
+      "loss": 0.6211,
+      "step": 27910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.928364451331403e-07,
+      "loss": 0.791,
+      "step": 27911
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.92484948362938e-07,
+      "loss": 0.6265,
+      "step": 27912
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.92133552644868e-07,
+      "loss": 0.6565,
+      "step": 27913
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.917822579827037e-07,
+      "loss": 1.3018,
+      "step": 27914
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.91431064380219e-07,
+      "loss": 0.8828,
+      "step": 27915
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.910799718411897e-07,
+      "loss": 1.0732,
+      "step": 27916
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.907289803693839e-07,
+      "loss": 1.2412,
+      "step": 27917
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.903780899685707e-07,
+      "loss": 0.8677,
+      "step": 27918
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.900273006425205e-07,
+      "loss": 0.8887,
+      "step": 27919
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.896766123950048e-07,
+      "loss": 1.1104,
+      "step": 27920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.893260252297872e-07,
+      "loss": 0.7266,
+      "step": 27921
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.889755391506314e-07,
+      "loss": 0.8691,
+      "step": 27922
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.886251541613075e-07,
+      "loss": 0.9736,
+      "step": 27923
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.88274870265575e-07,
+      "loss": 0.938,
+      "step": 27924
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.879246874671985e-07,
+      "loss": 0.7017,
+      "step": 27925
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.875746057699383e-07,
+      "loss": 0.7793,
+      "step": 27926
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.872246251775571e-07,
+      "loss": 1.2236,
+      "step": 27927
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.868747456938118e-07,
+      "loss": 1.0537,
+      "step": 27928
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.865249673224616e-07,
+      "loss": 0.8169,
+      "step": 27929
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.861752900672624e-07,
+      "loss": 0.832,
+      "step": 27930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.858257139319745e-07,
+      "loss": 0.8223,
+      "step": 27931
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.854762389203494e-07,
+      "loss": 0.709,
+      "step": 27932
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.851268650361397e-07,
+      "loss": 0.7979,
+      "step": 27933
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.847775922831012e-07,
+      "loss": 1.0693,
+      "step": 27934
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.844284206649853e-07,
+      "loss": 1.2246,
+      "step": 27935
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.840793501855424e-07,
+      "loss": 1.6094,
+      "step": 27936
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.837303808485207e-07,
+      "loss": 0.7251,
+      "step": 27937
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.833815126576714e-07,
+      "loss": 0.998,
+      "step": 27938
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.830327456167395e-07,
+      "loss": 0.9102,
+      "step": 27939
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.826840797294708e-07,
+      "loss": 0.8252,
+      "step": 27940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.823355149996124e-07,
+      "loss": 1.2754,
+      "step": 27941
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.81987051430909e-07,
+      "loss": 0.968,
+      "step": 27942
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.816386890271031e-07,
+      "loss": 0.6865,
+      "step": 27943
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.81290427791934e-07,
+      "loss": 1.1123,
+      "step": 27944
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.809422677291454e-07,
+      "loss": 0.8145,
+      "step": 27945
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.805942088424777e-07,
+      "loss": 0.9453,
+      "step": 27946
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.802462511356677e-07,
+      "loss": 0.8555,
+      "step": 27947
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.798983946124537e-07,
+      "loss": 1.1396,
+      "step": 27948
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.795506392765726e-07,
+      "loss": 0.7588,
+      "step": 27949
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.792029851317582e-07,
+      "loss": 0.9199,
+      "step": 27950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.788554321817486e-07,
+      "loss": 1.0293,
+      "step": 27951
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.785079804302718e-07,
+      "loss": 0.8154,
+      "step": 27952
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.78160629881066e-07,
+      "loss": 0.7314,
+      "step": 27953
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.778133805378583e-07,
+      "loss": 1.3047,
+      "step": 27954
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.774662324043779e-07,
+      "loss": 0.9434,
+      "step": 27955
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.771191854843561e-07,
+      "loss": 1.042,
+      "step": 27956
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.767722397815212e-07,
+      "loss": 1.5332,
+      "step": 27957
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.76425395299599e-07,
+      "loss": 0.9551,
+      "step": 27958
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.760786520423145e-07,
+      "loss": 0.8789,
+      "step": 27959
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.757320100133922e-07,
+      "loss": 1.3672,
+      "step": 27960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.753854692165573e-07,
+      "loss": 0.8994,
+      "step": 27961
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.750390296555319e-07,
+      "loss": 1.499,
+      "step": 27962
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.746926913340345e-07,
+      "loss": 1.0122,
+      "step": 27963
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.743464542557898e-07,
+      "loss": 0.7461,
+      "step": 27964
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.740003184245124e-07,
+      "loss": 1.3896,
+      "step": 27965
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.736542838439252e-07,
+      "loss": 1.168,
+      "step": 27966
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.733083505177406e-07,
+      "loss": 0.6436,
+      "step": 27967
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.729625184496768e-07,
+      "loss": 0.7446,
+      "step": 27968
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.726167876434497e-07,
+      "loss": 1.0068,
+      "step": 27969
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.722711581027695e-07,
+      "loss": 0.7158,
+      "step": 27970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.719256298313514e-07,
+      "loss": 1.2344,
+      "step": 27971
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.715802028329065e-07,
+      "loss": 0.7998,
+      "step": 27972
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.712348771111464e-07,
+      "loss": 0.8701,
+      "step": 27973
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.708896526697782e-07,
+      "loss": 1.0029,
+      "step": 27974
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.705445295125123e-07,
+      "loss": 0.4844,
+      "step": 27975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.701995076430533e-07,
+      "loss": 1.0576,
+      "step": 27976
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.698545870651084e-07,
+      "loss": 1.3379,
+      "step": 27977
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.695097677823846e-07,
+      "loss": 0.625,
+      "step": 27978
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.691650497985845e-07,
+      "loss": 1.3555,
+      "step": 27979
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.688204331174108e-07,
+      "loss": 0.7969,
+      "step": 27980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.684759177425625e-07,
+      "loss": 0.8818,
+      "step": 27981
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.681315036777457e-07,
+      "loss": 0.8428,
+      "step": 27982
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.677871909266564e-07,
+      "loss": 0.7891,
+      "step": 27983
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.674429794929948e-07,
+      "loss": 0.7354,
+      "step": 27984
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.670988693804558e-07,
+      "loss": 1.1104,
+      "step": 27985
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.667548605927365e-07,
+      "loss": 1.0693,
+      "step": 27986
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.664109531335349e-07,
+      "loss": 0.8271,
+      "step": 27987
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.660671470065415e-07,
+      "loss": 0.707,
+      "step": 27988
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.657234422154534e-07,
+      "loss": 0.8906,
+      "step": 27989
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.653798387639586e-07,
+      "loss": 1.0469,
+      "step": 27990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.650363366557487e-07,
+      "loss": 0.9395,
+      "step": 27991
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.646929358945152e-07,
+      "loss": 1.1572,
+      "step": 27992
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.643496364839462e-07,
+      "loss": 1.123,
+      "step": 27993
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.640064384277288e-07,
+      "loss": 0.6582,
+      "step": 27994
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.63663341729549e-07,
+      "loss": 0.8633,
+      "step": 27995
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.633203463930937e-07,
+      "loss": 0.6763,
+      "step": 27996
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.629774524220455e-07,
+      "loss": 0.8647,
+      "step": 27997
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.626346598200893e-07,
+      "loss": 0.9341,
+      "step": 27998
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.622919685909056e-07,
+      "loss": 0.9385,
+      "step": 27999
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.619493787381769e-07,
+      "loss": 1.1201,
+      "step": 28000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.616068902655835e-07,
+      "loss": 1.167,
+      "step": 28001
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.612645031768027e-07,
+      "loss": 0.8779,
+      "step": 28002
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.609222174755113e-07,
+      "loss": 1.0654,
+      "step": 28003
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.605800331653899e-07,
+      "loss": 0.918,
+      "step": 28004
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.602379502501132e-07,
+      "loss": 0.8896,
+      "step": 28005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.598959687333516e-07,
+      "loss": 1.2842,
+      "step": 28006
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.595540886187811e-07,
+      "loss": 1.3652,
+      "step": 28007
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.592123099100766e-07,
+      "loss": 1.0127,
+      "step": 28008
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.588706326109072e-07,
+      "loss": 0.7656,
+      "step": 28009
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.585290567249424e-07,
+      "loss": 0.8066,
+      "step": 28010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.581875822558513e-07,
+      "loss": 1.0303,
+      "step": 28011
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.578462092073023e-07,
+      "loss": 0.9678,
+      "step": 28012
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.575049375829645e-07,
+      "loss": 0.9668,
+      "step": 28013
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.571637673864993e-07,
+      "loss": 0.8071,
+      "step": 28014
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.568226986215764e-07,
+      "loss": 0.8057,
+      "step": 28015
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.564817312918558e-07,
+      "loss": 1.0371,
+      "step": 28016
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.561408654009992e-07,
+      "loss": 1.0791,
+      "step": 28017
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.558001009526715e-07,
+      "loss": 1.29,
+      "step": 28018
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.554594379505318e-07,
+      "loss": 1.0811,
+      "step": 28019
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.551188763982396e-07,
+      "loss": 0.9961,
+      "step": 28020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.547784162994518e-07,
+      "loss": 0.8433,
+      "step": 28021
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.544380576578256e-07,
+      "loss": 0.8306,
+      "step": 28022
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.540978004770192e-07,
+      "loss": 0.9209,
+      "step": 28023
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.537576447606863e-07,
+      "loss": 1.1592,
+      "step": 28024
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.534175905124794e-07,
+      "loss": 0.78,
+      "step": 28025
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.530776377360536e-07,
+      "loss": 0.8105,
+      "step": 28026
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.527377864350592e-07,
+      "loss": 0.9199,
+      "step": 28027
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.523980366131454e-07,
+      "loss": 0.6401,
+      "step": 28028
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.520583882739628e-07,
+      "loss": 0.9673,
+      "step": 28029
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.517188414211616e-07,
+      "loss": 1.0684,
+      "step": 28030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.513793960583879e-07,
+      "loss": 1.0527,
+      "step": 28031
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.510400521892856e-07,
+      "loss": 1.2344,
+      "step": 28032
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.507008098175027e-07,
+      "loss": 1.0498,
+      "step": 28033
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.503616689466829e-07,
+      "loss": 1.1514,
+      "step": 28034
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.500226295804689e-07,
+      "loss": 1.2725,
+      "step": 28035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.496836917225012e-07,
+      "loss": 1.4365,
+      "step": 28036
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.493448553764225e-07,
+      "loss": 0.8389,
+      "step": 28037
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.490061205458697e-07,
+      "loss": 1.1592,
+      "step": 28038
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.486674872344844e-07,
+      "loss": 1.1738,
+      "step": 28039
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.483289554459026e-07,
+      "loss": 0.6938,
+      "step": 28040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.479905251837614e-07,
+      "loss": 0.7295,
+      "step": 28041
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.476521964516957e-07,
+      "loss": 1.3242,
+      "step": 28042
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.473139692533391e-07,
+      "loss": 1.0684,
+      "step": 28043
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.469758435923245e-07,
+      "loss": 1.0684,
+      "step": 28044
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.466378194722854e-07,
+      "loss": 0.7324,
+      "step": 28045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.462998968968525e-07,
+      "loss": 0.8799,
+      "step": 28046
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.459620758696549e-07,
+      "loss": 1.1025,
+      "step": 28047
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.456243563943197e-07,
+      "loss": 0.7197,
+      "step": 28048
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.452867384744787e-07,
+      "loss": 0.8086,
+      "step": 28049
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.449492221137564e-07,
+      "loss": 1.0078,
+      "step": 28050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.44611807315777e-07,
+      "loss": 0.8955,
+      "step": 28051
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.442744940841671e-07,
+      "loss": 0.7441,
+      "step": 28052
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.439372824225475e-07,
+      "loss": 0.959,
+      "step": 28053
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.43600172334543e-07,
+      "loss": 0.8711,
+      "step": 28054
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.432631638237718e-07,
+      "loss": 1.042,
+      "step": 28055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.429262568938575e-07,
+      "loss": 0.8115,
+      "step": 28056
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.425894515484164e-07,
+      "loss": 1.2061,
+      "step": 28057
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.422527477910666e-07,
+      "loss": 0.6353,
+      "step": 28058
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.419161456254252e-07,
+      "loss": 0.856,
+      "step": 28059
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.415796450551092e-07,
+      "loss": 1.1289,
+      "step": 28060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.412432460837314e-07,
+      "loss": 0.6829,
+      "step": 28061
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.409069487149054e-07,
+      "loss": 0.9199,
+      "step": 28062
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.405707529522453e-07,
+      "loss": 1.1152,
+      "step": 28063
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.402346587993589e-07,
+      "loss": 0.583,
+      "step": 28064
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.398986662598615e-07,
+      "loss": 0.959,
+      "step": 28065
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.395627753373567e-07,
+      "loss": 0.5942,
+      "step": 28066
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.392269860354571e-07,
+      "loss": 1.0986,
+      "step": 28067
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.388912983577677e-07,
+      "loss": 1.0469,
+      "step": 28068
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.385557123078911e-07,
+      "loss": 0.9062,
+      "step": 28069
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.382202278894388e-07,
+      "loss": 1.1152,
+      "step": 28070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.378848451060104e-07,
+      "loss": 1.0537,
+      "step": 28071
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.375495639612094e-07,
+      "loss": 0.9551,
+      "step": 28072
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.372143844586341e-07,
+      "loss": 0.9741,
+      "step": 28073
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.368793066018895e-07,
+      "loss": 0.9385,
+      "step": 28074
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.365443303945728e-07,
+      "loss": 0.9424,
+      "step": 28075
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.362094558402819e-07,
+      "loss": 0.9824,
+      "step": 28076
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.358746829426165e-07,
+      "loss": 1.0332,
+      "step": 28077
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.35540011705169e-07,
+      "loss": 1.2314,
+      "step": 28078
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.352054421315356e-07,
+      "loss": 0.9551,
+      "step": 28079
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.348709742253099e-07,
+      "loss": 1.0439,
+      "step": 28080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.34536607990086e-07,
+      "loss": 1.0264,
+      "step": 28081
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.342023434294552e-07,
+      "loss": 1.2637,
+      "step": 28082
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.33868180547008e-07,
+      "loss": 1.1006,
+      "step": 28083
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.335341193463295e-07,
+      "loss": 1.1572,
+      "step": 28084
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.332001598310154e-07,
+      "loss": 0.8906,
+      "step": 28085
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.328663020046487e-07,
+      "loss": 0.9395,
+      "step": 28086
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.325325458708163e-07,
+      "loss": 0.8926,
+      "step": 28087
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.321988914331034e-07,
+      "loss": 0.8594,
+      "step": 28088
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.318653386950945e-07,
+      "loss": 0.5255,
+      "step": 28089
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.315318876603714e-07,
+      "loss": 1.3359,
+      "step": 28090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.311985383325146e-07,
+      "loss": 0.9941,
+      "step": 28091
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.308652907151102e-07,
+      "loss": 0.7637,
+      "step": 28092
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.305321448117329e-07,
+      "loss": 1.0645,
+      "step": 28093
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.30199100625961e-07,
+      "loss": 0.8154,
+      "step": 28094
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.29866158161374e-07,
+      "loss": 1.0615,
+      "step": 28095
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2953331742155e-07,
+      "loss": 1.1592,
+      "step": 28096
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.292005784100618e-07,
+      "loss": 1.1553,
+      "step": 28097
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.28867941130482e-07,
+      "loss": 1.0947,
+      "step": 28098
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.285354055863867e-07,
+      "loss": 0.8613,
+      "step": 28099
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.282029717813453e-07,
+      "loss": 0.9932,
+      "step": 28100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.278706397189315e-07,
+      "loss": 1.1753,
+      "step": 28101
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.275384094027125e-07,
+      "loss": 1.1016,
+      "step": 28102
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.272062808362599e-07,
+      "loss": 0.8384,
+      "step": 28103
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.268742540231398e-07,
+      "loss": 0.9316,
+      "step": 28104
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.265423289669158e-07,
+      "loss": 0.7549,
+      "step": 28105
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.262105056711575e-07,
+      "loss": 0.875,
+      "step": 28106
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.258787841394275e-07,
+      "loss": 0.876,
+      "step": 28107
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.255471643752907e-07,
+      "loss": 1.1592,
+      "step": 28108
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.252156463823055e-07,
+      "loss": 1.1387,
+      "step": 28109
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.248842301640356e-07,
+      "loss": 0.8369,
+      "step": 28110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.245529157240425e-07,
+      "loss": 1.0596,
+      "step": 28111
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.242217030658825e-07,
+      "loss": 1.1504,
+      "step": 28112
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.238905921931125e-07,
+      "loss": 1.0498,
+      "step": 28113
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.235595831092921e-07,
+      "loss": 0.8687,
+      "step": 28114
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.23228675817975e-07,
+      "loss": 1.3516,
+      "step": 28115
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.228978703227161e-07,
+      "loss": 0.9658,
+      "step": 28116
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.225671666270682e-07,
+      "loss": 0.7402,
+      "step": 28117
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.222365647345862e-07,
+      "loss": 1.085,
+      "step": 28118
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.219060646488184e-07,
+      "loss": 0.8457,
+      "step": 28119
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.215756663733151e-07,
+      "loss": 1.0508,
+      "step": 28120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.21245369911626e-07,
+      "loss": 1.5977,
+      "step": 28121
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.209151752673003e-07,
+      "loss": 0.8711,
+      "step": 28122
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.20585082443884e-07,
+      "loss": 0.752,
+      "step": 28123
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.202550914449212e-07,
+      "loss": 0.8145,
+      "step": 28124
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.199252022739587e-07,
+      "loss": 1.0225,
+      "step": 28125
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.195954149345384e-07,
+      "loss": 0.958,
+      "step": 28126
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.192657294302039e-07,
+      "loss": 0.7656,
+      "step": 28127
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.189361457644959e-07,
+      "loss": 0.835,
+      "step": 28128
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.186066639409559e-07,
+      "loss": 1.1094,
+      "step": 28129
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.182772839631223e-07,
+      "loss": 0.5474,
+      "step": 28130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.179480058345321e-07,
+      "loss": 1.043,
+      "step": 28131
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.176188295587226e-07,
+      "loss": 0.8525,
+      "step": 28132
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.172897551392308e-07,
+      "loss": 0.835,
+      "step": 28133
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.16960782579592e-07,
+      "loss": 0.876,
+      "step": 28134
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.166319118833385e-07,
+      "loss": 0.9189,
+      "step": 28135
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.163031430540022e-07,
+      "loss": 0.6816,
+      "step": 28136
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.159744760951169e-07,
+      "loss": 1.0371,
+      "step": 28137
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.156459110102119e-07,
+      "loss": 0.8896,
+      "step": 28138
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.153174478028155e-07,
+      "loss": 0.9795,
+      "step": 28139
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.149890864764584e-07,
+      "loss": 1.7031,
+      "step": 28140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.146608270346642e-07,
+      "loss": 0.8438,
+      "step": 28141
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.143326694809636e-07,
+      "loss": 0.9722,
+      "step": 28142
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.14004613818877e-07,
+      "loss": 0.7207,
+      "step": 28143
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.136766600519316e-07,
+      "loss": 1.0947,
+      "step": 28144
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.13348808183648e-07,
+      "loss": 0.957,
+      "step": 28145
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.130210582175477e-07,
+      "loss": 1.1436,
+      "step": 28146
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.126934101571524e-07,
+      "loss": 1.3174,
+      "step": 28147
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.123658640059837e-07,
+      "loss": 0.8511,
+      "step": 28148
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.120384197675565e-07,
+      "loss": 1.2227,
+      "step": 28149
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.117110774453882e-07,
+      "loss": 1.2705,
+      "step": 28150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.113838370429969e-07,
+      "loss": 0.793,
+      "step": 28151
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.110566985638965e-07,
+      "loss": 0.814,
+      "step": 28152
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.10729662011602e-07,
+      "loss": 0.8213,
+      "step": 28153
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.104027273896239e-07,
+      "loss": 0.791,
+      "step": 28154
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.100758947014783e-07,
+      "loss": 0.7598,
+      "step": 28155
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.097491639506724e-07,
+      "loss": 0.8018,
+      "step": 28156
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.094225351407156e-07,
+      "loss": 0.8589,
+      "step": 28157
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.090960082751184e-07,
+      "loss": 1.0596,
+      "step": 28158
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.08769583357388e-07,
+      "loss": 1.0059,
+      "step": 28159
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.084432603910305e-07,
+      "loss": 0.6914,
+      "step": 28160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.081170393795498e-07,
+      "loss": 0.9502,
+      "step": 28161
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.077909203264509e-07,
+      "loss": 1.0049,
+      "step": 28162
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.074649032352386e-07,
+      "loss": 1.0127,
+      "step": 28163
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.071389881094125e-07,
+      "loss": 0.9014,
+      "step": 28164
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.06813174952474e-07,
+      "loss": 1.207,
+      "step": 28165
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.06487463767924e-07,
+      "loss": 1.1113,
+      "step": 28166
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.061618545592584e-07,
+      "loss": 0.8804,
+      "step": 28167
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.058363473299777e-07,
+      "loss": 1.1152,
+      "step": 28168
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.05510942083578e-07,
+      "loss": 0.8584,
+      "step": 28169
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.051856388235543e-07,
+      "loss": 1.0371,
+      "step": 28170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.048604375534016e-07,
+      "loss": 1.293,
+      "step": 28171
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.045353382766083e-07,
+      "loss": 0.6123,
+      "step": 28172
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.042103409966736e-07,
+      "loss": 1.1553,
+      "step": 28173
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.03885445717085e-07,
+      "loss": 1.0332,
+      "step": 28174
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.035606524413328e-07,
+      "loss": 1.0371,
+      "step": 28175
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.032359611729043e-07,
+      "loss": 1.0645,
+      "step": 28176
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.029113719152879e-07,
+      "loss": 1.0308,
+      "step": 28177
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.025868846719728e-07,
+      "loss": 1.126,
+      "step": 28178
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.02262499446442e-07,
+      "loss": 1.5771,
+      "step": 28179
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.019382162421815e-07,
+      "loss": 1.1846,
+      "step": 28180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.01614035062673e-07,
+      "loss": 1.0215,
+      "step": 28181
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.012899559114004e-07,
+      "loss": 1.2617,
+      "step": 28182
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.009659787918408e-07,
+      "loss": 1.2031,
+      "step": 28183
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.006421037074794e-07,
+      "loss": 1.0718,
+      "step": 28184
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.003183306617943e-07,
+      "loss": 0.7227,
+      "step": 28185
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.999946596582595e-07,
+      "loss": 0.6582,
+      "step": 28186
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.996710907003566e-07,
+      "loss": 0.9814,
+      "step": 28187
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.993476237915574e-07,
+      "loss": 1.2627,
+      "step": 28188
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.990242589353389e-07,
+      "loss": 1.0957,
+      "step": 28189
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.98700996135173e-07,
+      "loss": 0.8623,
+      "step": 28190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.983778353945345e-07,
+      "loss": 0.7432,
+      "step": 28191
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.980547767168931e-07,
+      "loss": 0.6729,
+      "step": 28192
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.977318201057168e-07,
+      "loss": 0.9004,
+      "step": 28193
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.974089655644776e-07,
+      "loss": 0.6016,
+      "step": 28194
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.970862130966436e-07,
+      "loss": 0.9883,
+      "step": 28195
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.967635627056811e-07,
+      "loss": 1.1753,
+      "step": 28196
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.964410143950538e-07,
+      "loss": 1.4268,
+      "step": 28197
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.961185681682279e-07,
+      "loss": 0.9053,
+      "step": 28198
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.957962240286684e-07,
+      "loss": 0.9424,
+      "step": 28199
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.954739819798382e-07,
+      "loss": 0.8535,
+      "step": 28200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.951518420251944e-07,
+      "loss": 0.7451,
+      "step": 28201
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.94829804168202e-07,
+      "loss": 1.0391,
+      "step": 28202
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.945078684123173e-07,
+      "loss": 1.0674,
+      "step": 28203
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.941860347609995e-07,
+      "loss": 1.1357,
+      "step": 28204
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.938643032177049e-07,
+      "loss": 1.0342,
+      "step": 28205
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.935426737858917e-07,
+      "loss": 1.0869,
+      "step": 28206
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.932211464690129e-07,
+      "loss": 0.8525,
+      "step": 28207
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.928997212705211e-07,
+      "loss": 1.4512,
+      "step": 28208
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.925783981938703e-07,
+      "loss": 0.7822,
+      "step": 28209
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.922571772425133e-07,
+      "loss": 0.842,
+      "step": 28210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.919360584198996e-07,
+      "loss": 0.8267,
+      "step": 28211
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.916150417294762e-07,
+      "loss": 1.0664,
+      "step": 28212
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.912941271746952e-07,
+      "loss": 0.7427,
+      "step": 28213
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.909733147590001e-07,
+      "loss": 0.9395,
+      "step": 28214
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.906526044858395e-07,
+      "loss": 0.8086,
+      "step": 28215
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.903319963586573e-07,
+      "loss": 0.7014,
+      "step": 28216
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.900114903808984e-07,
+      "loss": 0.7539,
+      "step": 28217
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.896910865560045e-07,
+      "loss": 0.8125,
+      "step": 28218
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.893707848874174e-07,
+      "loss": 0.6484,
+      "step": 28219
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.890505853785765e-07,
+      "loss": 0.7915,
+      "step": 28220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.887304880329246e-07,
+      "loss": 0.9619,
+      "step": 28221
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.884104928538979e-07,
+      "loss": 0.6167,
+      "step": 28222
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.880905998449325e-07,
+      "loss": 0.9102,
+      "step": 28223
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.877708090094657e-07,
+      "loss": 0.7678,
+      "step": 28224
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.874511203509347e-07,
+      "loss": 0.7046,
+      "step": 28225
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.871315338727711e-07,
+      "loss": 1.1104,
+      "step": 28226
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.868120495784067e-07,
+      "loss": 1.0234,
+      "step": 28227
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.864926674712766e-07,
+      "loss": 1.2236,
+      "step": 28228
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.86173387554808e-07,
+      "loss": 1.2402,
+      "step": 28229
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.858542098324349e-07,
+      "loss": 0.9766,
+      "step": 28230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.855351343075798e-07,
+      "loss": 0.8979,
+      "step": 28231
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.852161609836759e-07,
+      "loss": 1.1074,
+      "step": 28232
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.848972898641468e-07,
+      "loss": 1.2322,
+      "step": 28233
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.845785209524156e-07,
+      "loss": 0.9629,
+      "step": 28234
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.842598542519094e-07,
+      "loss": 1.0293,
+      "step": 28235
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.839412897660523e-07,
+      "loss": 1.2109,
+      "step": 28236
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.836228274982646e-07,
+      "loss": 1.0166,
+      "step": 28237
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.833044674519649e-07,
+      "loss": 0.8877,
+      "step": 28238
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.829862096305749e-07,
+      "loss": 1.2578,
+      "step": 28239
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.826680540375139e-07,
+      "loss": 0.9121,
+      "step": 28240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.823500006761995e-07,
+      "loss": 0.9199,
+      "step": 28241
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.820320495500464e-07,
+      "loss": 0.8447,
+      "step": 28242
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.81714200662472e-07,
+      "loss": 0.9326,
+      "step": 28243
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.813964540168892e-07,
+      "loss": 0.6816,
+      "step": 28244
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.810788096167096e-07,
+      "loss": 0.9453,
+      "step": 28245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.807612674653472e-07,
+      "loss": 0.9746,
+      "step": 28246
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.804438275662137e-07,
+      "loss": 0.9526,
+      "step": 28247
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.801264899227188e-07,
+      "loss": 1.3418,
+      "step": 28248
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.798092545382682e-07,
+      "loss": 1.1567,
+      "step": 28249
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.794921214162717e-07,
+      "loss": 0.8359,
+      "step": 28250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.791750905601367e-07,
+      "loss": 0.8452,
+      "step": 28251
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.78858161973268e-07,
+      "loss": 1.1162,
+      "step": 28252
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.785413356590685e-07,
+      "loss": 1.1211,
+      "step": 28253
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.782246116209443e-07,
+      "loss": 1.2168,
+      "step": 28254
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.77907989862294e-07,
+      "loss": 0.9482,
+      "step": 28255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.775914703865214e-07,
+      "loss": 0.7925,
+      "step": 28256
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.772750531970238e-07,
+      "loss": 1.084,
+      "step": 28257
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.76958738297204e-07,
+      "loss": 1.1045,
+      "step": 28258
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.766425256904572e-07,
+      "loss": 1.0898,
+      "step": 28259
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7632641538017724e-07,
+      "loss": 0.8608,
+      "step": 28260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.760104073697658e-07,
+      "loss": 0.9209,
+      "step": 28261
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7569450166261356e-07,
+      "loss": 0.8877,
+      "step": 28262
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7537869826211556e-07,
+      "loss": 1.0781,
+      "step": 28263
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.750629971716614e-07,
+      "loss": 1.3564,
+      "step": 28264
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.747473983946438e-07,
+      "loss": 1.0049,
+      "step": 28265
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.744319019344534e-07,
+      "loss": 0.8398,
+      "step": 28266
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.741165077944776e-07,
+      "loss": 0.9775,
+      "step": 28267
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7380121597810694e-07,
+      "loss": 0.958,
+      "step": 28268
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7348602648872644e-07,
+      "loss": 1.3691,
+      "step": 28269
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7317093932972125e-07,
+      "loss": 0.8657,
+      "step": 28270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.728559545044753e-07,
+      "loss": 1.3291,
+      "step": 28271
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.725410720163748e-07,
+      "loss": 0.895,
+      "step": 28272
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.722262918688003e-07,
+      "loss": 1.0244,
+      "step": 28273
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7191161406513363e-07,
+      "loss": 0.8862,
+      "step": 28274
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7159703860875317e-07,
+      "loss": 0.918,
+      "step": 28275
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7128256550303954e-07,
+      "loss": 0.9619,
+      "step": 28276
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.709681947513722e-07,
+      "loss": 0.7188,
+      "step": 28277
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.706539263571241e-07,
+      "loss": 1.033,
+      "step": 28278
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.703397603236748e-07,
+      "loss": 0.6831,
+      "step": 28279
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.700256966543981e-07,
+      "loss": 0.9502,
+      "step": 28280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.697117353526648e-07,
+      "loss": 0.9209,
+      "step": 28281
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6939787642184986e-07,
+      "loss": 1.1177,
+      "step": 28282
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.690841198653262e-07,
+      "loss": 1.0771,
+      "step": 28283
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.687704656864611e-07,
+      "loss": 0.7607,
+      "step": 28284
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6845691388862303e-07,
+      "loss": 1.2344,
+      "step": 28285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6814346447518255e-07,
+      "loss": 0.708,
+      "step": 28286
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.67830117449507e-07,
+      "loss": 0.7607,
+      "step": 28287
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.675168728149615e-07,
+      "loss": 0.7363,
+      "step": 28288
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6720373057490776e-07,
+      "loss": 0.7715,
+      "step": 28289
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.668906907327142e-07,
+      "loss": 0.7688,
+      "step": 28290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6657775329174037e-07,
+      "loss": 0.8525,
+      "step": 28291
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6626491825534914e-07,
+      "loss": 0.9531,
+      "step": 28292
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6595218562689893e-07,
+      "loss": 0.9668,
+      "step": 28293
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6563955540975145e-07,
+      "loss": 0.8809,
+      "step": 28294
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6532702760726414e-07,
+      "loss": 0.8853,
+      "step": 28295
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6501460222279304e-07,
+      "loss": 1.1348,
+      "step": 28296
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6470227925969446e-07,
+      "loss": 1.1465,
+      "step": 28297
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6439005872132457e-07,
+      "loss": 1.1553,
+      "step": 28298
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.640779406110363e-07,
+      "loss": 1.0088,
+      "step": 28299
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.637659249321802e-07,
+      "loss": 0.8823,
+      "step": 28300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.634540116881103e-07,
+      "loss": 1.0693,
+      "step": 28301
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6314220088217845e-07,
+      "loss": 0.9531,
+      "step": 28302
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.628304925177318e-07,
+      "loss": 1.127,
+      "step": 28303
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.625188865981167e-07,
+      "loss": 0.835,
+      "step": 28304
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.622073831266849e-07,
+      "loss": 1.0811,
+      "step": 28305
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.618959821067792e-07,
+      "loss": 1.0352,
+      "step": 28306
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.615846835417448e-07,
+      "loss": 1.4102,
+      "step": 28307
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6127348743492563e-07,
+      "loss": 0.9922,
+      "step": 28308
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6096239378966676e-07,
+      "loss": 0.8848,
+      "step": 28309
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6065140260930785e-07,
+      "loss": 0.7935,
+      "step": 28310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.603405138971895e-07,
+      "loss": 0.9355,
+      "step": 28311
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.600297276566501e-07,
+      "loss": 1.3008,
+      "step": 28312
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.597190438910304e-07,
+      "loss": 0.9922,
+      "step": 28313
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5940846260366766e-07,
+      "loss": 0.9653,
+      "step": 28314
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5909798379789484e-07,
+      "loss": 0.998,
+      "step": 28315
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5878760747705033e-07,
+      "loss": 0.9561,
+      "step": 28316
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.584773336444659e-07,
+      "loss": 0.9863,
+      "step": 28317
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5816716230347555e-07,
+      "loss": 0.8154,
+      "step": 28318
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5785709345741e-07,
+      "loss": 1.5664,
+      "step": 28319
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.575471271096021e-07,
+      "loss": 1.1094,
+      "step": 28320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5723726326338035e-07,
+      "loss": 0.9785,
+      "step": 28321
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.569275019220709e-07,
+      "loss": 0.6016,
+      "step": 28322
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5661784308900334e-07,
+      "loss": 0.8389,
+      "step": 28323
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5630828676750504e-07,
+      "loss": 1.3086,
+      "step": 28324
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5599883296089996e-07,
+      "loss": 0.7871,
+      "step": 28325
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.556894816725099e-07,
+      "loss": 0.9307,
+      "step": 28326
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5538023290566e-07,
+      "loss": 0.9106,
+      "step": 28327
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5507108666367316e-07,
+      "loss": 1.5273,
+      "step": 28328
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.547620429498678e-07,
+      "loss": 1.2451,
+      "step": 28329
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5445310176756466e-07,
+      "loss": 0.8828,
+      "step": 28330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.541442631200832e-07,
+      "loss": 0.7217,
+      "step": 28331
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5383552701073976e-07,
+      "loss": 1.0967,
+      "step": 28332
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.535268934428494e-07,
+      "loss": 1.2334,
+      "step": 28333
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.532183624197284e-07,
+      "loss": 1.1123,
+      "step": 28334
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5290993394469295e-07,
+      "loss": 0.7466,
+      "step": 28335
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5260160802105377e-07,
+      "loss": 1.0127,
+      "step": 28336
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.522933846521227e-07,
+      "loss": 1.251,
+      "step": 28337
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.519852638412103e-07,
+      "loss": 1.2979,
+      "step": 28338
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5167724559162963e-07,
+      "loss": 1.0303,
+      "step": 28339
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.513693299066857e-07,
+      "loss": 0.8813,
+      "step": 28340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5106151678968703e-07,
+      "loss": 0.9263,
+      "step": 28341
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5075380624394096e-07,
+      "loss": 0.7646,
+      "step": 28342
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5044619827275037e-07,
+      "loss": 0.9307,
+      "step": 28343
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.501386928794238e-07,
+      "loss": 1.1426,
+      "step": 28344
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4983129006725856e-07,
+      "loss": 0.6895,
+      "step": 28345
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4952398983956205e-07,
+      "loss": 0.999,
+      "step": 28346
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4921679219963376e-07,
+      "loss": 1.2139,
+      "step": 28347
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4890969715076893e-07,
+      "loss": 1.0908,
+      "step": 28348
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4860270469627267e-07,
+      "loss": 0.8774,
+      "step": 28349
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4829581483944006e-07,
+      "loss": 1.0498,
+      "step": 28350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4798902758356853e-07,
+      "loss": 0.8481,
+      "step": 28351
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.476823429319499e-07,
+      "loss": 0.918,
+      "step": 28352
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.473757608878815e-07,
+      "loss": 0.7798,
+      "step": 28353
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4706928145465735e-07,
+      "loss": 1.1836,
+      "step": 28354
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4676290463556817e-07,
+      "loss": 0.9604,
+      "step": 28355
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.464566304339035e-07,
+      "loss": 1.0371,
+      "step": 28356
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4615045885295527e-07,
+      "loss": 1.3691,
+      "step": 28357
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4584438989601185e-07,
+      "loss": 0.9805,
+      "step": 28358
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.455384235663596e-07,
+      "loss": 0.7568,
+      "step": 28359
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.45232559867288e-07,
+      "loss": 1.1914,
+      "step": 28360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.449267988020811e-07,
+      "loss": 1.1162,
+      "step": 28361
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4462114037402304e-07,
+      "loss": 1.3379,
+      "step": 28362
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4431558458639447e-07,
+      "loss": 0.7949,
+      "step": 28363
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.440101314424816e-07,
+      "loss": 1.0732,
+      "step": 28364
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4370478094556523e-07,
+      "loss": 1.2168,
+      "step": 28365
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.433995330989216e-07,
+      "loss": 0.8027,
+      "step": 28366
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.430943879058347e-07,
+      "loss": 1.1094,
+      "step": 28367
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4278934536957865e-07,
+      "loss": 0.8887,
+      "step": 28368
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4248440549343075e-07,
+      "loss": 0.9814,
+      "step": 28369
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.421795682806662e-07,
+      "loss": 1.0664,
+      "step": 28370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.418748337345613e-07,
+      "loss": 1.4902,
+      "step": 28371
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4157020185838895e-07,
+      "loss": 1.2412,
+      "step": 28372
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4126567265542097e-07,
+      "loss": 0.9087,
+      "step": 28373
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.409612461289248e-07,
+      "loss": 0.7021,
+      "step": 28374
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4065692228217773e-07,
+      "loss": 1.0342,
+      "step": 28375
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4035270111844385e-07,
+      "loss": 1.0073,
+      "step": 28376
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.400485826409906e-07,
+      "loss": 0.9731,
+      "step": 28377
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3974456685308864e-07,
+      "loss": 1.4355,
+      "step": 28378
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3944065375799873e-07,
+      "loss": 0.8652,
+      "step": 28379
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3913684335898933e-07,
+      "loss": 1.1201,
+      "step": 28380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.388331356593212e-07,
+      "loss": 0.916,
+      "step": 28381
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3852953066225947e-07,
+      "loss": 0.7979,
+      "step": 28382
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3822602837106265e-07,
+      "loss": 1.0107,
+      "step": 28383
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.379226287889915e-07,
+      "loss": 1.0684,
+      "step": 28384
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.376193319193045e-07,
+      "loss": 0.8125,
+      "step": 28385
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3731613776526236e-07,
+      "loss": 0.917,
+      "step": 28386
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.370130463301203e-07,
+      "loss": 0.9316,
+      "step": 28387
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3671005761713126e-07,
+      "loss": 1.1025,
+      "step": 28388
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3640717162955373e-07,
+      "loss": 0.7812,
+      "step": 28389
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3610438837064063e-07,
+      "loss": 0.9805,
+      "step": 28390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.358017078436427e-07,
+      "loss": 0.9023,
+      "step": 28391
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3549913005181075e-07,
+      "loss": 1.4062,
+      "step": 28392
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.351966549983977e-07,
+      "loss": 0.7725,
+      "step": 28393
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.34894282686652e-07,
+      "loss": 1.2061,
+      "step": 28394
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.345920131198189e-07,
+      "loss": 0.8398,
+      "step": 28395
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.342898463011469e-07,
+      "loss": 1.2715,
+      "step": 28396
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3398778223388225e-07,
+      "loss": 0.9043,
+      "step": 28397
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.336858209212713e-07,
+      "loss": 1.0547,
+      "step": 28398
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3338396236655367e-07,
+      "loss": 0.9336,
+      "step": 28399
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.330822065729734e-07,
+      "loss": 1.0029,
+      "step": 28400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.327805535437735e-07,
+      "loss": 0.875,
+      "step": 28401
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.324790032821935e-07,
+      "loss": 1.1064,
+      "step": 28402
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3217755579146983e-07,
+      "loss": 0.8794,
+      "step": 28403
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3187621107484425e-07,
+      "loss": 1.1523,
+      "step": 28404
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3157496913555197e-07,
+      "loss": 0.8032,
+      "step": 28405
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.312738299768293e-07,
+      "loss": 0.8018,
+      "step": 28406
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3097279360190926e-07,
+      "loss": 1.2168,
+      "step": 28407
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3067186001402914e-07,
+      "loss": 0.8667,
+      "step": 28408
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.303710292164187e-07,
+      "loss": 0.9795,
+      "step": 28409
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.300703012123075e-07,
+      "loss": 1.0674,
+      "step": 28410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.297696760049297e-07,
+      "loss": 1.3359,
+      "step": 28411
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.294691535975137e-07,
+      "loss": 1.085,
+      "step": 28412
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.291687339932871e-07,
+      "loss": 0.9004,
+      "step": 28413
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.288684171954749e-07,
+      "loss": 0.9424,
+      "step": 28414
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2856820320730573e-07,
+      "loss": 0.8999,
+      "step": 28415
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.282680920320037e-07,
+      "loss": 0.8052,
+      "step": 28416
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2796808367279284e-07,
+      "loss": 0.8809,
+      "step": 28417
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.27668178132894e-07,
+      "loss": 1.0098,
+      "step": 28418
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.273683754155311e-07,
+      "loss": 1.1006,
+      "step": 28419
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2706867552392173e-07,
+      "loss": 1.2168,
+      "step": 28420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.267690784612877e-07,
+      "loss": 0.9775,
+      "step": 28421
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.264695842308453e-07,
+      "loss": 0.8203,
+      "step": 28422
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2617019283581417e-07,
+      "loss": 0.9121,
+      "step": 28423
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2587090427940736e-07,
+      "loss": 0.6841,
+      "step": 28424
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2557171856484e-07,
+      "loss": 1.3105,
+      "step": 28425
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.252726356953274e-07,
+      "loss": 0.7646,
+      "step": 28426
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.249736556740824e-07,
+      "loss": 0.9814,
+      "step": 28427
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.246747785043148e-07,
+      "loss": 0.9932,
+      "step": 28428
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.243760041892353e-07,
+      "loss": 0.9609,
+      "step": 28429
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.240773327320558e-07,
+      "loss": 0.8252,
+      "step": 28430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.237787641359803e-07,
+      "loss": 0.9443,
+      "step": 28431
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.234802984042197e-07,
+      "loss": 0.7461,
+      "step": 28432
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2318193553997687e-07,
+      "loss": 0.8538,
+      "step": 28433
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2288367554645936e-07,
+      "loss": 0.6758,
+      "step": 28434
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.225855184268701e-07,
+      "loss": 0.8594,
+      "step": 28435
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.22287464184411e-07,
+      "loss": 1.1758,
+      "step": 28436
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2198951282228393e-07,
+      "loss": 1.0083,
+      "step": 28437
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.216916643436908e-07,
+      "loss": 1.1748,
+      "step": 28438
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.21393918751829e-07,
+      "loss": 1.2949,
+      "step": 28439
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.210962760498971e-07,
+      "loss": 0.834,
+      "step": 28440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2079873624109366e-07,
+      "loss": 0.9863,
+      "step": 28441
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2050129932861394e-07,
+      "loss": 0.7129,
+      "step": 28442
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2020396531565424e-07,
+      "loss": 1.1279,
+      "step": 28443
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.199067342054053e-07,
+      "loss": 0.8301,
+      "step": 28444
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.196096060010624e-07,
+      "loss": 0.7651,
+      "step": 28445
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1931258070581514e-07,
+      "loss": 0.7993,
+      "step": 28446
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1901565832285775e-07,
+      "loss": 0.916,
+      "step": 28447
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.187188388553742e-07,
+      "loss": 0.8672,
+      "step": 28448
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.184221223065588e-07,
+      "loss": 0.7842,
+      "step": 28449
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1812550867959545e-07,
+      "loss": 0.6782,
+      "step": 28450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.178289979776673e-07,
+      "loss": 0.4937,
+      "step": 28451
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.175325902039662e-07,
+      "loss": 0.96,
+      "step": 28452
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.172362853616718e-07,
+      "loss": 0.6667,
+      "step": 28453
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.169400834539683e-07,
+      "loss": 0.9316,
+      "step": 28454
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.166439844840353e-07,
+      "loss": 1.2412,
+      "step": 28455
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1634798845505585e-07,
+      "loss": 1.1602,
+      "step": 28456
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.160520953702074e-07,
+      "loss": 0.8843,
+      "step": 28457
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1575630523266963e-07,
+      "loss": 0.6743,
+      "step": 28458
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1546061804562e-07,
+      "loss": 0.6685,
+      "step": 28459
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1516503381223483e-07,
+      "loss": 1.1787,
+      "step": 28460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.148695525356894e-07,
+      "loss": 0.8682,
+      "step": 28461
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1457417421915335e-07,
+      "loss": 0.8154,
+      "step": 28462
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.142788988658053e-07,
+      "loss": 1.1621,
+      "step": 28463
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.139837264788149e-07,
+      "loss": 0.9399,
+      "step": 28464
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1368865706135186e-07,
+      "loss": 0.6411,
+      "step": 28465
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1339369061658696e-07,
+      "loss": 0.9258,
+      "step": 28466
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1309882714768765e-07,
+      "loss": 1.3438,
+      "step": 28467
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.128040666578226e-07,
+      "loss": 0.9819,
+      "step": 28468
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.125094091501558e-07,
+      "loss": 0.77,
+      "step": 28469
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.122148546278548e-07,
+      "loss": 1.1074,
+      "step": 28470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.119204030940827e-07,
+      "loss": 0.9033,
+      "step": 28471
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1162605455200123e-07,
+      "loss": 1.1133,
+      "step": 28472
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.113318090047724e-07,
+      "loss": 1.0479,
+      "step": 28473
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.110376664555593e-07,
+      "loss": 1.041,
+      "step": 28474
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.107436269075205e-07,
+      "loss": 1.2051,
+      "step": 28475
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.104496903638111e-07,
+      "loss": 1.0225,
+      "step": 28476
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.101558568275921e-07,
+      "loss": 1.2998,
+      "step": 28477
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0986212630201974e-07,
+      "loss": 1.208,
+      "step": 28478
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0956849879024816e-07,
+      "loss": 1.1816,
+      "step": 28479
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.092749742954305e-07,
+      "loss": 0.897,
+      "step": 28480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0898155282072084e-07,
+      "loss": 0.7275,
+      "step": 28481
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.086882343692711e-07,
+      "loss": 1.021,
+      "step": 28482
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0839501894423205e-07,
+      "loss": 0.9395,
+      "step": 28483
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0810190654875235e-07,
+      "loss": 0.6094,
+      "step": 28484
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.078088971859817e-07,
+      "loss": 0.7778,
+      "step": 28485
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.075159908590665e-07,
+      "loss": 1.0947,
+      "step": 28486
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.07223187571153e-07,
+      "loss": 1.2295,
+      "step": 28487
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.069304873253876e-07,
+      "loss": 0.8662,
+      "step": 28488
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.066378901249135e-07,
+      "loss": 0.96,
+      "step": 28489
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.063453959728747e-07,
+      "loss": 1.1104,
+      "step": 28490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0605300487241096e-07,
+      "loss": 0.6377,
+      "step": 28491
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.057607168266653e-07,
+      "loss": 0.7549,
+      "step": 28492
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.054685318387741e-07,
+      "loss": 0.7588,
+      "step": 28493
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0517644991188043e-07,
+      "loss": 0.9839,
+      "step": 28494
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.048844710491184e-07,
+      "loss": 1.0352,
+      "step": 28495
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.045925952536267e-07,
+      "loss": 1.2188,
+      "step": 28496
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.043008225285383e-07,
+      "loss": 1.2764,
+      "step": 28497
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.040091528769874e-07,
+      "loss": 0.8811,
+      "step": 28498
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.037175863021081e-07,
+      "loss": 0.9668,
+      "step": 28499
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0342612280703353e-07,
+      "loss": 0.9219,
+      "step": 28500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0313476239489227e-07,
+      "loss": 1.0957,
+      "step": 28501
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.02843505068814e-07,
+      "loss": 0.7734,
+      "step": 28502
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0255235083192846e-07,
+      "loss": 1.0038,
+      "step": 28503
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0226129968736316e-07,
+      "loss": 1.0312,
+      "step": 28504
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.019703516382445e-07,
+      "loss": 0.9424,
+      "step": 28505
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.016795066876955e-07,
+      "loss": 0.8384,
+      "step": 28506
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.013887648388437e-07,
+      "loss": 1.0254,
+      "step": 28507
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.010981260948099e-07,
+      "loss": 0.8018,
+      "step": 28508
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.008075904587172e-07,
+      "loss": 1.0566,
+      "step": 28509
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0051715793368526e-07,
+      "loss": 0.874,
+      "step": 28510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.00226828522835e-07,
+      "loss": 0.6426,
+      "step": 28511
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9993660222928385e-07,
+      "loss": 1.1113,
+      "step": 28512
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.996464790561505e-07,
+      "loss": 0.7695,
+      "step": 28513
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9935645900654906e-07,
+      "loss": 0.8057,
+      "step": 28514
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.990665420835993e-07,
+      "loss": 1.041,
+      "step": 28515
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9877672829041204e-07,
+      "loss": 0.9736,
+      "step": 28516
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9848701763009923e-07,
+      "loss": 0.9902,
+      "step": 28517
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.981974101057762e-07,
+      "loss": 0.8774,
+      "step": 28518
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.979079057205504e-07,
+      "loss": 0.8545,
+      "step": 28519
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.976185044775349e-07,
+      "loss": 0.9619,
+      "step": 28520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.973292063798362e-07,
+      "loss": 1.175,
+      "step": 28521
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9704001143056284e-07,
+      "loss": 1.0049,
+      "step": 28522
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.967509196328212e-07,
+      "loss": 1.1572,
+      "step": 28523
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9646193098971555e-07,
+      "loss": 0.9863,
+      "step": 28524
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9617304550435e-07,
+      "loss": 1.2832,
+      "step": 28525
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.958842631798299e-07,
+      "loss": 1.2852,
+      "step": 28526
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.95595584019256e-07,
+      "loss": 1.0391,
+      "step": 28527
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.953070080257282e-07,
+      "loss": 0.9004,
+      "step": 28528
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9501853520234614e-07,
+      "loss": 0.8618,
+      "step": 28529
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9473016555221064e-07,
+      "loss": 1.1465,
+      "step": 28530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9444189907841935e-07,
+      "loss": 0.6914,
+      "step": 28531
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.941537357840652e-07,
+      "loss": 0.6708,
+      "step": 28532
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9386567567224696e-07,
+      "loss": 0.9688,
+      "step": 28533
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9357771874605655e-07,
+      "loss": 0.9326,
+      "step": 28534
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.932898650085892e-07,
+      "loss": 0.9561,
+      "step": 28535
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9300211446293477e-07,
+      "loss": 0.793,
+      "step": 28536
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.927144671121863e-07,
+      "loss": 0.708,
+      "step": 28537
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.924269229594335e-07,
+      "loss": 1.124,
+      "step": 28538
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.921394820077618e-07,
+      "loss": 1.1816,
+      "step": 28539
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9185214426026294e-07,
+      "loss": 1.3242,
+      "step": 28540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.915649097200214e-07,
+      "loss": 0.8115,
+      "step": 28541
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.912777783901234e-07,
+      "loss": 1.1387,
+      "step": 28542
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.90990750273651e-07,
+      "loss": 1.1299,
+      "step": 28543
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9070382537368947e-07,
+      "loss": 0.8682,
+      "step": 28544
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.904170036933208e-07,
+      "loss": 0.8906,
+      "step": 28545
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.901302852356259e-07,
+      "loss": 0.9883,
+      "step": 28546
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.898436700036834e-07,
+      "loss": 0.729,
+      "step": 28547
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8955715800057303e-07,
+      "loss": 1.0381,
+      "step": 28548
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8927074922937235e-07,
+      "loss": 0.9082,
+      "step": 28549
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.889844436931556e-07,
+      "loss": 1.3613,
+      "step": 28550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8869824139500244e-07,
+      "loss": 1.167,
+      "step": 28551
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8841214233798606e-07,
+      "loss": 1.1816,
+      "step": 28552
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8812614652517733e-07,
+      "loss": 1.0869,
+      "step": 28553
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.878402539596482e-07,
+      "loss": 0.8389,
+      "step": 28554
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.875544646444707e-07,
+      "loss": 0.8447,
+      "step": 28555
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.872687785827167e-07,
+      "loss": 1.0908,
+      "step": 28556
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.869831957774517e-07,
+      "loss": 0.8604,
+      "step": 28557
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.866977162317465e-07,
+      "loss": 0.8564,
+      "step": 28558
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8641233994866525e-07,
+      "loss": 0.8271,
+      "step": 28559
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.861270669312733e-07,
+      "loss": 0.9004,
+      "step": 28560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.85841897182635e-07,
+      "loss": 1.1768,
+      "step": 28561
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.855568307058155e-07,
+      "loss": 0.8154,
+      "step": 28562
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.852718675038758e-07,
+      "loss": 0.9971,
+      "step": 28563
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8498700757987564e-07,
+      "loss": 0.9746,
+      "step": 28564
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.847022509368747e-07,
+      "loss": 1.0176,
+      "step": 28565
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.844175975779352e-07,
+      "loss": 0.8574,
+      "step": 28566
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.841330475061122e-07,
+      "loss": 0.9629,
+      "step": 28567
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.838486007244613e-07,
+      "loss": 1.0391,
+      "step": 28568
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8356425723604094e-07,
+      "loss": 0.9458,
+      "step": 28569
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8328001704390107e-07,
+      "loss": 1.0908,
+      "step": 28570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.829958801511002e-07,
+      "loss": 0.8066,
+      "step": 28571
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8271184656068604e-07,
+      "loss": 1.2676,
+      "step": 28572
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8242791627571163e-07,
+      "loss": 0.8867,
+      "step": 28573
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8214408929922674e-07,
+      "loss": 1.1455,
+      "step": 28574
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.81860365634279e-07,
+      "loss": 0.771,
+      "step": 28575
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.81576745283917e-07,
+      "loss": 0.7466,
+      "step": 28576
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.812932282511883e-07,
+      "loss": 0.9282,
+      "step": 28577
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.810098145391372e-07,
+      "loss": 1.2168,
+      "step": 28578
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.807265041508068e-07,
+      "loss": 0.9258,
+      "step": 28579
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.804432970892424e-07,
+      "loss": 1.1377,
+      "step": 28580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.801601933574839e-07,
+      "loss": 0.8652,
+      "step": 28581
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.798771929585754e-07,
+      "loss": 0.8223,
+      "step": 28582
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7959429589555453e-07,
+      "loss": 0.9072,
+      "step": 28583
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.793115021714611e-07,
+      "loss": 0.6768,
+      "step": 28584
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.790288117893326e-07,
+      "loss": 0.8945,
+      "step": 28585
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7874622475220336e-07,
+      "loss": 0.6313,
+      "step": 28586
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.784637410631109e-07,
+      "loss": 1.042,
+      "step": 28587
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7818136072509173e-07,
+      "loss": 0.8481,
+      "step": 28588
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7789908374117556e-07,
+      "loss": 1.0781,
+      "step": 28589
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.776169101143945e-07,
+      "loss": 0.6562,
+      "step": 28590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7733483984778165e-07,
+      "loss": 0.8472,
+      "step": 28591
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7705287294436677e-07,
+      "loss": 0.8945,
+      "step": 28592
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.767710094071775e-07,
+      "loss": 1.1709,
+      "step": 28593
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7648924923924137e-07,
+      "loss": 1.1797,
+      "step": 28594
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7620759244358597e-07,
+      "loss": 0.9961,
+      "step": 28595
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7592603902323553e-07,
+      "loss": 0.7598,
+      "step": 28596
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.756445889812166e-07,
+      "loss": 1.127,
+      "step": 28597
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.753632423205489e-07,
+      "loss": 0.7061,
+      "step": 28598
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.750819990442589e-07,
+      "loss": 1.0869,
+      "step": 28599
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7480085915536425e-07,
+      "loss": 1.0439,
+      "step": 28600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.745198226568847e-07,
+      "loss": 0.9463,
+      "step": 28601
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7423888955184116e-07,
+      "loss": 0.936,
+      "step": 28602
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.739580598432513e-07,
+      "loss": 0.731,
+      "step": 28603
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7367733353413036e-07,
+      "loss": 0.9756,
+      "step": 28604
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7339671062749384e-07,
+      "loss": 0.7588,
+      "step": 28605
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7311619112635587e-07,
+      "loss": 0.9248,
+      "step": 28606
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.728357750337319e-07,
+      "loss": 0.8359,
+      "step": 28607
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.725554623526317e-07,
+      "loss": 0.8525,
+      "step": 28608
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7227525308606626e-07,
+      "loss": 0.8633,
+      "step": 28609
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7199514723704754e-07,
+      "loss": 0.9053,
+      "step": 28610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7171514480858317e-07,
+      "loss": 1.0225,
+      "step": 28611
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7143524580367964e-07,
+      "loss": 1.1582,
+      "step": 28612
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.711554502253445e-07,
+      "loss": 0.9795,
+      "step": 28613
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.708757580765843e-07,
+      "loss": 0.9307,
+      "step": 28614
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.705961693604032e-07,
+      "loss": 0.957,
+      "step": 28615
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7031668407980224e-07,
+      "loss": 0.6304,
+      "step": 28616
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7003730223778455e-07,
+      "loss": 1.4766,
+      "step": 28617
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6975802383735324e-07,
+      "loss": 0.708,
+      "step": 28618
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6947884888150597e-07,
+      "loss": 0.403,
+      "step": 28619
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.691997773732414e-07,
+      "loss": 0.6846,
+      "step": 28620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.689208093155594e-07,
+      "loss": 0.9536,
+      "step": 28621
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.686419447114531e-07,
+      "loss": 0.8125,
+      "step": 28622
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.683631835639212e-07,
+      "loss": 1.1162,
+      "step": 28623
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.680845258759569e-07,
+      "loss": 0.8379,
+      "step": 28624
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6780597165055334e-07,
+      "loss": 0.8682,
+      "step": 28625
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6752752089070254e-07,
+      "loss": 1.0454,
+      "step": 28626
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.672491735993955e-07,
+      "loss": 1.0166,
+      "step": 28627
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.669709297796209e-07,
+      "loss": 0.8779,
+      "step": 28628
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.666927894343708e-07,
+      "loss": 0.873,
+      "step": 28629
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.664147525666306e-07,
+      "loss": 0.6777,
+      "step": 28630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6613681917938683e-07,
+      "loss": 1.0283,
+      "step": 28631
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.658589892756248e-07,
+      "loss": 1.0078,
+      "step": 28632
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.655812628583311e-07,
+      "loss": 1.1055,
+      "step": 28633
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6530363993048654e-07,
+      "loss": 0.7017,
+      "step": 28634
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.650261204950734e-07,
+      "loss": 0.917,
+      "step": 28635
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.647487045550746e-07,
+      "loss": 1.2256,
+      "step": 28636
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6447139211346684e-07,
+      "loss": 1.0,
+      "step": 28637
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6419418317323207e-07,
+      "loss": 0.4683,
+      "step": 28638
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6391707773734687e-07,
+      "loss": 1.1494,
+      "step": 28639
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6364007580878766e-07,
+      "loss": 0.9185,
+      "step": 28640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6336317739052997e-07,
+      "loss": 0.9268,
+      "step": 28641
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.630863824855468e-07,
+      "loss": 1.0576,
+      "step": 28642
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6280969109681263e-07,
+      "loss": 1.1357,
+      "step": 28643
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.625331032273016e-07,
+      "loss": 1.2461,
+      "step": 28644
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6225661887998145e-07,
+      "loss": 1.0283,
+      "step": 28645
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.61980238057823e-07,
+      "loss": 1.1484,
+      "step": 28646
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.617039607637962e-07,
+      "loss": 0.7466,
+      "step": 28647
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6142778700086646e-07,
+      "loss": 0.9062,
+      "step": 28648
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6115171677200246e-07,
+      "loss": 0.9893,
+      "step": 28649
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6087575008016963e-07,
+      "loss": 0.916,
+      "step": 28650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6059988692833117e-07,
+      "loss": 0.9268,
+      "step": 28651
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.603241273194513e-07,
+      "loss": 1.0703,
+      "step": 28652
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6004847125649e-07,
+      "loss": 0.9453,
+      "step": 28653
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.597729187424104e-07,
+      "loss": 0.6943,
+      "step": 28654
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.594974697801734e-07,
+      "loss": 0.7339,
+      "step": 28655
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.592221243727345e-07,
+      "loss": 0.9395,
+      "step": 28656
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.589468825230535e-07,
+      "loss": 0.875,
+      "step": 28657
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.58671744234087e-07,
+      "loss": 1.1299,
+      "step": 28658
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5839670950878925e-07,
+      "loss": 0.9492,
+      "step": 28659
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5812177835011564e-07,
+      "loss": 1.2402,
+      "step": 28660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.578469507610194e-07,
+      "loss": 1.0811,
+      "step": 28661
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5757222674445257e-07,
+      "loss": 0.8579,
+      "step": 28662
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5729760630336395e-07,
+      "loss": 1.3164,
+      "step": 28663
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.570230894407056e-07,
+      "loss": 0.8086,
+      "step": 28664
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.567486761594274e-07,
+      "loss": 1.249,
+      "step": 28665
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.564743664624759e-07,
+      "loss": 1.1211,
+      "step": 28666
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.562001603527954e-07,
+      "loss": 1.1387,
+      "step": 28667
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5592605783333455e-07,
+      "loss": 0.7305,
+      "step": 28668
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.556520589070356e-07,
+      "loss": 0.8877,
+      "step": 28669
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5537816357684384e-07,
+      "loss": 0.6978,
+      "step": 28670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.551043718456981e-07,
+      "loss": 1.165,
+      "step": 28671
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.548306837165427e-07,
+      "loss": 1.3076,
+      "step": 28672
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.545570991923153e-07,
+      "loss": 0.7949,
+      "step": 28673
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.542836182759546e-07,
+      "loss": 0.9482,
+      "step": 28674
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.540102409703994e-07,
+      "loss": 0.998,
+      "step": 28675
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.537369672785873e-07,
+      "loss": 1.0908,
+      "step": 28676
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.534637972034516e-07,
+      "loss": 0.7769,
+      "step": 28677
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5319073074792655e-07,
+      "loss": 0.9165,
+      "step": 28678
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5291776791494536e-07,
+      "loss": 1.4199,
+      "step": 28679
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5264490870744237e-07,
+      "loss": 1.1108,
+      "step": 28680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.523721531283475e-07,
+      "loss": 0.7227,
+      "step": 28681
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5209950118058835e-07,
+      "loss": 1.0166,
+      "step": 28682
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.518269528670959e-07,
+      "loss": 1.0107,
+      "step": 28683
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5155450819079675e-07,
+      "loss": 0.9692,
+      "step": 28684
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.512821671546185e-07,
+      "loss": 1.0811,
+      "step": 28685
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.510099297614844e-07,
+      "loss": 0.7332,
+      "step": 28686
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5073779601432214e-07,
+      "loss": 1.1387,
+      "step": 28687
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.504657659160526e-07,
+      "loss": 0.8535,
+      "step": 28688
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5019383946959696e-07,
+      "loss": 1.3135,
+      "step": 28689
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.499220166778783e-07,
+      "loss": 1.2871,
+      "step": 28690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.496502975438154e-07,
+      "loss": 0.8994,
+      "step": 28691
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.493786820703282e-07,
+      "loss": 0.834,
+      "step": 28692
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.49107170260331e-07,
+      "loss": 0.5464,
+      "step": 28693
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4883576211674266e-07,
+      "loss": 1.0557,
+      "step": 28694
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.485644576424796e-07,
+      "loss": 1.3428,
+      "step": 28695
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.482932568404551e-07,
+      "loss": 0.5996,
+      "step": 28696
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.480221597135802e-07,
+      "loss": 1.0449,
+      "step": 28697
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4775116626477034e-07,
+      "loss": 0.8242,
+      "step": 28698
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4748027649693536e-07,
+      "loss": 0.7993,
+      "step": 28699
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4720949041298304e-07,
+      "loss": 1.2988,
+      "step": 28700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4693880801582313e-07,
+      "loss": 0.7881,
+      "step": 28701
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.466682293083645e-07,
+      "loss": 0.9307,
+      "step": 28702
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4639775429351373e-07,
+      "loss": 0.917,
+      "step": 28703
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4612738297417404e-07,
+      "loss": 1.0059,
+      "step": 28704
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.458571153532497e-07,
+      "loss": 0.7207,
+      "step": 28705
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.455869514336474e-07,
+      "loss": 1.2666,
+      "step": 28706
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4531689121826585e-07,
+      "loss": 0.9502,
+      "step": 28707
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.450469347100061e-07,
+      "loss": 0.8057,
+      "step": 28708
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.447770819117691e-07,
+      "loss": 0.7944,
+      "step": 28709
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4450733282645255e-07,
+      "loss": 1.3418,
+      "step": 28710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.442376874569553e-07,
+      "loss": 0.6958,
+      "step": 28711
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.439681458061717e-07,
+      "loss": 1.0625,
+      "step": 28712
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4369870787699934e-07,
+      "loss": 0.7363,
+      "step": 28713
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4342937367233156e-07,
+      "loss": 0.874,
+      "step": 28714
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.431601431950593e-07,
+      "loss": 1.3989,
+      "step": 28715
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.428910164480781e-07,
+      "loss": 0.6084,
+      "step": 28716
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.426219934342767e-07,
+      "loss": 1.2422,
+      "step": 28717
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4235307415654617e-07,
+      "loss": 0.8232,
+      "step": 28718
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4208425861777415e-07,
+      "loss": 0.8613,
+      "step": 28719
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4181554682084726e-07,
+      "loss": 0.8115,
+      "step": 28720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.415469387686543e-07,
+      "loss": 1.0596,
+      "step": 28721
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.412784344640796e-07,
+      "loss": 0.9102,
+      "step": 28722
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.410100339100053e-07,
+      "loss": 0.916,
+      "step": 28723
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4074173710931804e-07,
+      "loss": 0.8477,
+      "step": 28724
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.404735440648965e-07,
+      "loss": 0.6987,
+      "step": 28725
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.402054547796241e-07,
+      "loss": 1.1084,
+      "step": 28726
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.399374692563773e-07,
+      "loss": 0.6294,
+      "step": 28727
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3966958749803934e-07,
+      "loss": 1.3281,
+      "step": 28728
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3940180950748357e-07,
+      "loss": 0.812,
+      "step": 28729
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3913413528758877e-07,
+      "loss": 0.9336,
+      "step": 28730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3886656484122817e-07,
+      "loss": 1.0635,
+      "step": 28731
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3859909817127836e-07,
+      "loss": 0.9561,
+      "step": 28732
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3833173528061035e-07,
+      "loss": 0.8408,
+      "step": 28733
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.380644761720964e-07,
+      "loss": 1.333,
+      "step": 28734
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3779732084860963e-07,
+      "loss": 0.947,
+      "step": 28735
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.375302693130156e-07,
+      "loss": 0.9131,
+      "step": 28736
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3726332156818533e-07,
+      "loss": 1.0156,
+      "step": 28737
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3699647761698763e-07,
+      "loss": 0.8203,
+      "step": 28738
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.367297374622869e-07,
+      "loss": 1.0322,
+      "step": 28739
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3646310110694967e-07,
+      "loss": 1.332,
+      "step": 28740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.36196568553836e-07,
+      "loss": 0.7036,
+      "step": 28741
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3593013980581457e-07,
+      "loss": 0.8848,
+      "step": 28742
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3566381486574427e-07,
+      "loss": 1.2598,
+      "step": 28743
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3539759373648615e-07,
+      "loss": 1.0186,
+      "step": 28744
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3513147642089905e-07,
+      "loss": 1.0254,
+      "step": 28745
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.348654629218429e-07,
+      "loss": 0.6763,
+      "step": 28746
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.345995532421753e-07,
+      "loss": 1.1572,
+      "step": 28747
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.343337473847519e-07,
+      "loss": 0.937,
+      "step": 28748
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.340680453524281e-07,
+      "loss": 1.1191,
+      "step": 28749
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.338024471480583e-07,
+      "loss": 0.8574,
+      "step": 28750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.335369527744925e-07,
+      "loss": 1.0752,
+      "step": 28751
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.332715622345861e-07,
+      "loss": 1.2451,
+      "step": 28752
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3300627553119027e-07,
+      "loss": 1.3203,
+      "step": 28753
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.327410926671526e-07,
+      "loss": 0.8252,
+      "step": 28754
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.324760136453209e-07,
+      "loss": 1.1045,
+      "step": 28755
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.32211038468544e-07,
+      "loss": 1.2334,
+      "step": 28756
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3194616713966845e-07,
+      "loss": 1.0547,
+      "step": 28757
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.316813996615398e-07,
+      "loss": 0.9644,
+      "step": 28758
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.314167360369991e-07,
+      "loss": 0.8896,
+      "step": 28759
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3115217626889294e-07,
+      "loss": 1.1924,
+      "step": 28760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.308877203600613e-07,
+      "loss": 0.937,
+      "step": 28761
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.306233683133442e-07,
+      "loss": 1.1055,
+      "step": 28762
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3035912013158147e-07,
+      "loss": 0.9341,
+      "step": 28763
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.300949758176142e-07,
+      "loss": 0.707,
+      "step": 28764
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.298309353742768e-07,
+      "loss": 0.9775,
+      "step": 28765
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.295669988044059e-07,
+      "loss": 1.2744,
+      "step": 28766
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2930316611083703e-07,
+      "loss": 1.0732,
+      "step": 28767
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2903943729640564e-07,
+      "loss": 1.2637,
+      "step": 28768
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2877581236394286e-07,
+      "loss": 1.1758,
+      "step": 28769
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.285122913162797e-07,
+      "loss": 0.9424,
+      "step": 28770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.282488741562495e-07,
+      "loss": 0.9316,
+      "step": 28771
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2798556088667886e-07,
+      "loss": 1.0073,
+      "step": 28772
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.277223515103989e-07,
+      "loss": 1.1396,
+      "step": 28773
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2745924603023395e-07,
+      "loss": 1.061,
+      "step": 28774
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2719624444901285e-07,
+      "loss": 1.0059,
+      "step": 28775
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2693334676956013e-07,
+      "loss": 0.9492,
+      "step": 28776
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2667055299469783e-07,
+      "loss": 1.2148,
+      "step": 28777
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.264078631272505e-07,
+      "loss": 1.0732,
+      "step": 28778
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.261452771700413e-07,
+      "loss": 0.8652,
+      "step": 28779
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2588279512588803e-07,
+      "loss": 0.8271,
+      "step": 28780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2562041699761074e-07,
+      "loss": 1.3164,
+      "step": 28781
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2535814278802814e-07,
+      "loss": 0.8115,
+      "step": 28782
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2509597249995915e-07,
+      "loss": 1.0967,
+      "step": 28783
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.248339061362182e-07,
+      "loss": 1.2393,
+      "step": 28784
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2457194369961973e-07,
+      "loss": 1.124,
+      "step": 28785
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2431008519297927e-07,
+      "loss": 0.748,
+      "step": 28786
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.240483306191078e-07,
+      "loss": 0.9092,
+      "step": 28787
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.237866799808187e-07,
+      "loss": 0.8389,
+      "step": 28788
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.235251332809208e-07,
+      "loss": 0.7588,
+      "step": 28789
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.232636905222264e-07,
+      "loss": 1.1338,
+      "step": 28790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.230023517075409e-07,
+      "loss": 0.8262,
+      "step": 28791
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2274111683967214e-07,
+      "loss": 1.0605,
+      "step": 28792
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.224799859214267e-07,
+      "loss": 0.8193,
+      "step": 28793
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.222189589556102e-07,
+      "loss": 0.6846,
+      "step": 28794
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2195803594502695e-07,
+      "loss": 0.877,
+      "step": 28795
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.216972168924759e-07,
+      "loss": 0.7822,
+      "step": 28796
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.214365018007637e-07,
+      "loss": 1.2354,
+      "step": 28797
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.211758906726869e-07,
+      "loss": 1.1436,
+      "step": 28798
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.209153835110479e-07,
+      "loss": 0.8208,
+      "step": 28799
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2065498031864206e-07,
+      "loss": 1.0264,
+      "step": 28800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2039468109826833e-07,
+      "loss": 1.0479,
+      "step": 28801
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2013448585272333e-07,
+      "loss": 1.1909,
+      "step": 28802
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.198743945848004e-07,
+      "loss": 0.8413,
+      "step": 28803
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.19614407297294e-07,
+      "loss": 0.6606,
+      "step": 28804
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.193545239929974e-07,
+      "loss": 0.8076,
+      "step": 28805
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.190947446747017e-07,
+      "loss": 1.0234,
+      "step": 28806
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.188350693451969e-07,
+      "loss": 0.7144,
+      "step": 28807
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1857549800727195e-07,
+      "loss": 0.8857,
+      "step": 28808
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1831603066371676e-07,
+      "loss": 1.335,
+      "step": 28809
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.18056667317318e-07,
+      "loss": 0.8794,
+      "step": 28810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1779740797086013e-07,
+      "loss": 0.9688,
+      "step": 28811
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.175382526271298e-07,
+      "loss": 0.8169,
+      "step": 28812
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1727920128891034e-07,
+      "loss": 1.2842,
+      "step": 28813
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1702025395898395e-07,
+      "loss": 1.1777,
+      "step": 28814
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1676141064013066e-07,
+      "loss": 0.8069,
+      "step": 28815
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.165026713351349e-07,
+      "loss": 0.936,
+      "step": 28816
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.162440360467722e-07,
+      "loss": 1.1504,
+      "step": 28817
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1598550477782263e-07,
+      "loss": 0.54,
+      "step": 28818
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1572707753106167e-07,
+      "loss": 0.7515,
+      "step": 28819
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1546875430926717e-07,
+      "loss": 0.8184,
+      "step": 28820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.152105351152135e-07,
+      "loss": 1.0068,
+      "step": 28821
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.149524199516729e-07,
+      "loss": 0.7124,
+      "step": 28822
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1469440882141986e-07,
+      "loss": 0.918,
+      "step": 28823
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.144365017272233e-07,
+      "loss": 1.0449,
+      "step": 28824
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1417869867185755e-07,
+      "loss": 1.0146,
+      "step": 28825
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.139209996580872e-07,
+      "loss": 0.9346,
+      "step": 28826
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.136634046886844e-07,
+      "loss": 0.8066,
+      "step": 28827
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1340591376641473e-07,
+      "loss": 1.1504,
+      "step": 28828
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.131485268940404e-07,
+      "loss": 0.8057,
+      "step": 28829
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.128912440743326e-07,
+      "loss": 0.7949,
+      "step": 28830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.126340653100524e-07,
+      "loss": 0.9521,
+      "step": 28831
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.123769906039609e-07,
+      "loss": 1.002,
+      "step": 28832
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1212001995881925e-07,
+      "loss": 0.8994,
+      "step": 28833
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.118631533773886e-07,
+      "loss": 1.1191,
+      "step": 28834
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1160639086243005e-07,
+      "loss": 0.9268,
+      "step": 28835
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.113497324166992e-07,
+      "loss": 1.1128,
+      "step": 28836
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.110931780429538e-07,
+      "loss": 0.6865,
+      "step": 28837
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1083672774395055e-07,
+      "loss": 1.2363,
+      "step": 28838
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1058038152244174e-07,
+      "loss": 0.8291,
+      "step": 28839
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.103241393811829e-07,
+      "loss": 1.4053,
+      "step": 28840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1006800132292625e-07,
+      "loss": 1.0664,
+      "step": 28841
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.098119673504241e-07,
+      "loss": 0.8594,
+      "step": 28842
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0955603746642414e-07,
+      "loss": 1.2559,
+      "step": 28843
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.093002116736754e-07,
+      "loss": 1.0254,
+      "step": 28844
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.090444899749301e-07,
+      "loss": 1.0498,
+      "step": 28845
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0878887237293155e-07,
+      "loss": 0.9966,
+      "step": 28846
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.085333588704253e-07,
+      "loss": 0.9619,
+      "step": 28847
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.082779494701582e-07,
+      "loss": 0.7231,
+      "step": 28848
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0802264417487126e-07,
+      "loss": 1.3149,
+      "step": 28849
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.07767442987309e-07,
+      "loss": 0.7319,
+      "step": 28850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.075123459102114e-07,
+      "loss": 1.1016,
+      "step": 28851
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0725735294631965e-07,
+      "loss": 1.1055,
+      "step": 28852
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.070024640983715e-07,
+      "loss": 0.9912,
+      "step": 28853
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.06747679369106e-07,
+      "loss": 0.6931,
+      "step": 28854
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.064929987612586e-07,
+      "loss": 0.8945,
+      "step": 28855
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0623842227756715e-07,
+      "loss": 0.917,
+      "step": 28856
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0598394992076507e-07,
+      "loss": 1.1357,
+      "step": 28857
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0572958169358455e-07,
+      "loss": 0.9502,
+      "step": 28858
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0547531759876014e-07,
+      "loss": 0.5593,
+      "step": 28859
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.052211576390207e-07,
+      "loss": 0.4094,
+      "step": 28860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0496710181709857e-07,
+      "loss": 1.2979,
+      "step": 28861
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0471315013572144e-07,
+      "loss": 0.9629,
+      "step": 28862
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0445930259761835e-07,
+      "loss": 0.8628,
+      "step": 28863
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0420555920551484e-07,
+      "loss": 0.9902,
+      "step": 28864
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0395191996213656e-07,
+      "loss": 0.8286,
+      "step": 28865
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.03698384870208e-07,
+      "loss": 1.3721,
+      "step": 28866
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0344495393245356e-07,
+      "loss": 1.0127,
+      "step": 28867
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0319162715159447e-07,
+      "loss": 0.9219,
+      "step": 28868
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0293840453035186e-07,
+      "loss": 1.2812,
+      "step": 28869
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.026852860714469e-07,
+      "loss": 0.624,
+      "step": 28870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.024322717775974e-07,
+      "loss": 0.9844,
+      "step": 28871
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0217936165152227e-07,
+      "loss": 0.6533,
+      "step": 28872
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.019265556959361e-07,
+      "loss": 0.9512,
+      "step": 28873
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.016738539135566e-07,
+      "loss": 1.1992,
+      "step": 28874
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.014212563070962e-07,
+      "loss": 0.7061,
+      "step": 28875
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0116876287927033e-07,
+      "loss": 1.335,
+      "step": 28876
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0091637363278914e-07,
+      "loss": 1.2529,
+      "step": 28877
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.00664088570366e-07,
+      "loss": 1.0166,
+      "step": 28878
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0041190769470986e-07,
+      "loss": 1.1299,
+      "step": 28879
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0015983100852854e-07,
+      "loss": 1.2695,
+      "step": 28880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9990785851452985e-07,
+      "loss": 1.3594,
+      "step": 28881
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.996559902154228e-07,
+      "loss": 0.668,
+      "step": 28882
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.994042261139107e-07,
+      "loss": 1.0869,
+      "step": 28883
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.991525662126971e-07,
+      "loss": 0.79,
+      "step": 28884
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.989010105144874e-07,
+      "loss": 1.0908,
+      "step": 28885
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.986495590219818e-07,
+      "loss": 1.0361,
+      "step": 28886
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9839821173788364e-07,
+      "loss": 1.4395,
+      "step": 28887
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.981469686648908e-07,
+      "loss": 0.9258,
+      "step": 28888
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.978958298057033e-07,
+      "loss": 1.2256,
+      "step": 28889
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9764479516301794e-07,
+      "loss": 0.915,
+      "step": 28890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9739386473953137e-07,
+      "loss": 0.9512,
+      "step": 28891
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.971430385379381e-07,
+      "loss": 0.9492,
+      "step": 28892
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9689231656093496e-07,
+      "loss": 0.833,
+      "step": 28893
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.966416988112131e-07,
+      "loss": 1.252,
+      "step": 28894
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.963911852914647e-07,
+      "loss": 0.835,
+      "step": 28895
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.961407760043811e-07,
+      "loss": 1.2593,
+      "step": 28896
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9589047095265223e-07,
+      "loss": 0.8311,
+      "step": 28897
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9564027013896713e-07,
+      "loss": 0.7754,
+      "step": 28898
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.953901735660125e-07,
+      "loss": 1.0215,
+      "step": 28899
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9514018123647514e-07,
+      "loss": 1.0596,
+      "step": 28900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.948902931530395e-07,
+      "loss": 0.6309,
+      "step": 28901
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9464050931839237e-07,
+      "loss": 0.9287,
+      "step": 28902
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.943908297352127e-07,
+      "loss": 0.6475,
+      "step": 28903
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9414125440618725e-07,
+      "loss": 0.7029,
+      "step": 28904
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9389178333399383e-07,
+      "loss": 0.6833,
+      "step": 28905
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.936424165213125e-07,
+      "loss": 0.71,
+      "step": 28906
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9339315397082127e-07,
+      "loss": 1.4824,
+      "step": 28907
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.931439956852e-07,
+      "loss": 0.7676,
+      "step": 28908
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.928949416671234e-07,
+      "loss": 1.0215,
+      "step": 28909
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.926459919192659e-07,
+      "loss": 0.6294,
+      "step": 28910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9239714644430316e-07,
+      "loss": 0.8784,
+      "step": 28911
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9214840524490863e-07,
+      "loss": 0.9473,
+      "step": 28912
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9189976832375343e-07,
+      "loss": 0.7095,
+      "step": 28913
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.916512356835066e-07,
+      "loss": 0.8701,
+      "step": 28914
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9140280732684047e-07,
+      "loss": 1.0674,
+      "step": 28915
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.911544832564217e-07,
+      "loss": 1.0869,
+      "step": 28916
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9090626347491824e-07,
+      "loss": 0.8745,
+      "step": 28917
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.906581479849968e-07,
+      "loss": 0.7363,
+      "step": 28918
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9041013678932304e-07,
+      "loss": 1.2295,
+      "step": 28919
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.901622298905604e-07,
+      "loss": 1.1338,
+      "step": 28920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.899144272913712e-07,
+      "loss": 0.8799,
+      "step": 28921
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.896667289944166e-07,
+      "loss": 1.0713,
+      "step": 28922
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.894191350023601e-07,
+      "loss": 1.0537,
+      "step": 28923
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8917164531785946e-07,
+      "loss": 0.9277,
+      "step": 28924
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8892425994357264e-07,
+      "loss": 1.207,
+      "step": 28925
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.886769788821575e-07,
+      "loss": 0.9126,
+      "step": 28926
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.884298021362697e-07,
+      "loss": 1.0791,
+      "step": 28927
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8818272970856596e-07,
+      "loss": 1.1709,
+      "step": 28928
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.879357616016998e-07,
+      "loss": 0.6721,
+      "step": 28929
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8768889781832343e-07,
+      "loss": 0.8296,
+      "step": 28930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8744213836108815e-07,
+      "loss": 1.3418,
+      "step": 28931
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.871954832326429e-07,
+      "loss": 0.9473,
+      "step": 28932
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.869489324356423e-07,
+      "loss": 0.9463,
+      "step": 28933
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8670248597273074e-07,
+      "loss": 0.9629,
+      "step": 28934
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.864561438465563e-07,
+      "loss": 0.7556,
+      "step": 28935
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8620990605976564e-07,
+      "loss": 1.1738,
+      "step": 28936
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8596377261500217e-07,
+      "loss": 0.8218,
+      "step": 28937
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8571774351491277e-07,
+      "loss": 1.0322,
+      "step": 28938
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.854718187621364e-07,
+      "loss": 1.0625,
+      "step": 28939
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8522599835931864e-07,
+      "loss": 1.2832,
+      "step": 28940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.849802823090975e-07,
+      "loss": 1.0518,
+      "step": 28941
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8473467061411184e-07,
+      "loss": 1.1943,
+      "step": 28942
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.844891632770008e-07,
+      "loss": 0.7773,
+      "step": 28943
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8424376030040335e-07,
+      "loss": 1.1182,
+      "step": 28944
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.83998461686954e-07,
+      "loss": 0.9541,
+      "step": 28945
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.837532674392862e-07,
+      "loss": 0.9023,
+      "step": 28946
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8350817756003567e-07,
+      "loss": 0.8955,
+      "step": 28947
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.832631920518336e-07,
+      "loss": 1.2617,
+      "step": 28948
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8301831091731457e-07,
+      "loss": 0.8984,
+      "step": 28949
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.827735341591042e-07,
+      "loss": 0.7573,
+      "step": 28950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8252886177983716e-07,
+      "loss": 1.3691,
+      "step": 28951
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8228429378213796e-07,
+      "loss": 1.2715,
+      "step": 28952
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.820398301686333e-07,
+      "loss": 0.6948,
+      "step": 28953
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.817954709419513e-07,
+      "loss": 1.0542,
+      "step": 28954
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8155121610471624e-07,
+      "loss": 0.96,
+      "step": 28955
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.813070656595518e-07,
+      "loss": 1.0723,
+      "step": 28956
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.810630196090791e-07,
+      "loss": 0.916,
+      "step": 28957
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8081907795592166e-07,
+      "loss": 0.6782,
+      "step": 28958
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8057524070269957e-07,
+      "loss": 0.9492,
+      "step": 28959
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.803315078520308e-07,
+      "loss": 0.8398,
+      "step": 28960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8008787940653314e-07,
+      "loss": 1.1084,
+      "step": 28961
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.798443553688257e-07,
+      "loss": 1.1445,
+      "step": 28962
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.796009357415219e-07,
+      "loss": 0.9297,
+      "step": 28963
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.793576205272386e-07,
+      "loss": 0.8101,
+      "step": 28964
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7911440972858806e-07,
+      "loss": 0.9395,
+      "step": 28965
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.788713033481838e-07,
+      "loss": 1.0439,
+      "step": 28966
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7862830138863595e-07,
+      "loss": 0.9482,
+      "step": 28967
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7838540385255464e-07,
+      "loss": 1.2021,
+      "step": 28968
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7814261074255e-07,
+      "loss": 0.9775,
+      "step": 28969
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7789992206123106e-07,
+      "loss": 0.9814,
+      "step": 28970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7765733781120244e-07,
+      "loss": 1.2896,
+      "step": 28971
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7741485799507084e-07,
+      "loss": 0.9111,
+      "step": 28972
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7717248261543984e-07,
+      "loss": 0.4966,
+      "step": 28973
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7693021167491506e-07,
+      "loss": 0.8691,
+      "step": 28974
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.766880451760989e-07,
+      "loss": 0.709,
+      "step": 28975
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.764459831215893e-07,
+      "loss": 0.9463,
+      "step": 28976
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7620402551398974e-07,
+      "loss": 0.959,
+      "step": 28977
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.759621723558992e-07,
+      "loss": 1.0498,
+      "step": 28978
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7572042364991227e-07,
+      "loss": 0.9585,
+      "step": 28979
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.75478779398628e-07,
+      "loss": 1.0713,
+      "step": 28980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.752372396046432e-07,
+      "loss": 1.1289,
+      "step": 28981
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.749958042705514e-07,
+      "loss": 0.9893,
+      "step": 28982
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.747544733989449e-07,
+      "loss": 0.8154,
+      "step": 28983
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.745132469924161e-07,
+      "loss": 0.8506,
+      "step": 28984
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.742721250535596e-07,
+      "loss": 0.9102,
+      "step": 28985
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.740311075849611e-07,
+      "loss": 1.04,
+      "step": 28986
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7379019458921076e-07,
+      "loss": 1.084,
+      "step": 28987
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7354938606889867e-07,
+      "loss": 0.9668,
+      "step": 28988
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.733086820266073e-07,
+      "loss": 0.876,
+      "step": 28989
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.730680824649268e-07,
+      "loss": 1.0576,
+      "step": 28990
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.728275873864383e-07,
+      "loss": 1.0161,
+      "step": 28991
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.725871967937266e-07,
+      "loss": 0.8311,
+      "step": 28992
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723469106893739e-07,
+      "loss": 1.126,
+      "step": 28993
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.721067290759594e-07,
+      "loss": 1.1377,
+      "step": 28994
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7186665195606534e-07,
+      "loss": 1.0146,
+      "step": 28995
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.716266793322697e-07,
+      "loss": 0.7371,
+      "step": 28996
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.713868112071505e-07,
+      "loss": 1.001,
+      "step": 28997
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.711470475832834e-07,
+      "loss": 1.2256,
+      "step": 28998
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7090738846324405e-07,
+      "loss": 0.8574,
+      "step": 28999
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7066783384960825e-07,
+      "loss": 0.8472,
+      "step": 29000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.704283837449484e-07,
+      "loss": 0.7324,
+      "step": 29001
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.701890381518357e-07,
+      "loss": 0.9658,
+      "step": 29002
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.699497970728426e-07,
+      "loss": 0.9824,
+      "step": 29003
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6971066051053594e-07,
+      "loss": 1.3848,
+      "step": 29004
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.694716284674892e-07,
+      "loss": 0.7559,
+      "step": 29005
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6923270094626475e-07,
+      "loss": 1.0215,
+      "step": 29006
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.689938779494339e-07,
+      "loss": 1.0254,
+      "step": 29007
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.687551594795601e-07,
+      "loss": 0.6394,
+      "step": 29008
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.685165455392047e-07,
+      "loss": 1.1284,
+      "step": 29009
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.682780361309345e-07,
+      "loss": 0.9434,
+      "step": 29010
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.680396312573108e-07,
+      "loss": 0.9727,
+      "step": 29011
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6780133092089487e-07,
+      "loss": 0.9795,
+      "step": 29012
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6756313512424357e-07,
+      "loss": 0.7847,
+      "step": 29013
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.673250438699193e-07,
+      "loss": 0.9902,
+      "step": 29014
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6708705716047665e-07,
+      "loss": 1.0537,
+      "step": 29015
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.668491749984736e-07,
+      "loss": 1.1738,
+      "step": 29016
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.666113973864648e-07,
+      "loss": 0.5054,
+      "step": 29017
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6637372432700483e-07,
+      "loss": 0.8643,
+      "step": 29018
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.661361558226461e-07,
+      "loss": 0.8896,
+      "step": 29019
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.658986918759388e-07,
+      "loss": 0.8672,
+      "step": 29020
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.656613324894386e-07,
+      "loss": 1.1353,
+      "step": 29021
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.654240776656913e-07,
+      "loss": 1.1289,
+      "step": 29022
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6518692740724606e-07,
+      "loss": 1.1338,
+      "step": 29023
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.649498817166496e-07,
+      "loss": 1.1074,
+      "step": 29024
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.647129405964499e-07,
+      "loss": 0.5352,
+      "step": 29025
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6447610404919054e-07,
+      "loss": 1.0088,
+      "step": 29026
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.642393720774161e-07,
+      "loss": 0.9844,
+      "step": 29027
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6400274468367124e-07,
+      "loss": 1.02,
+      "step": 29028
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6376622187049506e-07,
+      "loss": 1.3691,
+      "step": 29029
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.635298036404288e-07,
+      "loss": 1.085,
+      "step": 29030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.632934899960127e-07,
+      "loss": 0.9443,
+      "step": 29031
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6305728093978465e-07,
+      "loss": 0.709,
+      "step": 29032
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6282117647428276e-07,
+      "loss": 0.5186,
+      "step": 29033
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6258517660204264e-07,
+      "loss": 0.9175,
+      "step": 29034
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.623492813255979e-07,
+      "loss": 0.9043,
+      "step": 29035
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.621134906474843e-07,
+      "loss": 0.9263,
+      "step": 29036
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.618778045702353e-07,
+      "loss": 0.749,
+      "step": 29037
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6164222309638e-07,
+      "loss": 1.168,
+      "step": 29038
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6140674622845083e-07,
+      "loss": 0.7988,
+      "step": 29039
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6117137396897695e-07,
+      "loss": 0.8662,
+      "step": 29040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6093610632048515e-07,
+      "loss": 1.0859,
+      "step": 29041
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6070094328550344e-07,
+      "loss": 0.8193,
+      "step": 29042
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6046588486655976e-07,
+      "loss": 0.7715,
+      "step": 29043
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.602309310661777e-07,
+      "loss": 1.2197,
+      "step": 29044
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.599960818868785e-07,
+      "loss": 0.7852,
+      "step": 29045
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.59761337331188e-07,
+      "loss": 1.498,
+      "step": 29046
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.595266974016286e-07,
+      "loss": 0.6562,
+      "step": 29047
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.592921621007183e-07,
+      "loss": 0.7905,
+      "step": 29048
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.59057731430975e-07,
+      "loss": 1.1113,
+      "step": 29049
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.588234053949212e-07,
+      "loss": 1.0938,
+      "step": 29050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5858918399507047e-07,
+      "loss": 0.771,
+      "step": 29051
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5835506723394075e-07,
+      "loss": 0.7266,
+      "step": 29052
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5812105511404563e-07,
+      "loss": 0.7617,
+      "step": 29053
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.578871476378997e-07,
+      "loss": 0.8926,
+      "step": 29054
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.576533448080143e-07,
+      "loss": 0.9297,
+      "step": 29055
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5741964662690187e-07,
+      "loss": 0.7412,
+      "step": 29056
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.571860530970716e-07,
+      "loss": 0.978,
+      "step": 29057
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5695256422103464e-07,
+      "loss": 0.9541,
+      "step": 29058
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5671918000129805e-07,
+      "loss": 0.8633,
+      "step": 29059
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.564859004403664e-07,
+      "loss": 0.9453,
+      "step": 29060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5625272554074877e-07,
+      "loss": 0.5625,
+      "step": 29061
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5601965530494987e-07,
+      "loss": 0.792,
+      "step": 29062
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5578668973547214e-07,
+      "loss": 1.1787,
+      "step": 29063
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5555382883481693e-07,
+      "loss": 0.7573,
+      "step": 29064
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.553210726054867e-07,
+      "loss": 1.0928,
+      "step": 29065
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5508842104998265e-07,
+      "loss": 0.9082,
+      "step": 29066
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.548558741708007e-07,
+      "loss": 1.1172,
+      "step": 29067
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5462343197044215e-07,
+      "loss": 0.9141,
+      "step": 29068
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5439109445140275e-07,
+      "loss": 0.7661,
+      "step": 29069
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.541588616161772e-07,
+      "loss": 1.2661,
+      "step": 29070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.539267334672613e-07,
+      "loss": 0.8994,
+      "step": 29071
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.536947100071463e-07,
+      "loss": 0.6851,
+      "step": 29072
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.534627912383281e-07,
+      "loss": 1.0859,
+      "step": 29073
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5323097716329683e-07,
+      "loss": 0.8022,
+      "step": 29074
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5299926778453943e-07,
+      "loss": 0.9854,
+      "step": 29075
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.527676631045484e-07,
+      "loss": 1.0244,
+      "step": 29076
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.525361631258094e-07,
+      "loss": 1.1406,
+      "step": 29077
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.523047678508106e-07,
+      "loss": 0.627,
+      "step": 29078
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.520734772820366e-07,
+      "loss": 0.9531,
+      "step": 29079
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.51842291421972e-07,
+      "loss": 0.9121,
+      "step": 29080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5161121027310164e-07,
+      "loss": 1.0156,
+      "step": 29081
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5138023383790454e-07,
+      "loss": 0.9697,
+      "step": 29082
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.511493621188632e-07,
+      "loss": 0.4463,
+      "step": 29083
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5091859511846006e-07,
+      "loss": 0.9277,
+      "step": 29084
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5068793283917094e-07,
+      "loss": 0.7197,
+      "step": 29085
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.504573752834738e-07,
+      "loss": 1.3613,
+      "step": 29086
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.502269224538456e-07,
+      "loss": 1.0762,
+      "step": 29087
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4999657435276323e-07,
+      "loss": 1.0645,
+      "step": 29088
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4976633098270033e-07,
+      "loss": 0.8545,
+      "step": 29089
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4953619234612816e-07,
+      "loss": 0.8462,
+      "step": 29090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.493061584455203e-07,
+      "loss": 1.1885,
+      "step": 29091
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4907622928334706e-07,
+      "loss": 1.1367,
+      "step": 29092
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.488464048620809e-07,
+      "loss": 0.9653,
+      "step": 29093
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4861668518418536e-07,
+      "loss": 0.9717,
+      "step": 29094
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4838707025213405e-07,
+      "loss": 1.0244,
+      "step": 29095
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4815756006838834e-07,
+      "loss": 0.9126,
+      "step": 29096
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.479281546354162e-07,
+      "loss": 1.145,
+      "step": 29097
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4769885395568017e-07,
+      "loss": 0.6909,
+      "step": 29098
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.47469658031646e-07,
+      "loss": 1.041,
+      "step": 29099
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.47240566865774e-07,
+      "loss": 0.7583,
+      "step": 29100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.470115804605233e-07,
+      "loss": 0.6064,
+      "step": 29101
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4678269881835637e-07,
+      "loss": 0.8506,
+      "step": 29102
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.465539219417301e-07,
+      "loss": 0.5076,
+      "step": 29103
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.463252498331037e-07,
+      "loss": 0.7256,
+      "step": 29104
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4609668249493066e-07,
+      "loss": 1.0508,
+      "step": 29105
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.45868219929668e-07,
+      "loss": 0.8604,
+      "step": 29106
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4563986213977044e-07,
+      "loss": 0.9458,
+      "step": 29107
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4541160912768814e-07,
+      "loss": 1.02,
+      "step": 29108
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.45183460895877e-07,
+      "loss": 1.1768,
+      "step": 29109
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.44955417446785e-07,
+      "loss": 1.0654,
+      "step": 29110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.447274787828624e-07,
+      "loss": 1.0371,
+      "step": 29111
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.444996449065562e-07,
+      "loss": 1.3018,
+      "step": 29112
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4427191582031553e-07,
+      "loss": 1.0361,
+      "step": 29113
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4404429152658726e-07,
+      "loss": 1.002,
+      "step": 29114
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4381677202781393e-07,
+      "loss": 0.7158,
+      "step": 29115
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.435893573264403e-07,
+      "loss": 0.8867,
+      "step": 29116
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.43362047424911e-07,
+      "loss": 1.2451,
+      "step": 29117
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.431348423256652e-07,
+      "loss": 0.6118,
+      "step": 29118
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.429077420311443e-07,
+      "loss": 1.1631,
+      "step": 29119
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.426807465437886e-07,
+      "loss": 1.3184,
+      "step": 29120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.424538558660361e-07,
+      "loss": 1.1494,
+      "step": 29121
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4222707000032376e-07,
+      "loss": 0.9385,
+      "step": 29122
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.420003889490863e-07,
+      "loss": 1.0283,
+      "step": 29123
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4177381271476064e-07,
+      "loss": 1.1943,
+      "step": 29124
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.415473412997804e-07,
+      "loss": 0.8711,
+      "step": 29125
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.41320974706577e-07,
+      "loss": 0.8857,
+      "step": 29126
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4109471293758293e-07,
+      "loss": 1.0078,
+      "step": 29127
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.408685559952295e-07,
+      "loss": 1.1406,
+      "step": 29128
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4064250388194266e-07,
+      "loss": 0.8193,
+      "step": 29129
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4041655660015374e-07,
+      "loss": 1.0801,
+      "step": 29130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4019071415228856e-07,
+      "loss": 1.0752,
+      "step": 29131
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.399649765407741e-07,
+      "loss": 1.2686,
+      "step": 29132
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.397393437680351e-07,
+      "loss": 0.8994,
+      "step": 29133
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.395138158364918e-07,
+      "loss": 0.7935,
+      "step": 29134
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.392883927485712e-07,
+      "loss": 1.1357,
+      "step": 29135
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.390630745066924e-07,
+      "loss": 0.9883,
+      "step": 29136
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.388378611132769e-07,
+      "loss": 1.1709,
+      "step": 29137
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.386127525707427e-07,
+      "loss": 0.6895,
+      "step": 29138
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3838774888150784e-07,
+      "loss": 1.0449,
+      "step": 29139
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3816285004799045e-07,
+      "loss": 0.6294,
+      "step": 29140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3793805607260522e-07,
+      "loss": 1.0107,
+      "step": 29141
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3771336695776804e-07,
+      "loss": 0.8867,
+      "step": 29142
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3748878270589136e-07,
+      "loss": 0.9453,
+      "step": 29143
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3726430331938665e-07,
+      "loss": 1.2148,
+      "step": 29144
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.370399288006664e-07,
+      "loss": 1.0771,
+      "step": 29145
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3681565915214088e-07,
+      "loss": 0.8984,
+      "step": 29146
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3659149437622043e-07,
+      "loss": 1.0098,
+      "step": 29147
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3636743447530973e-07,
+      "loss": 1.3018,
+      "step": 29148
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3614347945181692e-07,
+      "loss": 0.8926,
+      "step": 29149
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3591962930814893e-07,
+      "loss": 0.96,
+      "step": 29150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.356958840467094e-07,
+      "loss": 0.8555,
+      "step": 29151
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3547224366990196e-07,
+      "loss": 0.9023,
+      "step": 29152
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.35248708180128e-07,
+      "loss": 0.6494,
+      "step": 29153
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.35025277579789e-07,
+      "loss": 1.1504,
+      "step": 29154
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.348019518712852e-07,
+      "loss": 0.7891,
+      "step": 29155
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3457873105701578e-07,
+      "loss": 0.8125,
+      "step": 29156
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3435561513937778e-07,
+      "loss": 1.1191,
+      "step": 29157
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.341326041207692e-07,
+      "loss": 0.9727,
+      "step": 29158
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3390969800358377e-07,
+      "loss": 0.8516,
+      "step": 29159
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3368689679021616e-07,
+      "loss": 1.1162,
+      "step": 29160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3346420048306119e-07,
+      "loss": 1.1201,
+      "step": 29161
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.332416090845102e-07,
+      "loss": 0.9668,
+      "step": 29162
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3301912259695247e-07,
+      "loss": 1.0293,
+      "step": 29163
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.327967410227805e-07,
+      "loss": 0.9463,
+      "step": 29164
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3257446436438126e-07,
+      "loss": 0.9409,
+      "step": 29165
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3235229262414395e-07,
+      "loss": 0.928,
+      "step": 29166
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.321302258044522e-07,
+      "loss": 1.0684,
+      "step": 29167
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3190826390769527e-07,
+      "loss": 1.0967,
+      "step": 29168
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3168640693625455e-07,
+      "loss": 0.792,
+      "step": 29169
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.314646548925137e-07,
+      "loss": 0.6204,
+      "step": 29170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.312430077788541e-07,
+      "loss": 0.8184,
+      "step": 29171
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3102146559765835e-07,
+      "loss": 1.2749,
+      "step": 29172
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3080002835130678e-07,
+      "loss": 1.0571,
+      "step": 29173
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3057869604217409e-07,
+      "loss": 1.2246,
+      "step": 29174
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3035746867264065e-07,
+      "loss": 0.9946,
+      "step": 29175
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.301363462450834e-07,
+      "loss": 0.7725,
+      "step": 29176
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2991532876187604e-07,
+      "loss": 1.1465,
+      "step": 29177
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2969441622539223e-07,
+      "loss": 1.1846,
+      "step": 29178
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2947360863800672e-07,
+      "loss": 0.9282,
+      "step": 29179
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2925290600208981e-07,
+      "loss": 1.063,
+      "step": 29180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2903230832001412e-07,
+      "loss": 1.1943,
+      "step": 29181
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2881181559414656e-07,
+      "loss": 0.7603,
+      "step": 29182
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.285914278268575e-07,
+      "loss": 1.043,
+      "step": 29183
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2837114502051394e-07,
+      "loss": 1.0449,
+      "step": 29184
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2815096717748175e-07,
+      "loss": 1.3535,
+      "step": 29185
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.279308943001246e-07,
+      "loss": 0.9512,
+      "step": 29186
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.277109263908095e-07,
+      "loss": 0.48,
+      "step": 29187
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2749106345189787e-07,
+      "loss": 0.7036,
+      "step": 29188
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2727130548575115e-07,
+      "loss": 0.7427,
+      "step": 29189
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.270516524947297e-07,
+      "loss": 0.6772,
+      "step": 29190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2683210448119276e-07,
+      "loss": 1.0381,
+      "step": 29191
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.266126614474995e-07,
+      "loss": 1.0352,
+      "step": 29192
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2639332339600695e-07,
+      "loss": 0.8472,
+      "step": 29193
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.26174090329071e-07,
+      "loss": 1.0869,
+      "step": 29194
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2595496224904646e-07,
+      "loss": 0.7646,
+      "step": 29195
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2573593915828584e-07,
+      "loss": 1.0654,
+      "step": 29196
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2551702105914396e-07,
+      "loss": 0.8105,
+      "step": 29197
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2529820795397228e-07,
+      "loss": 1.416,
+      "step": 29198
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2507949984512e-07,
+      "loss": 1.0508,
+      "step": 29199
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2486089673493638e-07,
+      "loss": 1.2568,
+      "step": 29200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2464239862576954e-07,
+      "loss": 0.7236,
+      "step": 29201
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.244240055199687e-07,
+      "loss": 0.6118,
+      "step": 29202
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.242057174198775e-07,
+      "loss": 1.1816,
+      "step": 29203
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.239875343278408e-07,
+      "loss": 0.9927,
+      "step": 29204
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.237694562462034e-07,
+      "loss": 0.9121,
+      "step": 29205
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2355148317730667e-07,
+      "loss": 0.9482,
+      "step": 29206
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2333361512349218e-07,
+      "loss": 1.3789,
+      "step": 29207
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.231158520871013e-07,
+      "loss": 1.2949,
+      "step": 29208
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.228981940704722e-07,
+      "loss": 0.9609,
+      "step": 29209
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2268064107594302e-07,
+      "loss": 1.1211,
+      "step": 29210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2246319310584964e-07,
+      "loss": 1.0967,
+      "step": 29211
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.222458501625302e-07,
+      "loss": 0.8047,
+      "step": 29212
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.220286122483173e-07,
+      "loss": 1.1406,
+      "step": 29213
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.218114793655457e-07,
+      "loss": 1.1465,
+      "step": 29214
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.215944515165458e-07,
+      "loss": 1.0068,
+      "step": 29215
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2137752870365125e-07,
+      "loss": 1.0938,
+      "step": 29216
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2116071092919023e-07,
+      "loss": 0.6816,
+      "step": 29217
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2094399819549305e-07,
+      "loss": 0.8701,
+      "step": 29218
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2072739050488789e-07,
+      "loss": 0.8823,
+      "step": 29219
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2051088785970064e-07,
+      "loss": 1.166,
+      "step": 29220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2029449026225612e-07,
+      "loss": 0.9639,
+      "step": 29221
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2007819771487804e-07,
+      "loss": 1.3506,
+      "step": 29222
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.198620102198934e-07,
+      "loss": 0.8691,
+      "step": 29223
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1964592777962258e-07,
+      "loss": 0.832,
+      "step": 29224
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1942995039638482e-07,
+      "loss": 1.0,
+      "step": 29225
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1921407807250272e-07,
+      "loss": 1.1279,
+      "step": 29226
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1899831081029332e-07,
+      "loss": 0.9365,
+      "step": 29227
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1878264861207588e-07,
+      "loss": 0.6934,
+      "step": 29228
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.185670914801652e-07,
+      "loss": 0.9951,
+      "step": 29229
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.183516394168783e-07,
+      "loss": 1.1694,
+      "step": 29230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1813629242452893e-07,
+      "loss": 0.7751,
+      "step": 29231
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1792105050542856e-07,
+      "loss": 0.9893,
+      "step": 29232
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1770591366189199e-07,
+      "loss": 1.0098,
+      "step": 29233
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1749088189622848e-07,
+      "loss": 0.6494,
+      "step": 29234
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1727595521074951e-07,
+      "loss": 1.0352,
+      "step": 29235
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1706113360776104e-07,
+      "loss": 0.8047,
+      "step": 29236
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1684641708957232e-07,
+      "loss": 1.0381,
+      "step": 29237
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.166318056584904e-07,
+      "loss": 1.0562,
+      "step": 29238
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.164172993168201e-07,
+      "loss": 1.0654,
+      "step": 29239
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.16202898066864e-07,
+      "loss": 0.8535,
+      "step": 29240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1598860191092697e-07,
+      "loss": 1.1719,
+      "step": 29241
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1577441085130934e-07,
+      "loss": 0.585,
+      "step": 29242
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1556032489031375e-07,
+      "loss": 1.3145,
+      "step": 29243
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1534634403023834e-07,
+      "loss": 1.2148,
+      "step": 29244
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.151324682733824e-07,
+      "loss": 1.2695,
+      "step": 29245
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1491869762204298e-07,
+      "loss": 0.9229,
+      "step": 29246
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.14705032078516e-07,
+      "loss": 0.959,
+      "step": 29247
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1449147164509632e-07,
+      "loss": 1.3574,
+      "step": 29248
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1427801632407985e-07,
+      "loss": 1.3047,
+      "step": 29249
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.140646661177581e-07,
+      "loss": 0.8208,
+      "step": 29250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.138514210284226e-07,
+      "loss": 0.8408,
+      "step": 29251
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.136382810583637e-07,
+      "loss": 0.7856,
+      "step": 29252
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1342524620987183e-07,
+      "loss": 1.0532,
+      "step": 29253
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.132123164852351e-07,
+      "loss": 1.332,
+      "step": 29254
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1299949188673952e-07,
+      "loss": 1.1602,
+      "step": 29255
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1278677241667322e-07,
+      "loss": 1.2256,
+      "step": 29256
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1257415807731995e-07,
+      "loss": 0.9932,
+      "step": 29257
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1236164887096345e-07,
+      "loss": 0.875,
+      "step": 29258
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.121492447998863e-07,
+      "loss": 1.0181,
+      "step": 29259
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1193694586637115e-07,
+      "loss": 0.9199,
+      "step": 29260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1172475207269837e-07,
+      "loss": 1.1006,
+      "step": 29261
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1151266342114506e-07,
+      "loss": 0.7207,
+      "step": 29262
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.113006799139905e-07,
+      "loss": 0.8223,
+      "step": 29263
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1108880155351397e-07,
+      "loss": 0.5493,
+      "step": 29264
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.108770283419892e-07,
+      "loss": 0.856,
+      "step": 29265
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1066536028169105e-07,
+      "loss": 1.1602,
+      "step": 29266
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1045379737489325e-07,
+      "loss": 1.0303,
+      "step": 29267
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1024233962386843e-07,
+      "loss": 0.9473,
+      "step": 29268
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.100309870308892e-07,
+      "loss": 0.6899,
+      "step": 29269
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0981973959822378e-07,
+      "loss": 1.0381,
+      "step": 29270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0960859732814364e-07,
+      "loss": 0.9663,
+      "step": 29271
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.093975602229148e-07,
+      "loss": 0.689,
+      "step": 29272
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.091866282848043e-07,
+      "loss": 0.8262,
+      "step": 29273
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0897580151607811e-07,
+      "loss": 0.6914,
+      "step": 29274
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0876507991900219e-07,
+      "loss": 0.7734,
+      "step": 29275
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0855446349583918e-07,
+      "loss": 1.1484,
+      "step": 29276
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0834395224885062e-07,
+      "loss": 0.9961,
+      "step": 29277
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0813354618029803e-07,
+      "loss": 0.8779,
+      "step": 29278
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.079232452924429e-07,
+      "loss": 1.4971,
+      "step": 29279
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0771304958754345e-07,
+      "loss": 0.8359,
+      "step": 29280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0750295906785566e-07,
+      "loss": 1.3789,
+      "step": 29281
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0729297373563995e-07,
+      "loss": 1.2861,
+      "step": 29282
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0708309359314893e-07,
+      "loss": 1.0059,
+      "step": 29283
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0687331864263637e-07,
+      "loss": 1.6055,
+      "step": 29284
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0666364888635825e-07,
+      "loss": 0.9185,
+      "step": 29285
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0645408432656722e-07,
+      "loss": 0.7998,
+      "step": 29286
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0624462496551146e-07,
+      "loss": 1.5586,
+      "step": 29287
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.060352708054425e-07,
+      "loss": 1.0474,
+      "step": 29288
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0582602184860745e-07,
+      "loss": 0.7339,
+      "step": 29289
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0561687809725782e-07,
+      "loss": 0.9502,
+      "step": 29290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0540783955363742e-07,
+      "loss": 0.9736,
+      "step": 29291
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0519890621999107e-07,
+      "loss": 0.7129,
+      "step": 29292
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0499007809856475e-07,
+      "loss": 1.2061,
+      "step": 29293
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0478135519160002e-07,
+      "loss": 0.9102,
+      "step": 29294
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0457273750134066e-07,
+      "loss": 0.9482,
+      "step": 29295
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0436422503002484e-07,
+      "loss": 0.5625,
+      "step": 29296
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0415581777989634e-07,
+      "loss": 0.8184,
+      "step": 29297
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0394751575319115e-07,
+      "loss": 0.9727,
+      "step": 29298
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0373931895214639e-07,
+      "loss": 0.874,
+      "step": 29299
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.035312273790002e-07,
+      "loss": 0.7529,
+      "step": 29300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0332324103598643e-07,
+      "loss": 1.1172,
+      "step": 29301
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0311535992533993e-07,
+      "loss": 0.7578,
+      "step": 29302
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0290758404929333e-07,
+      "loss": 1.0732,
+      "step": 29303
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0269991341007823e-07,
+      "loss": 1.1797,
+      "step": 29304
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0249234800992612e-07,
+      "loss": 0.9385,
+      "step": 29305
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0228488785106636e-07,
+      "loss": 0.7139,
+      "step": 29306
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0207753293572718e-07,
+      "loss": 1.2148,
+      "step": 29307
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.018702832661368e-07,
+      "loss": 0.8828,
+      "step": 29308
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0166313884451895e-07,
+      "loss": 0.834,
+      "step": 29309
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0145609967309964e-07,
+      "loss": 0.9688,
+      "step": 29310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0124916575410602e-07,
+      "loss": 0.9053,
+      "step": 29311
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.010423370897574e-07,
+      "loss": 0.9121,
+      "step": 29312
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0083561368227645e-07,
+      "loss": 1.5137,
+      "step": 29313
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.006289955338836e-07,
+      "loss": 1.3125,
+      "step": 29314
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.004224826467982e-07,
+      "loss": 1.0039,
+      "step": 29315
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0021607502323849e-07,
+      "loss": 0.915,
+      "step": 29316
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0000977266542265e-07,
+      "loss": 1.0459,
+      "step": 29317
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.998035755755656e-07,
+      "loss": 1.2842,
+      "step": 29318
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.995974837558823e-07,
+      "loss": 1.2168,
+      "step": 29319
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.993914972085864e-07,
+      "loss": 1.0615,
+      "step": 29320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.991856159358918e-07,
+      "loss": 0.9395,
+      "step": 29321
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9897983994000892e-07,
+      "loss": 1.0332,
+      "step": 29322
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.987741692231482e-07,
+      "loss": 1.1973,
+      "step": 29323
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9856860378751896e-07,
+      "loss": 0.9219,
+      "step": 29324
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9836314363532837e-07,
+      "loss": 1.0098,
+      "step": 29325
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9815778876878578e-07,
+      "loss": 1.127,
+      "step": 29326
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9795253919009494e-07,
+      "loss": 1.0078,
+      "step": 29327
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9774739490146189e-07,
+      "loss": 0.917,
+      "step": 29328
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.975423559050893e-07,
+      "loss": 0.9375,
+      "step": 29329
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9733742220317876e-07,
+      "loss": 1.1436,
+      "step": 29330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9713259379793402e-07,
+      "loss": 0.916,
+      "step": 29331
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9692787069155339e-07,
+      "loss": 1.1523,
+      "step": 29332
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9672325288623728e-07,
+      "loss": 1.2695,
+      "step": 29333
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9651874038418283e-07,
+      "loss": 1.082,
+      "step": 29334
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9631433318758607e-07,
+      "loss": 1.2617,
+      "step": 29335
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9611003129864413e-07,
+      "loss": 0.9258,
+      "step": 29336
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9590583471955082e-07,
+      "loss": 1.1367,
+      "step": 29337
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9570174345249994e-07,
+      "loss": 1.1992,
+      "step": 29338
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9549775749968304e-07,
+      "loss": 1.085,
+      "step": 29339
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.952938768632906e-07,
+      "loss": 0.9287,
+      "step": 29340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9509010154551532e-07,
+      "loss": 1.1172,
+      "step": 29341
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9488643154854435e-07,
+      "loss": 0.8765,
+      "step": 29342
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9468286687456484e-07,
+      "loss": 0.7671,
+      "step": 29343
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.94479407525765e-07,
+      "loss": 0.8945,
+      "step": 29344
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9427605350432976e-07,
+      "loss": 0.8955,
+      "step": 29345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9407280481244184e-07,
+      "loss": 0.7891,
+      "step": 29346
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9386966145228615e-07,
+      "loss": 0.7891,
+      "step": 29347
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9366662342604425e-07,
+      "loss": 1.105,
+      "step": 29348
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9346369073589887e-07,
+      "loss": 0.8418,
+      "step": 29349
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9326086338402606e-07,
+      "loss": 1.3369,
+      "step": 29350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9305814137260737e-07,
+      "loss": 1.0,
+      "step": 29351
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9285552470381996e-07,
+      "loss": 0.6016,
+      "step": 29352
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9265301337984099e-07,
+      "loss": 0.7852,
+      "step": 29353
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9245060740284316e-07,
+      "loss": 1.4609,
+      "step": 29354
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.922483067750036e-07,
+      "loss": 0.6228,
+      "step": 29355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9204611149849284e-07,
+      "loss": 1.2285,
+      "step": 29356
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9184402157548464e-07,
+      "loss": 1.0039,
+      "step": 29357
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9164203700814842e-07,
+      "loss": 0.8809,
+      "step": 29358
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9144015779865465e-07,
+      "loss": 1.1035,
+      "step": 29359
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.912383839491727e-07,
+      "loss": 1.251,
+      "step": 29360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9103671546186752e-07,
+      "loss": 1.125,
+      "step": 29361
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9083515233890738e-07,
+      "loss": 1.1494,
+      "step": 29362
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9063369458245716e-07,
+      "loss": 0.8062,
+      "step": 29363
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9043234219468077e-07,
+      "loss": 1.1807,
+      "step": 29364
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9023109517773973e-07,
+      "loss": 0.7544,
+      "step": 29365
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9002995353379683e-07,
+      "loss": 0.6553,
+      "step": 29366
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8982891726501362e-07,
+      "loss": 0.8096,
+      "step": 29367
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.896279863735484e-07,
+      "loss": 0.6851,
+      "step": 29368
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8942716086155942e-07,
+      "loss": 0.9668,
+      "step": 29369
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.89226440731205e-07,
+      "loss": 0.8652,
+      "step": 29370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8902582598464003e-07,
+      "loss": 1.0244,
+      "step": 29371
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8882531662401836e-07,
+      "loss": 0.8086,
+      "step": 29372
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8862491265149495e-07,
+      "loss": 1.062,
+      "step": 29373
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8842461406922474e-07,
+      "loss": 0.9541,
+      "step": 29374
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.882244208793571e-07,
+      "loss": 0.7866,
+      "step": 29375
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8802433308404145e-07,
+      "loss": 0.6758,
+      "step": 29376
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8782435068542716e-07,
+      "loss": 1.3633,
+      "step": 29377
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8762447368566582e-07,
+      "loss": 0.917,
+      "step": 29378
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8742470208690132e-07,
+      "loss": 0.6725,
+      "step": 29379
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8722503589127862e-07,
+      "loss": 1.333,
+      "step": 29380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8702547510094592e-07,
+      "loss": 0.625,
+      "step": 29381
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.868260197180438e-07,
+      "loss": 1.2227,
+      "step": 29382
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.866266697447172e-07,
+      "loss": 1.0547,
+      "step": 29383
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8642742518310442e-07,
+      "loss": 0.8442,
+      "step": 29384
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8622828603534814e-07,
+      "loss": 1.1533,
+      "step": 29385
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8602925230358782e-07,
+      "loss": 1.0312,
+      "step": 29386
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8583032398995837e-07,
+      "loss": 1.1611,
+      "step": 29387
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8563150109659923e-07,
+      "loss": 0.6792,
+      "step": 29388
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8543278362564643e-07,
+      "loss": 1.0166,
+      "step": 29389
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8523417157923275e-07,
+      "loss": 1.0977,
+      "step": 29390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.85035664959492e-07,
+      "loss": 0.7266,
+      "step": 29391
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8483726376855694e-07,
+      "loss": 0.8735,
+      "step": 29392
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.846389680085592e-07,
+      "loss": 1.5859,
+      "step": 29393
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8444077768162815e-07,
+      "loss": 0.7607,
+      "step": 29394
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8424269278989214e-07,
+      "loss": 1.1865,
+      "step": 29395
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8404471333548058e-07,
+      "loss": 1.0811,
+      "step": 29396
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8384683932051728e-07,
+      "loss": 1.2871,
+      "step": 29397
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8364907074713057e-07,
+      "loss": 0.6523,
+      "step": 29398
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8345140761744428e-07,
+      "loss": 1.126,
+      "step": 29399
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.832538499335801e-07,
+      "loss": 0.4736,
+      "step": 29400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8305639769766182e-07,
+      "loss": 0.7139,
+      "step": 29401
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.828590509118089e-07,
+      "loss": 0.9287,
+      "step": 29402
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8266180957814183e-07,
+      "loss": 1.0029,
+      "step": 29403
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8246467369878007e-07,
+      "loss": 0.8174,
+      "step": 29404
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8226764327584078e-07,
+      "loss": 1.4121,
+      "step": 29405
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8207071831144006e-07,
+      "loss": 1.2148,
+      "step": 29406
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.818738988076929e-07,
+      "loss": 1.0537,
+      "step": 29407
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8167718476671426e-07,
+      "loss": 0.8447,
+      "step": 29408
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8148057619061577e-07,
+      "loss": 0.8286,
+      "step": 29409
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.812840730815113e-07,
+      "loss": 1.377,
+      "step": 29410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8108767544151028e-07,
+      "loss": 0.8677,
+      "step": 29411
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8089138327272326e-07,
+      "loss": 0.853,
+      "step": 29412
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.806951965772563e-07,
+      "loss": 0.7075,
+      "step": 29413
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8049911535722108e-07,
+      "loss": 0.8877,
+      "step": 29414
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8030313961472034e-07,
+      "loss": 1.2334,
+      "step": 29415
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8010726935186017e-07,
+      "loss": 0.9922,
+      "step": 29416
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7991150457074446e-07,
+      "loss": 0.9453,
+      "step": 29417
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7971584527347596e-07,
+      "loss": 1.1104,
+      "step": 29418
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7952029146215743e-07,
+      "loss": 1.3086,
+      "step": 29419
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7932484313888832e-07,
+      "loss": 0.7588,
+      "step": 29420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7912950030576916e-07,
+      "loss": 1.2852,
+      "step": 29421
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7893426296489714e-07,
+      "loss": 0.8799,
+      "step": 29422
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7873913111836838e-07,
+      "loss": 1.0166,
+      "step": 29423
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.785441047682801e-07,
+      "loss": 1.2061,
+      "step": 29424
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.783491839167295e-07,
+      "loss": 0.9385,
+      "step": 29425
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7815436856580714e-07,
+      "loss": 0.876,
+      "step": 29426
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7795965871760689e-07,
+      "loss": 0.9951,
+      "step": 29427
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.777650543742193e-07,
+      "loss": 0.7524,
+      "step": 29428
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7757055553773606e-07,
+      "loss": 1.0654,
+      "step": 29429
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7737616221024657e-07,
+      "loss": 0.9604,
+      "step": 29430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7718187439383805e-07,
+      "loss": 1.1777,
+      "step": 29431
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7698769209059773e-07,
+      "loss": 0.9668,
+      "step": 29432
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7679361530261064e-07,
+      "loss": 0.792,
+      "step": 29433
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7659964403196283e-07,
+      "loss": 0.9287,
+      "step": 29434
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.76405778280736e-07,
+      "loss": 1.1865,
+      "step": 29435
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.762120180510163e-07,
+      "loss": 1.0234,
+      "step": 29436
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7601836334488087e-07,
+      "loss": 0.9277,
+      "step": 29437
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7582481416441144e-07,
+      "loss": 1.2676,
+      "step": 29438
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7563137051168743e-07,
+      "loss": 1.0342,
+      "step": 29439
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.754380323887872e-07,
+      "loss": 0.9561,
+      "step": 29440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7524479979778687e-07,
+      "loss": 1.1084,
+      "step": 29441
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7505167274076141e-07,
+      "loss": 1.1968,
+      "step": 29442
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7485865121978697e-07,
+      "loss": 0.7422,
+      "step": 29443
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.746657352369352e-07,
+      "loss": 1.2158,
+      "step": 29444
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.744729247942789e-07,
+      "loss": 1.0908,
+      "step": 29445
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7428021989388866e-07,
+      "loss": 0.9336,
+      "step": 29446
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7408762053783613e-07,
+      "loss": 0.9097,
+      "step": 29447
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7389512672818964e-07,
+      "loss": 1.2393,
+      "step": 29448
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7370273846701535e-07,
+      "loss": 1.0693,
+      "step": 29449
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7351045575638047e-07,
+      "loss": 0.6318,
+      "step": 29450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7331827859835116e-07,
+      "loss": 0.8037,
+      "step": 29451
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.731262069949924e-07,
+      "loss": 0.915,
+      "step": 29452
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7293424094836475e-07,
+      "loss": 0.7056,
+      "step": 29453
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7274238046053216e-07,
+      "loss": 0.9155,
+      "step": 29454
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7255062553355628e-07,
+      "loss": 1.3262,
+      "step": 29455
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7235897616949548e-07,
+      "loss": 1.1914,
+      "step": 29456
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7216743237040702e-07,
+      "loss": 0.8389,
+      "step": 29457
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7197599413835143e-07,
+      "loss": 1.1221,
+      "step": 29458
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7178466147538376e-07,
+      "loss": 0.9375,
+      "step": 29459
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7159343438355902e-07,
+      "loss": 1.2891,
+      "step": 29460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7140231286493113e-07,
+      "loss": 1.1895,
+      "step": 29461
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7121129692155292e-07,
+      "loss": 0.8574,
+      "step": 29462
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7102038655547825e-07,
+      "loss": 1.3164,
+      "step": 29463
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7082958176875441e-07,
+      "loss": 0.6724,
+      "step": 29464
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7063888256343307e-07,
+      "loss": 1.0039,
+      "step": 29465
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7044828894156373e-07,
+      "loss": 1.0225,
+      "step": 29466
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.702578009051914e-07,
+      "loss": 1.0527,
+      "step": 29467
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.700674184563622e-07,
+      "loss": 1.2695,
+      "step": 29468
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6987714159712343e-07,
+      "loss": 0.8584,
+      "step": 29469
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6968697032951675e-07,
+      "loss": 1.4746,
+      "step": 29470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.694969046555861e-07,
+      "loss": 1.1318,
+      "step": 29471
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6930694457737318e-07,
+      "loss": 0.6675,
+      "step": 29472
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6911709009691746e-07,
+      "loss": 1.0225,
+      "step": 29473
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6892734121625954e-07,
+      "loss": 0.7363,
+      "step": 29474
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6873769793743667e-07,
+      "loss": 0.9951,
+      "step": 29475
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6854816026248498e-07,
+      "loss": 1.1934,
+      "step": 29476
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.68358728193444e-07,
+      "loss": 0.6924,
+      "step": 29477
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6816940173234543e-07,
+      "loss": 1.0781,
+      "step": 29478
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6798018088122202e-07,
+      "loss": 0.7988,
+      "step": 29479
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6779106564210888e-07,
+      "loss": 0.8623,
+      "step": 29480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.676020560170377e-07,
+      "loss": 1.0146,
+      "step": 29481
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6741315200803799e-07,
+      "loss": 1.1489,
+      "step": 29482
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6722435361713695e-07,
+      "loss": 0.9707,
+      "step": 29483
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6703566084636525e-07,
+      "loss": 0.7046,
+      "step": 29484
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6684707369774788e-07,
+      "loss": 0.6562,
+      "step": 29485
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6665859217331105e-07,
+      "loss": 0.9155,
+      "step": 29486
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6647021627507974e-07,
+      "loss": 1.0234,
+      "step": 29487
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6628194600507907e-07,
+      "loss": 0.6528,
+      "step": 29488
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.660937813653285e-07,
+      "loss": 1.0645,
+      "step": 29489
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6590572235784863e-07,
+      "loss": 1.1426,
+      "step": 29490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.657177689846634e-07,
+      "loss": 0.7173,
+      "step": 29491
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.655299212477901e-07,
+      "loss": 0.8652,
+      "step": 29492
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.653421791492449e-07,
+      "loss": 1.127,
+      "step": 29493
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6515454269104614e-07,
+      "loss": 0.707,
+      "step": 29494
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6496701187520782e-07,
+      "loss": 0.9912,
+      "step": 29495
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6477958670374605e-07,
+      "loss": 0.8618,
+      "step": 29496
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.645922671786726e-07,
+      "loss": 0.8613,
+      "step": 29497
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.644050533020014e-07,
+      "loss": 1.0713,
+      "step": 29498
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6421794507574196e-07,
+      "loss": 0.979,
+      "step": 29499
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6403094250190487e-07,
+      "loss": 1.3271,
+      "step": 29500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6384404558249632e-07,
+      "loss": 1.1846,
+      "step": 29501
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6365725431952917e-07,
+      "loss": 0.625,
+      "step": 29502
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6347056871500512e-07,
+      "loss": 0.9282,
+      "step": 29503
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6328398877093144e-07,
+      "loss": 1.0381,
+      "step": 29504
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.630975144893121e-07,
+      "loss": 0.9287,
+      "step": 29505
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6291114587214884e-07,
+      "loss": 0.7412,
+      "step": 29506
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6272488292144672e-07,
+      "loss": 0.8921,
+      "step": 29507
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6253872563920304e-07,
+      "loss": 0.8428,
+      "step": 29508
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.623526740274195e-07,
+      "loss": 0.8232,
+      "step": 29509
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6216672808809452e-07,
+      "loss": 0.9092,
+      "step": 29510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.619808878232243e-07,
+      "loss": 0.875,
+      "step": 29511
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.61795153234805e-07,
+      "loss": 0.6982,
+      "step": 29512
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.616095243248339e-07,
+      "loss": 1.0098,
+      "step": 29513
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6142400109530275e-07,
+      "loss": 1.0459,
+      "step": 29514
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6123858354820555e-07,
+      "loss": 1.5791,
+      "step": 29515
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6105327168553286e-07,
+      "loss": 0.7383,
+      "step": 29516
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6086806550927647e-07,
+      "loss": 0.6709,
+      "step": 29517
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6068296502142587e-07,
+      "loss": 0.9795,
+      "step": 29518
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.604979702239673e-07,
+      "loss": 0.7466,
+      "step": 29519
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6031308111889133e-07,
+      "loss": 1.001,
+      "step": 29520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6012829770817972e-07,
+      "loss": 1.248,
+      "step": 29521
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5994361999382202e-07,
+      "loss": 0.8809,
+      "step": 29522
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5975904797779774e-07,
+      "loss": 1.0986,
+      "step": 29523
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5957458166209195e-07,
+      "loss": 1.2021,
+      "step": 29524
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.593902210486853e-07,
+      "loss": 0.9258,
+      "step": 29525
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.592059661395584e-07,
+      "loss": 0.9453,
+      "step": 29526
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5902181693668972e-07,
+      "loss": 0.9956,
+      "step": 29527
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5883777344205986e-07,
+      "loss": 1.0098,
+      "step": 29528
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.586538356576428e-07,
+      "loss": 1.2324,
+      "step": 29529
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5847000358541475e-07,
+      "loss": 0.9775,
+      "step": 29530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.582862772273519e-07,
+      "loss": 0.7554,
+      "step": 29531
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.58102656585426e-07,
+      "loss": 1.1074,
+      "step": 29532
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5791914166161104e-07,
+      "loss": 1.2617,
+      "step": 29533
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5773573245787765e-07,
+      "loss": 0.9375,
+      "step": 29534
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5755242897619538e-07,
+      "loss": 0.6626,
+      "step": 29535
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5736923121853487e-07,
+      "loss": 0.9175,
+      "step": 29536
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.571861391868612e-07,
+      "loss": 1.0498,
+      "step": 29537
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5700315288314282e-07,
+      "loss": 0.7427,
+      "step": 29538
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5682027230934594e-07,
+      "loss": 0.7271,
+      "step": 29539
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.566374974674345e-07,
+      "loss": 0.8916,
+      "step": 29540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5645482835937032e-07,
+      "loss": 0.75,
+      "step": 29541
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5627226498711622e-07,
+      "loss": 0.8594,
+      "step": 29542
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5608980735263512e-07,
+      "loss": 0.9844,
+      "step": 29543
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5590745545788656e-07,
+      "loss": 0.8296,
+      "step": 29544
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.557252093048256e-07,
+      "loss": 0.916,
+      "step": 29545
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5554306889541514e-07,
+      "loss": 1.2334,
+      "step": 29546
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5536103423160697e-07,
+      "loss": 0.9199,
+      "step": 29547
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5517910531535952e-07,
+      "loss": 0.5981,
+      "step": 29548
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5499728214862563e-07,
+      "loss": 0.9941,
+      "step": 29549
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5481556473335933e-07,
+      "loss": 1.2412,
+      "step": 29550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5463395307151242e-07,
+      "loss": 1.0112,
+      "step": 29551
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5445244716503327e-07,
+      "loss": 1.3105,
+      "step": 29552
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.542710470158748e-07,
+      "loss": 1.0249,
+      "step": 29553
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5408975262598546e-07,
+      "loss": 0.8613,
+      "step": 29554
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5390856399731036e-07,
+      "loss": 0.959,
+      "step": 29555
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.537274811317968e-07,
+      "loss": 1.0127,
+      "step": 29556
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5354650403138992e-07,
+      "loss": 1.2559,
+      "step": 29557
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5336563269803372e-07,
+      "loss": 1.0205,
+      "step": 29558
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.531848671336722e-07,
+      "loss": 1.0654,
+      "step": 29559
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.530042073402449e-07,
+      "loss": 0.7695,
+      "step": 29560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5282365331969472e-07,
+      "loss": 0.6187,
+      "step": 29561
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5264320507395903e-07,
+      "loss": 1.2305,
+      "step": 29562
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5246286260497733e-07,
+      "loss": 0.9092,
+      "step": 29563
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.522826259146859e-07,
+      "loss": 0.8906,
+      "step": 29564
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5210249500502205e-07,
+      "loss": 1.2197,
+      "step": 29565
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.519224698779198e-07,
+      "loss": 0.7891,
+      "step": 29566
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5174255053531317e-07,
+      "loss": 0.7666,
+      "step": 29567
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5156273697913392e-07,
+      "loss": 1.04,
+      "step": 29568
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5138302921131497e-07,
+      "loss": 1.1494,
+      "step": 29569
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.51203427233787e-07,
+      "loss": 0.8242,
+      "step": 29570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5102393104847734e-07,
+      "loss": 0.871,
+      "step": 29571
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5084454065731556e-07,
+      "loss": 0.9053,
+      "step": 29572
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.506652560622268e-07,
+      "loss": 0.7913,
+      "step": 29573
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5048607726513953e-07,
+      "loss": 1.2119,
+      "step": 29574
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.503070042679755e-07,
+      "loss": 1.1445,
+      "step": 29575
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5012803707266098e-07,
+      "loss": 0.7363,
+      "step": 29576
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4994917568111776e-07,
+      "loss": 0.7485,
+      "step": 29577
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4977042009526433e-07,
+      "loss": 1.291,
+      "step": 29578
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4959177031702465e-07,
+      "loss": 1.502,
+      "step": 29579
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.494132263483161e-07,
+      "loss": 0.6616,
+      "step": 29580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.492347881910561e-07,
+      "loss": 1.2539,
+      "step": 29581
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4905645584716187e-07,
+      "loss": 1.1738,
+      "step": 29582
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4887822931854868e-07,
+      "loss": 0.6338,
+      "step": 29583
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.487001086071316e-07,
+      "loss": 1.0156,
+      "step": 29584
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4852209371482352e-07,
+      "loss": 1.1719,
+      "step": 29585
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4834418464353627e-07,
+      "loss": 0.9238,
+      "step": 29586
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.481663813951817e-07,
+      "loss": 1.0459,
+      "step": 29587
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.479886839716682e-07,
+      "loss": 1.0117,
+      "step": 29588
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4781109237490543e-07,
+      "loss": 0.9136,
+      "step": 29589
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4763360660680293e-07,
+      "loss": 1.0801,
+      "step": 29590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4745622666926473e-07,
+      "loss": 1.2021,
+      "step": 29591
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.47278952564196e-07,
+      "loss": 1.3242,
+      "step": 29592
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4710178429350187e-07,
+      "loss": 1.0472,
+      "step": 29593
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4692472185908635e-07,
+      "loss": 1.0938,
+      "step": 29594
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4674776526285018e-07,
+      "loss": 1.2002,
+      "step": 29595
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4657091450669291e-07,
+      "loss": 0.9561,
+      "step": 29596
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.463941695925175e-07,
+      "loss": 0.6648,
+      "step": 29597
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.462175305222202e-07,
+      "loss": 1.21,
+      "step": 29598
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4604099729769838e-07,
+      "loss": 0.9629,
+      "step": 29599
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4586456992084942e-07,
+      "loss": 0.9062,
+      "step": 29600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4568824839356732e-07,
+      "loss": 1.2129,
+      "step": 29601
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.455120327177473e-07,
+      "loss": 0.9668,
+      "step": 29602
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.453359228952811e-07,
+      "loss": 0.916,
+      "step": 29603
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4515991892805946e-07,
+      "loss": 1.0566,
+      "step": 29604
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4498402081797537e-07,
+      "loss": 1.0298,
+      "step": 29605
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4480822856691723e-07,
+      "loss": 0.5415,
+      "step": 29606
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4463254217677247e-07,
+      "loss": 0.9736,
+      "step": 29607
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.444569616494307e-07,
+      "loss": 1.0479,
+      "step": 29608
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.442814869867748e-07,
+      "loss": 0.957,
+      "step": 29609
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.441061181906922e-07,
+      "loss": 0.8809,
+      "step": 29610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4393085526306471e-07,
+      "loss": 1.1865,
+      "step": 29611
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.437556982057764e-07,
+      "loss": 0.7764,
+      "step": 29612
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4358064702070795e-07,
+      "loss": 0.8213,
+      "step": 29613
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.43405701709739e-07,
+      "loss": 1.3047,
+      "step": 29614
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4323086227475025e-07,
+      "loss": 0.9424,
+      "step": 29615
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.430561287176202e-07,
+      "loss": 0.6836,
+      "step": 29616
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4288150104022402e-07,
+      "loss": 1.0488,
+      "step": 29617
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.42706979244438e-07,
+      "loss": 1.0457,
+      "step": 29618
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4253256333213726e-07,
+      "loss": 0.896,
+      "step": 29619
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.423582533051948e-07,
+      "loss": 1.1035,
+      "step": 29620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4218404916548357e-07,
+      "loss": 1.1738,
+      "step": 29621
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4200995091487424e-07,
+      "loss": 0.731,
+      "step": 29622
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4183595855523757e-07,
+      "loss": 0.9551,
+      "step": 29623
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4166207208844208e-07,
+      "loss": 1.0552,
+      "step": 29624
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4148829151635512e-07,
+      "loss": 1.0615,
+      "step": 29625
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4131461684084413e-07,
+      "loss": 0.9424,
+      "step": 29626
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4114104806377427e-07,
+      "loss": 1.083,
+      "step": 29627
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.409675851870107e-07,
+      "loss": 0.5713,
+      "step": 29628
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.407942282124164e-07,
+      "loss": 1.0996,
+      "step": 29629
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4062097714185207e-07,
+      "loss": 0.9834,
+      "step": 29630
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.404478319771807e-07,
+      "loss": 0.7607,
+      "step": 29631
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.402747927202608e-07,
+      "loss": 0.8877,
+      "step": 29632
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4010185937295195e-07,
+      "loss": 0.7573,
+      "step": 29633
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.399290319371116e-07,
+      "loss": 0.8232,
+      "step": 29634
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3975631041459492e-07,
+      "loss": 0.7393,
+      "step": 29635
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3958369480725931e-07,
+      "loss": 1.3613,
+      "step": 29636
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3941118511695663e-07,
+      "loss": 1.0449,
+      "step": 29637
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3923878134554314e-07,
+      "loss": 1.001,
+      "step": 29638
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.390664834948685e-07,
+      "loss": 1.0708,
+      "step": 29639
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3889429156678235e-07,
+      "loss": 0.8594,
+      "step": 29640
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3872220556313653e-07,
+      "loss": 0.6748,
+      "step": 29641
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3855022548577845e-07,
+      "loss": 1.0947,
+      "step": 29642
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3837835133655663e-07,
+      "loss": 1.0645,
+      "step": 29643
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3820658311731516e-07,
+      "loss": 0.8506,
+      "step": 29644
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3803492082990145e-07,
+      "loss": 0.6714,
+      "step": 29645
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3786336447615845e-07,
+      "loss": 0.8691,
+      "step": 29646
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3769191405792805e-07,
+      "loss": 1.1504,
+      "step": 29647
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.375205695770532e-07,
+      "loss": 0.9106,
+      "step": 29648
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.373493310353735e-07,
+      "loss": 0.9883,
+      "step": 29649
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.371781984347298e-07,
+      "loss": 0.9111,
+      "step": 29650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3700717177695832e-07,
+      "loss": 0.5228,
+      "step": 29651
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3683625106389652e-07,
+      "loss": 0.9902,
+      "step": 29652
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366654362973818e-07,
+      "loss": 0.4957,
+      "step": 29653
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.364947274792483e-07,
+      "loss": 0.8799,
+      "step": 29654
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3632412461133004e-07,
+      "loss": 0.4817,
+      "step": 29655
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3615362769545782e-07,
+      "loss": 1.0967,
+      "step": 29656
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.359832367334657e-07,
+      "loss": 1.0488,
+      "step": 29657
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3581295172718224e-07,
+      "loss": 1.0195,
+      "step": 29658
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3564277267843595e-07,
+      "loss": 0.9365,
+      "step": 29659
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3547269958905763e-07,
+      "loss": 0.9189,
+      "step": 29660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3530273246087133e-07,
+      "loss": 0.7422,
+      "step": 29661
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3513287129570342e-07,
+      "loss": 0.5464,
+      "step": 29662
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3496311609537905e-07,
+      "loss": 0.7529,
+      "step": 29663
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3479346686172236e-07,
+      "loss": 0.4893,
+      "step": 29664
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3462392359655518e-07,
+      "loss": 1.2324,
+      "step": 29665
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.344544863016961e-07,
+      "loss": 1.0488,
+      "step": 29666
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3428515497896922e-07,
+      "loss": 0.874,
+      "step": 29667
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3411592963019083e-07,
+      "loss": 0.875,
+      "step": 29668
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.339468102571806e-07,
+      "loss": 0.9717,
+      "step": 29669
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3377779686175152e-07,
+      "loss": 0.9795,
+      "step": 29670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3360888944572325e-07,
+      "loss": 1.4316,
+      "step": 29671
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.334400880109088e-07,
+      "loss": 1.0039,
+      "step": 29672
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3327139255912004e-07,
+      "loss": 1.0244,
+      "step": 29673
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3310280309216994e-07,
+      "loss": 1.1514,
+      "step": 29674
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3293431961187043e-07,
+      "loss": 1.083,
+      "step": 29675
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3276594212002892e-07,
+      "loss": 0.5991,
+      "step": 29676
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3259767061845618e-07,
+      "loss": 0.7979,
+      "step": 29677
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3242950510895968e-07,
+      "loss": 0.9932,
+      "step": 29678
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3226144559334464e-07,
+      "loss": 1.1172,
+      "step": 29679
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3209349207341738e-07,
+      "loss": 1.0088,
+      "step": 29680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3192564455097978e-07,
+      "loss": 0.9541,
+      "step": 29681
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3175790302783708e-07,
+      "loss": 1.0527,
+      "step": 29682
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3159026750579118e-07,
+      "loss": 1.1465,
+      "step": 29683
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3142273798664174e-07,
+      "loss": 1.1611,
+      "step": 29684
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.312553144721884e-07,
+      "loss": 0.8877,
+      "step": 29685
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.310879969642298e-07,
+      "loss": 0.9697,
+      "step": 29686
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3092078546456332e-07,
+      "loss": 0.6494,
+      "step": 29687
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3075367997498422e-07,
+      "loss": 1.0547,
+      "step": 29688
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3058668049728884e-07,
+      "loss": 1.0391,
+      "step": 29689
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.304197870332713e-07,
+      "loss": 0.8511,
+      "step": 29690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3025299958472237e-07,
+      "loss": 0.7866,
+      "step": 29691
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3008631815343288e-07,
+      "loss": 0.8164,
+      "step": 29692
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.29919742741198e-07,
+      "loss": 0.855,
+      "step": 29693
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2975327334980303e-07,
+      "loss": 1.1348,
+      "step": 29694
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2958690998103652e-07,
+      "loss": 1.1631,
+      "step": 29695
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.294206526366859e-07,
+      "loss": 0.9922,
+      "step": 29696
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2925450131853756e-07,
+      "loss": 1.082,
+      "step": 29697
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2908845602837673e-07,
+      "loss": 0.7109,
+      "step": 29698
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2892251676798528e-07,
+      "loss": 0.708,
+      "step": 29699
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2875668353914738e-07,
+      "loss": 1.0186,
+      "step": 29700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.285909563436427e-07,
+      "loss": 0.6909,
+      "step": 29701
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2842533518325318e-07,
+      "loss": 0.998,
+      "step": 29702
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2825982005975512e-07,
+      "loss": 0.8896,
+      "step": 29703
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2809441097493046e-07,
+      "loss": 0.8682,
+      "step": 29704
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2792910793055336e-07,
+      "loss": 0.9075,
+      "step": 29705
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.27763910928399e-07,
+      "loss": 0.9473,
+      "step": 29706
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.275988199702427e-07,
+      "loss": 1.1172,
+      "step": 29707
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2743383505785856e-07,
+      "loss": 1.1084,
+      "step": 29708
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2726895619301737e-07,
+      "loss": 1.0918,
+      "step": 29709
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2710418337749109e-07,
+      "loss": 0.6606,
+      "step": 29710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2693951661305049e-07,
+      "loss": 0.8857,
+      "step": 29711
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2677495590146304e-07,
+      "loss": 1.0034,
+      "step": 29712
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2661050124449514e-07,
+      "loss": 0.7202,
+      "step": 29713
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2644615264391536e-07,
+      "loss": 0.6543,
+      "step": 29714
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2628191010148893e-07,
+      "loss": 0.8755,
+      "step": 29715
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2611777361898004e-07,
+      "loss": 1.2705,
+      "step": 29716
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2595374319815167e-07,
+      "loss": 0.9658,
+      "step": 29717
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2578981884076468e-07,
+      "loss": 0.707,
+      "step": 29718
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2562600054858098e-07,
+      "loss": 1.0918,
+      "step": 29719
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2546228832336028e-07,
+      "loss": 1.0032,
+      "step": 29720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2529868216686004e-07,
+      "loss": 0.73,
+      "step": 29721
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2513518208083996e-07,
+      "loss": 1.0596,
+      "step": 29722
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2497178806705424e-07,
+      "loss": 0.6904,
+      "step": 29723
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.248085001272581e-07,
+      "loss": 0.8286,
+      "step": 29724
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.246453182632057e-07,
+      "loss": 1.168,
+      "step": 29725
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.244822424766512e-07,
+      "loss": 0.6455,
+      "step": 29726
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.243192727693454e-07,
+      "loss": 1.3809,
+      "step": 29727
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2415640914303805e-07,
+      "loss": 0.9092,
+      "step": 29728
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2399365159947886e-07,
+      "loss": 1.2793,
+      "step": 29729
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.238310001404175e-07,
+      "loss": 0.5796,
+      "step": 29730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2366845476759925e-07,
+      "loss": 1.1582,
+      "step": 29731
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2350601548277164e-07,
+      "loss": 0.9922,
+      "step": 29732
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2334368228767767e-07,
+      "loss": 1.1592,
+      "step": 29733
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2318145518406376e-07,
+      "loss": 0.9043,
+      "step": 29734
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2301933417367073e-07,
+      "loss": 1.1768,
+      "step": 29735
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.228573192582405e-07,
+      "loss": 0.8018,
+      "step": 29736
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2269541043951283e-07,
+      "loss": 0.9248,
+      "step": 29737
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.225336077192274e-07,
+      "loss": 0.626,
+      "step": 29738
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2237191109912173e-07,
+      "loss": 1.6855,
+      "step": 29739
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.222103205809333e-07,
+      "loss": 0.7959,
+      "step": 29740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.220488361663974e-07,
+      "loss": 1.1211,
+      "step": 29741
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.218874578572493e-07,
+      "loss": 0.9453,
+      "step": 29742
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.217261856552221e-07,
+      "loss": 0.9121,
+      "step": 29743
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2156501956204657e-07,
+      "loss": 1.125,
+      "step": 29744
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2140395957945694e-07,
+      "loss": 0.7173,
+      "step": 29745
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.212430057091818e-07,
+      "loss": 0.8594,
+      "step": 29746
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2108215795294863e-07,
+      "loss": 1.2554,
+      "step": 29747
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2092141631248832e-07,
+      "loss": 1.1147,
+      "step": 29748
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.207607807895239e-07,
+      "loss": 0.7432,
+      "step": 29749
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.20600251385784e-07,
+      "loss": 0.6011,
+      "step": 29750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.204398281029906e-07,
+      "loss": 1.0156,
+      "step": 29751
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2027951094286894e-07,
+      "loss": 1.1797,
+      "step": 29752
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2011929990713988e-07,
+      "loss": 0.7603,
+      "step": 29753
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.199591949975243e-07,
+      "loss": 0.8242,
+      "step": 29754
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1979919621574188e-07,
+      "loss": 1.1787,
+      "step": 29755
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.196393035635113e-07,
+      "loss": 1.1084,
+      "step": 29756
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1947951704255224e-07,
+      "loss": 1.1099,
+      "step": 29757
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1931983665457669e-07,
+      "loss": 1.0049,
+      "step": 29758
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1916026240130441e-07,
+      "loss": 0.7949,
+      "step": 29759
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1900079428444623e-07,
+      "loss": 0.7349,
+      "step": 29760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1884143230571744e-07,
+      "loss": 1.0957,
+      "step": 29761
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1868217646682889e-07,
+      "loss": 1.0479,
+      "step": 29762
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1852302676949035e-07,
+      "loss": 0.9963,
+      "step": 29763
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1836398321541265e-07,
+      "loss": 0.9131,
+      "step": 29764
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1820504580630332e-07,
+      "loss": 0.7114,
+      "step": 29765
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1804621454386988e-07,
+      "loss": 0.7666,
+      "step": 29766
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1788748942981876e-07,
+      "loss": 0.8174,
+      "step": 29767
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1772887046585524e-07,
+      "loss": 0.9941,
+      "step": 29768
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.175703576536813e-07,
+      "loss": 0.8965,
+      "step": 29769
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1741195099500113e-07,
+      "loss": 0.8906,
+      "step": 29770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1725365049151672e-07,
+      "loss": 0.8311,
+      "step": 29771
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1709545614492779e-07,
+      "loss": 0.9727,
+      "step": 29772
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.16937367956933e-07,
+      "loss": 0.6934,
+      "step": 29773
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.167793859292321e-07,
+      "loss": 0.6421,
+      "step": 29774
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1662151006352041e-07,
+      "loss": 1.2192,
+      "step": 29775
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1646374036149433e-07,
+      "loss": 0.9185,
+      "step": 29776
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1630607682484917e-07,
+      "loss": 1.0967,
+      "step": 29777
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1614851945527806e-07,
+      "loss": 1.4707,
+      "step": 29778
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1599106825447292e-07,
+      "loss": 1.1289,
+      "step": 29779
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1583372322412467e-07,
+      "loss": 1.0664,
+      "step": 29780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1567648436592527e-07,
+      "loss": 1.0068,
+      "step": 29781
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1551935168156336e-07,
+      "loss": 0.9316,
+      "step": 29782
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1536232517272538e-07,
+      "loss": 0.9561,
+      "step": 29783
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1520540484109887e-07,
+      "loss": 0.8252,
+      "step": 29784
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1504859068836915e-07,
+      "loss": 1.0605,
+      "step": 29785
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1489188271622154e-07,
+      "loss": 0.8516,
+      "step": 29786
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.14735280926338e-07,
+      "loss": 1.3076,
+      "step": 29787
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1457878532040278e-07,
+      "loss": 0.9795,
+      "step": 29788
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1442239590009451e-07,
+      "loss": 0.75,
+      "step": 29789
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1426611266709408e-07,
+      "loss": 0.7793,
+      "step": 29790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1410993562307903e-07,
+      "loss": 1.0073,
+      "step": 29791
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1395386476973024e-07,
+      "loss": 1.0713,
+      "step": 29792
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1379790010872083e-07,
+      "loss": 1.1475,
+      "step": 29793
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.136420416417272e-07,
+      "loss": 0.9385,
+      "step": 29794
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1348628937042472e-07,
+      "loss": 0.8672,
+      "step": 29795
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1333064329648424e-07,
+      "loss": 0.7319,
+      "step": 29796
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1317510342158e-07,
+      "loss": 1.0747,
+      "step": 29797
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1301966974738065e-07,
+      "loss": 1.1631,
+      "step": 29798
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.128643422755571e-07,
+      "loss": 0.8467,
+      "step": 29799
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1270912100777798e-07,
+      "loss": 1.2705,
+      "step": 29800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1255400594570864e-07,
+      "loss": 1.1724,
+      "step": 29801
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1239899709101776e-07,
+      "loss": 1.0791,
+      "step": 29802
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1224409444536843e-07,
+      "loss": 0.8623,
+      "step": 29803
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1208929801042712e-07,
+      "loss": 1.0176,
+      "step": 29804
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.119346077878536e-07,
+      "loss": 1.0156,
+      "step": 29805
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1178002377930986e-07,
+      "loss": 0.6583,
+      "step": 29806
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1162554598645903e-07,
+      "loss": 0.9609,
+      "step": 29807
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1147117441095867e-07,
+      "loss": 0.9717,
+      "step": 29808
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1131690905446635e-07,
+      "loss": 1.0078,
+      "step": 29809
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1116274991864073e-07,
+      "loss": 1.2305,
+      "step": 29810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1100869700513606e-07,
+      "loss": 1.0439,
+      "step": 29811
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1085475031560877e-07,
+      "loss": 0.8003,
+      "step": 29812
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1070090985171089e-07,
+      "loss": 1.0537,
+      "step": 29813
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1054717561509553e-07,
+      "loss": 0.9258,
+      "step": 29814
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1039354760741582e-07,
+      "loss": 0.981,
+      "step": 29815
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1024002583031823e-07,
+      "loss": 0.8887,
+      "step": 29816
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1008661028545475e-07,
+      "loss": 1.1318,
+      "step": 29817
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0993330097447297e-07,
+      "loss": 0.8179,
+      "step": 29818
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0978009789901933e-07,
+      "loss": 0.7051,
+      "step": 29819
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0962700106073921e-07,
+      "loss": 0.6899,
+      "step": 29820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0947401046127682e-07,
+      "loss": 1.1543,
+      "step": 29821
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.093211261022764e-07,
+      "loss": 0.9971,
+      "step": 29822
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0916834798538e-07,
+      "loss": 1.1758,
+      "step": 29823
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0901567611222741e-07,
+      "loss": 0.7778,
+      "step": 29824
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0886311048446174e-07,
+      "loss": 0.7888,
+      "step": 29825
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0871065110371726e-07,
+      "loss": 0.9707,
+      "step": 29826
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0855829797163597e-07,
+      "loss": 0.7095,
+      "step": 29827
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0840605108985214e-07,
+      "loss": 1.0693,
+      "step": 29828
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0825391046000111e-07,
+      "loss": 1.2178,
+      "step": 29829
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0810187608371825e-07,
+      "loss": 0.6987,
+      "step": 29830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0794994796263558e-07,
+      "loss": 0.9209,
+      "step": 29831
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0779812609838513e-07,
+      "loss": 0.959,
+      "step": 29832
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0764641049259784e-07,
+      "loss": 0.9355,
+      "step": 29833
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0749480114690458e-07,
+      "loss": 1.0044,
+      "step": 29834
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0734329806293187e-07,
+      "loss": 1.1475,
+      "step": 29835
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0719190124230838e-07,
+      "loss": 0.8916,
+      "step": 29836
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0704061068666061e-07,
+      "loss": 0.9868,
+      "step": 29837
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.068894263976128e-07,
+      "loss": 0.9795,
+      "step": 29838
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.067383483767892e-07,
+      "loss": 0.7861,
+      "step": 29839
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0658737662581298e-07,
+      "loss": 1.0195,
+      "step": 29840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0643651114630504e-07,
+      "loss": 1.0967,
+      "step": 29841
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0628575193988633e-07,
+      "loss": 1.2031,
+      "step": 29842
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0613509900817664e-07,
+      "loss": 0.9912,
+      "step": 29843
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0598455235279359e-07,
+      "loss": 0.9741,
+      "step": 29844
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0583411197535587e-07,
+      "loss": 1.0752,
+      "step": 29845
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0568377787747663e-07,
+      "loss": 0.9316,
+      "step": 29846
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0553355006077237e-07,
+      "loss": 1.0723,
+      "step": 29847
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0538342852685734e-07,
+      "loss": 0.7466,
+      "step": 29848
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.052334132773436e-07,
+      "loss": 1.3096,
+      "step": 29849
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0508350431384207e-07,
+      "loss": 0.8184,
+      "step": 29850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0493370163796368e-07,
+      "loss": 0.7236,
+      "step": 29851
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0478400525131715e-07,
+      "loss": 0.9663,
+      "step": 29852
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.046344151555112e-07,
+      "loss": 0.9248,
+      "step": 29853
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.044849313521501e-07,
+      "loss": 1.1846,
+      "step": 29854
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0433555384284366e-07,
+      "loss": 0.9316,
+      "step": 29855
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0418628262919395e-07,
+      "loss": 1.2695,
+      "step": 29856
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0403711771280301e-07,
+      "loss": 0.6685,
+      "step": 29857
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0388805909527621e-07,
+      "loss": 0.9316,
+      "step": 29858
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.037391067782123e-07,
+      "loss": 0.8857,
+      "step": 29859
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0359026076321333e-07,
+      "loss": 1.0508,
+      "step": 29860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0344152105187688e-07,
+      "loss": 0.6345,
+      "step": 29861
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0329288764580059e-07,
+      "loss": 0.9023,
+      "step": 29862
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0314436054658095e-07,
+      "loss": 0.5342,
+      "step": 29863
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0299593975581446e-07,
+      "loss": 1.168,
+      "step": 29864
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0284762527509429e-07,
+      "loss": 1.0283,
+      "step": 29865
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0269941710601361e-07,
+      "loss": 1.124,
+      "step": 29866
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0255131525016559e-07,
+      "loss": 0.8643,
+      "step": 29867
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0240331970913897e-07,
+      "loss": 1.0117,
+      "step": 29868
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0225543048452579e-07,
+      "loss": 0.811,
+      "step": 29869
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.021076475779137e-07,
+      "loss": 0.5493,
+      "step": 29870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0195997099089027e-07,
+      "loss": 1.0049,
+      "step": 29871
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0181240072503984e-07,
+      "loss": 1.0791,
+      "step": 29872
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0166493678195e-07,
+      "loss": 0.833,
+      "step": 29873
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0151757916320504e-07,
+      "loss": 0.8174,
+      "step": 29874
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0137032787038592e-07,
+      "loss": 0.7686,
+      "step": 29875
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0122318290507582e-07,
+      "loss": 0.8174,
+      "step": 29876
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0107614426885459e-07,
+      "loss": 0.835,
+      "step": 29877
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0092921196330097e-07,
+      "loss": 1.2598,
+      "step": 29878
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0078238598999367e-07,
+      "loss": 1.1328,
+      "step": 29879
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0063566635051148e-07,
+      "loss": 0.9893,
+      "step": 29880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0048905304642975e-07,
+      "loss": 1.0371,
+      "step": 29881
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.003425460793217e-07,
+      "loss": 1.209,
+      "step": 29882
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0019614545076162e-07,
+      "loss": 1.2783,
+      "step": 29883
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.000498511623238e-07,
+      "loss": 0.98,
+      "step": 29884
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.990366321557809e-08,
+      "loss": 1.25,
+      "step": 29885
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.975758161209548e-08,
+      "loss": 1.1738,
+      "step": 29886
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.961160635344469e-08,
+      "loss": 0.8169,
+      "step": 29887
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.946573744119336e-08,
+      "loss": 0.7505,
+      "step": 29888
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.931997487690914e-08,
+      "loss": 1.0898,
+      "step": 29889
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.917431866215743e-08,
+      "loss": 0.9268,
+      "step": 29890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.902876879850365e-08,
+      "loss": 0.8027,
+      "step": 29891
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.888332528750988e-08,
+      "loss": 1.4688,
+      "step": 29892
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.873798813073931e-08,
+      "loss": 1.1895,
+      "step": 29893
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.859275732975183e-08,
+      "loss": 0.751,
+      "step": 29894
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.844763288610947e-08,
+      "loss": 0.6338,
+      "step": 29895
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.830261480136994e-08,
+      "loss": 0.6392,
+      "step": 29896
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.815770307709083e-08,
+      "loss": 0.8467,
+      "step": 29897
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.801289771482981e-08,
+      "loss": 0.8467,
+      "step": 29898
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.786819871614006e-08,
+      "loss": 0.8091,
+      "step": 29899
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.772360608257814e-08,
+      "loss": 1.5332,
+      "step": 29900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.757911981569613e-08,
+      "loss": 0.8345,
+      "step": 29901
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.743473991704721e-08,
+      "loss": 0.8613,
+      "step": 29902
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.729046638818129e-08,
+      "loss": 0.6401,
+      "step": 29903
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.714629923064711e-08,
+      "loss": 1.0234,
+      "step": 29904
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.700223844599454e-08,
+      "loss": 0.9346,
+      "step": 29905
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.685828403577124e-08,
+      "loss": 0.9453,
+      "step": 29906
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.671443600152375e-08,
+      "loss": 0.7383,
+      "step": 29907
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.657069434479527e-08,
+      "loss": 1.0605,
+      "step": 29908
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.642705906713235e-08,
+      "loss": 0.8145,
+      "step": 29909
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.628353017007596e-08,
+      "loss": 1.1514,
+      "step": 29910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.614010765516935e-08,
+      "loss": 1.1011,
+      "step": 29911
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.599679152395125e-08,
+      "loss": 1.0303,
+      "step": 29912
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.585358177796267e-08,
+      "loss": 0.8057,
+      "step": 29913
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.571047841874126e-08,
+      "loss": 1.0625,
+      "step": 29914
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.556748144782469e-08,
+      "loss": 1.2617,
+      "step": 29915
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.542459086674838e-08,
+      "loss": 1.0498,
+      "step": 29916
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.528180667704778e-08,
+      "loss": 1.4302,
+      "step": 29917
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.513912888025611e-08,
+      "loss": 1.0322,
+      "step": 29918
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.499655747790549e-08,
+      "loss": 0.8013,
+      "step": 29919
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.4854092471528e-08,
+      "loss": 1.3896,
+      "step": 29920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.471173386265575e-08,
+      "loss": 0.895,
+      "step": 29921
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.456948165281421e-08,
+      "loss": 1.0332,
+      "step": 29922
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.442733584353436e-08,
+      "loss": 1.5703,
+      "step": 29923
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.428529643634165e-08,
+      "loss": 0.832,
+      "step": 29924
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.414336343276265e-08,
+      "loss": 1.0264,
+      "step": 29925
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.400153683432168e-08,
+      "loss": 0.85,
+      "step": 29926
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.385981664254085e-08,
+      "loss": 0.9487,
+      "step": 29927
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.37182028589445e-08,
+      "loss": 0.8809,
+      "step": 29928
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.357669548505367e-08,
+      "loss": 0.9341,
+      "step": 29929
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.343529452238598e-08,
+      "loss": 0.8926,
+      "step": 29930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.329399997246247e-08,
+      "loss": 0.7979,
+      "step": 29931
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.315281183680192e-08,
+      "loss": 1.0059,
+      "step": 29932
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.301173011691756e-08,
+      "loss": 1.1016,
+      "step": 29933
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.287075481432705e-08,
+      "loss": 1.1738,
+      "step": 29934
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.272988593054477e-08,
+      "loss": 1.1426,
+      "step": 29935
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.258912346708394e-08,
+      "loss": 1.0859,
+      "step": 29936
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.244846742545555e-08,
+      "loss": 0.7573,
+      "step": 29937
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.230791780717064e-08,
+      "loss": 1.4199,
+      "step": 29938
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.216747461373909e-08,
+      "loss": 0.9609,
+      "step": 29939
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.202713784666972e-08,
+      "loss": 0.5718,
+      "step": 29940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.188690750747131e-08,
+      "loss": 1.0225,
+      "step": 29941
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.17467835976471e-08,
+      "loss": 0.8652,
+      "step": 29942
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.160676611870478e-08,
+      "loss": 0.7471,
+      "step": 29943
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.146685507214759e-08,
+      "loss": 0.6997,
+      "step": 29944
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.132705045947876e-08,
+      "loss": 0.917,
+      "step": 29945
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.118735228219821e-08,
+      "loss": 0.9287,
+      "step": 29946
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.104776054180919e-08,
+      "loss": 0.874,
+      "step": 29947
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.09082752398105e-08,
+      "loss": 1.2773,
+      "step": 29948
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.076889637769759e-08,
+      "loss": 0.9229,
+      "step": 29949
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.06296239569715e-08,
+      "loss": 0.8472,
+      "step": 29950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.049045797912659e-08,
+      "loss": 0.6284,
+      "step": 29951
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.035139844565721e-08,
+      "loss": 1.3848,
+      "step": 29952
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.021244535805773e-08,
+      "loss": 1.0068,
+      "step": 29953
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.00735987178214e-08,
+      "loss": 0.9805,
+      "step": 29954
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.993485852643813e-08,
+      "loss": 0.8032,
+      "step": 29955
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.979622478539895e-08,
+      "loss": 1.1973,
+      "step": 29956
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.965769749619269e-08,
+      "loss": 1.0381,
+      "step": 29957
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.951927666030813e-08,
+      "loss": 0.9375,
+      "step": 29958
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.938096227923188e-08,
+      "loss": 0.4269,
+      "step": 29959
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.92427543544494e-08,
+      "loss": 1.1592,
+      "step": 29960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.910465288744507e-08,
+      "loss": 1.1562,
+      "step": 29961
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.896665787970326e-08,
+      "loss": 1.0449,
+      "step": 29962
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.882876933270502e-08,
+      "loss": 0.3591,
+      "step": 29963
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.869098724793246e-08,
+      "loss": 0.8374,
+      "step": 29964
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.855331162686553e-08,
+      "loss": 1.3008,
+      "step": 29965
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.841574247098194e-08,
+      "loss": 1.0908,
+      "step": 29966
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.827827978176051e-08,
+      "loss": 1.1299,
+      "step": 29967
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.814092356067782e-08,
+      "loss": 0.7129,
+      "step": 29968
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.800367380920938e-08,
+      "loss": 0.8096,
+      "step": 29969
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.786653052882843e-08,
+      "loss": 1.0938,
+      "step": 29970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.772949372100825e-08,
+      "loss": 0.9531,
+      "step": 29971
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.759256338722321e-08,
+      "loss": 1.2012,
+      "step": 29972
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.745573952894104e-08,
+      "loss": 0.9336,
+      "step": 29973
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.731902214763388e-08,
+      "loss": 1.5039,
+      "step": 29974
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.718241124476834e-08,
+      "loss": 0.8281,
+      "step": 29975
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.704590682181213e-08,
+      "loss": 1.1885,
+      "step": 29976
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.690950888023409e-08,
+      "loss": 1.2988,
+      "step": 29977
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.677321742149525e-08,
+      "loss": 0.8223,
+      "step": 29978
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.663703244706333e-08,
+      "loss": 0.9307,
+      "step": 29979
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.650095395839941e-08,
+      "loss": 1.0596,
+      "step": 29980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.636498195696453e-08,
+      "loss": 1.0205,
+      "step": 29981
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.622911644421971e-08,
+      "loss": 1.1387,
+      "step": 29982
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.609335742162605e-08,
+      "loss": 1.1934,
+      "step": 29983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.595770489064014e-08,
+      "loss": 1.0693,
+      "step": 29984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.58221588527186e-08,
+      "loss": 0.7041,
+      "step": 29985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.568671930931915e-08,
+      "loss": 1.04,
+      "step": 29986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.555138626189619e-08,
+      "loss": 1.1807,
+      "step": 29987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.541615971190299e-08,
+      "loss": 1.1582,
+      "step": 29988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.52810396607906e-08,
+      "loss": 0.8676,
+      "step": 29989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.514602611001344e-08,
+      "loss": 0.9307,
+      "step": 29990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.501111906102033e-08,
+      "loss": 1.0039,
+      "step": 29991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.487631851525901e-08,
+      "loss": 1.0703,
+      "step": 29992
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.474162447417944e-08,
+      "loss": 1.0303,
+      "step": 29993
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.460703693922934e-08,
+      "loss": 1.0381,
+      "step": 29994
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.447255591185199e-08,
+      "loss": 0.9287,
+      "step": 29995
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.433818139349181e-08,
+      "loss": 0.7896,
+      "step": 29996
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.42039133855932e-08,
+      "loss": 1.6055,
+      "step": 29997
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.406975188959943e-08,
+      "loss": 1.1709,
+      "step": 29998
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.393569690695158e-08,
+      "loss": 1.3066,
+      "step": 29999
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.380174843908629e-08,
+      "loss": 0.9717,
+      "step": 30000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.366790648744682e-08,
+      "loss": 0.7505,
+      "step": 30001
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.353417105346762e-08,
+      "loss": 1.1182,
+      "step": 30002
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.34005421385864e-08,
+      "loss": 1.2246,
+      "step": 30003
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.326701974423867e-08,
+      "loss": 1.1592,
+      "step": 30004
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.313360387185998e-08,
+      "loss": 0.9043,
+      "step": 30005
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.300029452288028e-08,
+      "loss": 0.7456,
+      "step": 30006
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.286709169873396e-08,
+      "loss": 0.9414,
+      "step": 30007
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.273399540085104e-08,
+      "loss": 1.3438,
+      "step": 30008
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.260100563066143e-08,
+      "loss": 1.2139,
+      "step": 30009
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.246812238959401e-08,
+      "loss": 0.8911,
+      "step": 30010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.23353456790743e-08,
+      "loss": 0.6348,
+      "step": 30011
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.220267550053007e-08,
+      "loss": 0.7207,
+      "step": 30012
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.207011185538793e-08,
+      "loss": 1.2598,
+      "step": 30013
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.193765474506898e-08,
+      "loss": 0.8887,
+      "step": 30014
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.180530417099764e-08,
+      "loss": 0.9043,
+      "step": 30015
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.167306013459498e-08,
+      "loss": 1.3867,
+      "step": 30016
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.154092263728208e-08,
+      "loss": 0.7866,
+      "step": 30017
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.140889168047672e-08,
+      "loss": 0.8447,
+      "step": 30018
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.127696726559886e-08,
+      "loss": 0.8179,
+      "step": 30019
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.114514939406626e-08,
+      "loss": 1.3262,
+      "step": 30020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.101343806729223e-08,
+      "loss": 1.1533,
+      "step": 30021
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.088183328669341e-08,
+      "loss": 1.1377,
+      "step": 30022
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.075033505368312e-08,
+      "loss": 0.7969,
+      "step": 30023
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.061894336967357e-08,
+      "loss": 1.1514,
+      "step": 30024
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.048765823607697e-08,
+      "loss": 0.6777,
+      "step": 30025
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.035647965430215e-08,
+      "loss": 0.5969,
+      "step": 30026
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.022540762576025e-08,
+      "loss": 1.0889,
+      "step": 30027
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.009444215185569e-08,
+      "loss": 0.8215,
+      "step": 30028
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.996358323399956e-08,
+      "loss": 0.7456,
+      "step": 30029
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.983283087359406e-08,
+      "loss": 0.8691,
+      "step": 30030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.970218507204586e-08,
+      "loss": 0.7227,
+      "step": 30031
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.957164583075716e-08,
+      "loss": 1.0645,
+      "step": 30032
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.944121315113018e-08,
+      "loss": 1.1929,
+      "step": 30033
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.931088703456601e-08,
+      "loss": 0.9277,
+      "step": 30034
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.918066748246578e-08,
+      "loss": 1.1367,
+      "step": 30035
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.905055449622612e-08,
+      "loss": 0.8027,
+      "step": 30036
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.892054807724703e-08,
+      "loss": 1.166,
+      "step": 30037
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.879064822692295e-08,
+      "loss": 0.8091,
+      "step": 30038
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.866085494665054e-08,
+      "loss": 1.1875,
+      "step": 30039
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.853116823782314e-08,
+      "loss": 0.8682,
+      "step": 30040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.840158810183406e-08,
+      "loss": 0.8589,
+      "step": 30041
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.827211454007666e-08,
+      "loss": 1.207,
+      "step": 30042
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.814274755393869e-08,
+      "loss": 0.8936,
+      "step": 30043
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.801348714481238e-08,
+      "loss": 0.6992,
+      "step": 30044
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.78843333140844e-08,
+      "loss": 0.6519,
+      "step": 30045
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.775528606314365e-08,
+      "loss": 1.001,
+      "step": 30046
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.762634539337566e-08,
+      "loss": 1.2549,
+      "step": 30047
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.74975113061649e-08,
+      "loss": 1.1426,
+      "step": 30048
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.736878380289581e-08,
+      "loss": 0.7939,
+      "step": 30049
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.724016288495173e-08,
+      "loss": 0.8633,
+      "step": 30050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.711164855371378e-08,
+      "loss": 0.6802,
+      "step": 30051
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.698324081056085e-08,
+      "loss": 0.6084,
+      "step": 30052
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.685493965687519e-08,
+      "loss": 0.4775,
+      "step": 30053
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.672674509403233e-08,
+      "loss": 1.2109,
+      "step": 30054
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.659865712341119e-08,
+      "loss": 0.6455,
+      "step": 30055
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.647067574638623e-08,
+      "loss": 1.0859,
+      "step": 30056
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.6342800964333e-08,
+      "loss": 1.2109,
+      "step": 30057
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.621503277862485e-08,
+      "loss": 0.7349,
+      "step": 30058
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.608737119063292e-08,
+      "loss": 0.752,
+      "step": 30059
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.595981620173165e-08,
+      "loss": 0.8228,
+      "step": 30060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.583236781328884e-08,
+      "loss": 0.751,
+      "step": 30061
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.570502602667451e-08,
+      "loss": 0.9082,
+      "step": 30062
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.557779084325534e-08,
+      "loss": 1.2266,
+      "step": 30063
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.545066226439801e-08,
+      "loss": 0.7695,
+      "step": 30064
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.532364029147032e-08,
+      "loss": 0.8354,
+      "step": 30065
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.519672492583341e-08,
+      "loss": 1.3516,
+      "step": 30066
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.506991616885395e-08,
+      "loss": 0.8682,
+      "step": 30067
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.494321402189197e-08,
+      "loss": 0.8662,
+      "step": 30068
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.48166184863086e-08,
+      "loss": 0.875,
+      "step": 30069
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.469012956346389e-08,
+      "loss": 1.1992,
+      "step": 30070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.456374725471782e-08,
+      "loss": 1.0371,
+      "step": 30071
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.4437471561426e-08,
+      "loss": 0.915,
+      "step": 30072
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.431130248494622e-08,
+      "loss": 1.3174,
+      "step": 30073
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.418524002663296e-08,
+      "loss": 0.9072,
+      "step": 30074
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.405928418783959e-08,
+      "loss": 0.8701,
+      "step": 30075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.393343496992056e-08,
+      "loss": 0.769,
+      "step": 30076
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.380769237422702e-08,
+      "loss": 0.8613,
+      "step": 30077
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.368205640211013e-08,
+      "loss": 1.2432,
+      "step": 30078
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.355652705491989e-08,
+      "loss": 1.1084,
+      "step": 30079
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.343110433400191e-08,
+      "loss": 0.5251,
+      "step": 30080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.330578824070622e-08,
+      "loss": 1.1328,
+      "step": 30081
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.318057877637841e-08,
+      "loss": 1.1367,
+      "step": 30082
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.305547594236406e-08,
+      "loss": 1.0322,
+      "step": 30083
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.293047974000433e-08,
+      "loss": 1.2432,
+      "step": 30084
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.280559017064481e-08,
+      "loss": 1.3789,
+      "step": 30085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.268080723562554e-08,
+      "loss": 0.8848,
+      "step": 30086
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.255613093628656e-08,
+      "loss": 0.9072,
+      "step": 30087
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.2431561273969e-08,
+      "loss": 1.0459,
+      "step": 30088
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.230709825000848e-08,
+      "loss": 1.1147,
+      "step": 30089
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.218274186574281e-08,
+      "loss": 0.4707,
+      "step": 30090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.205849212250981e-08,
+      "loss": 0.6626,
+      "step": 30091
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.193434902164064e-08,
+      "loss": 0.9531,
+      "step": 30092
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.181031256447202e-08,
+      "loss": 1.0156,
+      "step": 30093
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.168638275233397e-08,
+      "loss": 1.1572,
+      "step": 30094
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.156255958655878e-08,
+      "loss": 0.8569,
+      "step": 30095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.143884306847537e-08,
+      "loss": 0.7363,
+      "step": 30096
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.131523319941491e-08,
+      "loss": 1.1992,
+      "step": 30097
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.119172998070412e-08,
+      "loss": 0.8657,
+      "step": 30098
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.106833341366859e-08,
+      "loss": 1.0684,
+      "step": 30099
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.094504349963394e-08,
+      "loss": 1.2441,
+      "step": 30100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.082186023992688e-08,
+      "loss": 1.2305,
+      "step": 30101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.069878363586746e-08,
+      "loss": 0.9004,
+      "step": 30102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.057581368877909e-08,
+      "loss": 1.0176,
+      "step": 30103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.045295039998401e-08,
+      "loss": 0.7612,
+      "step": 30104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.033019377080008e-08,
+      "loss": 1.085,
+      "step": 30105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.020754380254625e-08,
+      "loss": 0.9385,
+      "step": 30106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.008500049653922e-08,
+      "loss": 0.8594,
+      "step": 30107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.996256385409795e-08,
+      "loss": 0.9346,
+      "step": 30108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.984023387653583e-08,
+      "loss": 0.9922,
+      "step": 30109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.971801056516626e-08,
+      "loss": 0.5894,
+      "step": 30110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.959589392130261e-08,
+      "loss": 1.2314,
+      "step": 30111
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.947388394625831e-08,
+      "loss": 1.1582,
+      "step": 30112
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.935198064134119e-08,
+      "loss": 0.9746,
+      "step": 30113
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.923018400786241e-08,
+      "loss": 0.9575,
+      "step": 30114
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.910849404713094e-08,
+      "loss": 1.2461,
+      "step": 30115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.898691076045128e-08,
+      "loss": 0.8525,
+      "step": 30116
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.886543414913238e-08,
+      "loss": 0.9917,
+      "step": 30117
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.874406421447654e-08,
+      "loss": 0.9033,
+      "step": 30118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.86228009577894e-08,
+      "loss": 1.0703,
+      "step": 30119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.850164438037321e-08,
+      "loss": 1.0371,
+      "step": 30120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.838059448352807e-08,
+      "loss": 0.7417,
+      "step": 30121
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.825965126855517e-08,
+      "loss": 1.2383,
+      "step": 30122
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.813881473675344e-08,
+      "loss": 0.8438,
+      "step": 30123
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.801808488942186e-08,
+      "loss": 0.7178,
+      "step": 30124
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.789746172785605e-08,
+      "loss": 0.8687,
+      "step": 30125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.777694525335277e-08,
+      "loss": 0.6768,
+      "step": 30126
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.76565354672054e-08,
+      "loss": 0.9268,
+      "step": 30127
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.75362323707085e-08,
+      "loss": 0.6543,
+      "step": 30128
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.741603596515211e-08,
+      "loss": 0.873,
+      "step": 30129
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.729594625183078e-08,
+      "loss": 1.1602,
+      "step": 30130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.717596323203123e-08,
+      "loss": 1.0459,
+      "step": 30131
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.705608690704468e-08,
+      "loss": 1.0186,
+      "step": 30132
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.693631727815675e-08,
+      "loss": 0.8652,
+      "step": 30133
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.68166543466553e-08,
+      "loss": 0.8369,
+      "step": 30134
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.669709811382596e-08,
+      "loss": 1.4395,
+      "step": 30135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.657764858095106e-08,
+      "loss": 1.3184,
+      "step": 30136
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.645830574931622e-08,
+      "loss": 0.9873,
+      "step": 30137
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.633906962020153e-08,
+      "loss": 1.0173,
+      "step": 30138
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.621994019488931e-08,
+      "loss": 1.0176,
+      "step": 30139
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.61009174746574e-08,
+      "loss": 0.6621,
+      "step": 30140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.598200146078482e-08,
+      "loss": 1.1895,
+      "step": 30141
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.586319215454939e-08,
+      "loss": 0.9189,
+      "step": 30142
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.574448955722901e-08,
+      "loss": 0.7393,
+      "step": 30143
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.562589367009486e-08,
+      "loss": 0.9541,
+      "step": 30144
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.550740449442372e-08,
+      "loss": 1.0234,
+      "step": 30145
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.53890220314879e-08,
+      "loss": 0.9512,
+      "step": 30146
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.527074628255858e-08,
+      "loss": 0.7473,
+      "step": 30147
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.515257724890589e-08,
+      "loss": 1.0547,
+      "step": 30148
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.503451493179991e-08,
+      "loss": 1.0029,
+      "step": 30149
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.491655933250963e-08,
+      "loss": 0.9536,
+      "step": 30150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.479871045229957e-08,
+      "loss": 1.0645,
+      "step": 30151
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.468096829243654e-08,
+      "loss": 1.0469,
+      "step": 30152
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.456333285418726e-08,
+      "loss": 1.0967,
+      "step": 30153
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.444580413881407e-08,
+      "loss": 0.8945,
+      "step": 30154
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.43283821475782e-08,
+      "loss": 0.9854,
+      "step": 30155
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.421106688174194e-08,
+      "loss": 1.0469,
+      "step": 30156
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.409385834256653e-08,
+      "loss": 1.3672,
+      "step": 30157
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.397675653130875e-08,
+      "loss": 1.2334,
+      "step": 30158
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.385976144922867e-08,
+      "loss": 1.1709,
+      "step": 30159
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.374287309758088e-08,
+      "loss": 0.9502,
+      "step": 30160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.362609147762322e-08,
+      "loss": 0.918,
+      "step": 30161
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.350941659060806e-08,
+      "loss": 1.0791,
+      "step": 30162
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.339284843778993e-08,
+      "loss": 0.8594,
+      "step": 30163
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.327638702042005e-08,
+      "loss": 0.792,
+      "step": 30164
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.316003233975077e-08,
+      "loss": 1.0645,
+      "step": 30165
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.304378439703107e-08,
+      "loss": 1.2871,
+      "step": 30166
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.292764319350997e-08,
+      "loss": 0.6165,
+      "step": 30167
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.281160873043423e-08,
+      "loss": 1.0659,
+      "step": 30168
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.269568100905065e-08,
+      "loss": 0.8945,
+      "step": 30169
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.25798600306049e-08,
+      "loss": 0.6851,
+      "step": 30170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.246414579634041e-08,
+      "loss": 1.0708,
+      "step": 30171
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.234853830750065e-08,
+      "loss": 0.9746,
+      "step": 30172
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.223303756532684e-08,
+      "loss": 0.9414,
+      "step": 30173
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.211764357106132e-08,
+      "loss": 1.1816,
+      "step": 30174
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.200235632594198e-08,
+      "loss": 0.7214,
+      "step": 30175
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.188717583120674e-08,
+      "loss": 0.9082,
+      "step": 30176
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.177210208809458e-08,
+      "loss": 0.8164,
+      "step": 30177
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.165713509784011e-08,
+      "loss": 0.9375,
+      "step": 30178
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.154227486168008e-08,
+      "loss": 1.0615,
+      "step": 30179
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.142752138084573e-08,
+      "loss": 0.9082,
+      "step": 30180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.131287465657165e-08,
+      "loss": 1.1816,
+      "step": 30181
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.119833469008907e-08,
+      "loss": 1.0122,
+      "step": 30182
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.108390148262699e-08,
+      "loss": 1.0645,
+      "step": 30183
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.096957503541556e-08,
+      "loss": 1.1777,
+      "step": 30184
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.085535534968379e-08,
+      "loss": 0.7568,
+      "step": 30185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.074124242665736e-08,
+      "loss": 1.0537,
+      "step": 30186
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.062723626756195e-08,
+      "loss": 0.8354,
+      "step": 30187
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.051333687362327e-08,
+      "loss": 0.5522,
+      "step": 30188
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.039954424606365e-08,
+      "loss": 0.7798,
+      "step": 30189
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.028585838610656e-08,
+      "loss": 0.7129,
+      "step": 30190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.017227929497104e-08,
+      "loss": 0.9658,
+      "step": 30191
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.005880697387945e-08,
+      "loss": 1.0957,
+      "step": 30192
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.994544142404968e-08,
+      "loss": 0.7749,
+      "step": 30193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.983218264669965e-08,
+      "loss": 1.209,
+      "step": 30194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.971903064304507e-08,
+      "loss": 0.6855,
+      "step": 30195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.960598541430274e-08,
+      "loss": 0.9927,
+      "step": 30196
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.949304696168612e-08,
+      "loss": 1.2012,
+      "step": 30197
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.938021528640869e-08,
+      "loss": 0.9131,
+      "step": 30198
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.926749038968172e-08,
+      "loss": 1.1904,
+      "step": 30199
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9154872272717545e-08,
+      "loss": 1.1875,
+      "step": 30200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9042360936725216e-08,
+      "loss": 1.2197,
+      "step": 30201
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8929956382911545e-08,
+      "loss": 0.7314,
+      "step": 30202
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.881765861248667e-08,
+      "loss": 1.0674,
+      "step": 30203
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8705467626655186e-08,
+      "loss": 1.1221,
+      "step": 30204
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8593383426622795e-08,
+      "loss": 1.1777,
+      "step": 30205
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.848140601359409e-08,
+      "loss": 1.0713,
+      "step": 30206
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.836953538877033e-08,
+      "loss": 1.3574,
+      "step": 30207
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.825777155335388e-08,
+      "loss": 1.0801,
+      "step": 30208
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8146114508546015e-08,
+      "loss": 1.1162,
+      "step": 30209
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8034564255544655e-08,
+      "loss": 1.209,
+      "step": 30210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.792312079554996e-08,
+      "loss": 1.0127,
+      "step": 30211
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.781178412975763e-08,
+      "loss": 0.8052,
+      "step": 30212
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7700554259362274e-08,
+      "loss": 1.053,
+      "step": 30213
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.758943118556182e-08,
+      "loss": 0.9398,
+      "step": 30214
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7478414909548644e-08,
+      "loss": 0.9263,
+      "step": 30215
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7367505432514014e-08,
+      "loss": 0.6538,
+      "step": 30216
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.725670275565032e-08,
+      "loss": 0.6387,
+      "step": 30217
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.714600688014882e-08,
+      "loss": 1.0166,
+      "step": 30218
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7035417807196344e-08,
+      "loss": 0.6699,
+      "step": 30219
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.692493553798306e-08,
+      "loss": 0.9395,
+      "step": 30220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.681456007369357e-08,
+      "loss": 1.0078,
+      "step": 30221
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.67042914155147e-08,
+      "loss": 0.7754,
+      "step": 30222
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.659412956463106e-08,
+      "loss": 1.0098,
+      "step": 30223
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.648407452222504e-08,
+      "loss": 0.8853,
+      "step": 30224
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.637412628947903e-08,
+      "loss": 0.8291,
+      "step": 30225
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.62642848675754e-08,
+      "loss": 0.9292,
+      "step": 30226
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.615455025769323e-08,
+      "loss": 0.9219,
+      "step": 30227
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6044922461009346e-08,
+      "loss": 1.2109,
+      "step": 30228
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.593540147870502e-08,
+      "loss": 1.2617,
+      "step": 30229
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.582598731195377e-08,
+      "loss": 1.3838,
+      "step": 30230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.571667996193131e-08,
+      "loss": 0.9316,
+      "step": 30231
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.560747942981337e-08,
+      "loss": 0.8506,
+      "step": 30232
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.549838571677124e-08,
+      "loss": 1.1475,
+      "step": 30233
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5389398823978426e-08,
+      "loss": 0.8574,
+      "step": 30234
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.528051875260287e-08,
+      "loss": 0.8833,
+      "step": 30235
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5171745503816984e-08,
+      "loss": 0.79,
+      "step": 30236
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.506307907878761e-08,
+      "loss": 0.7183,
+      "step": 30237
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.495451947868269e-08,
+      "loss": 0.7715,
+      "step": 30238
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4846066704667965e-08,
+      "loss": 1.1836,
+      "step": 30239
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.473772075790806e-08,
+      "loss": 0.8984,
+      "step": 30240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.46294816395676e-08,
+      "loss": 0.6655,
+      "step": 30241
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4521349350808996e-08,
+      "loss": 1.2295,
+      "step": 30242
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4413323892793524e-08,
+      "loss": 1.0469,
+      "step": 30243
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.430540526668138e-08,
+      "loss": 0.9258,
+      "step": 30244
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.419759347363274e-08,
+      "loss": 1.1719,
+      "step": 30245
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.408988851480446e-08,
+      "loss": 1.1211,
+      "step": 30246
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.39822903913545e-08,
+      "loss": 1.0117,
+      "step": 30247
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.387479910443749e-08,
+      "loss": 1.1455,
+      "step": 30248
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.376741465521029e-08,
+      "loss": 1.0342,
+      "step": 30249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.366013704482309e-08,
+      "loss": 0.6196,
+      "step": 30250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.355296627443052e-08,
+      "loss": 0.7578,
+      "step": 30251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3445902345183875e-08,
+      "loss": 1.1836,
+      "step": 30252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3338945258232246e-08,
+      "loss": 1.2871,
+      "step": 30253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3232095014723594e-08,
+      "loss": 1.0742,
+      "step": 30254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.312535161580812e-08,
+      "loss": 1.0859,
+      "step": 30255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3018715062630455e-08,
+      "loss": 1.0557,
+      "step": 30256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2912185356337464e-08,
+      "loss": 0.6294,
+      "step": 30257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.280576249807268e-08,
+      "loss": 0.8833,
+      "step": 30258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2699446488979624e-08,
+      "loss": 0.8022,
+      "step": 30259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.259323733019961e-08,
+      "loss": 0.8862,
+      "step": 30260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2487135022873945e-08,
+      "loss": 1.1289,
+      "step": 30261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2381139568142835e-08,
+      "loss": 0.8037,
+      "step": 30262
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2275250967144254e-08,
+      "loss": 1.0464,
+      "step": 30263
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2169469221015066e-08,
+      "loss": 1.1572,
+      "step": 30264
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2063794330892146e-08,
+      "loss": 0.96,
+      "step": 30265
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.195822629791125e-08,
+      "loss": 0.6929,
+      "step": 30266
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1852765123205914e-08,
+      "loss": 1.1172,
+      "step": 30267
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.174741080790746e-08,
+      "loss": 1.1089,
+      "step": 30268
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.164216335315053e-08,
+      "loss": 0.835,
+      "step": 30269
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.153702276006312e-08,
+      "loss": 1.0322,
+      "step": 30270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.143198902977542e-08,
+      "loss": 1.0508,
+      "step": 30271
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.132706216341654e-08,
+      "loss": 1.6504,
+      "step": 30272
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1222242162112244e-08,
+      "loss": 0.7822,
+      "step": 30273
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.111752902698941e-08,
+      "loss": 0.7354,
+      "step": 30274
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.101292275917158e-08,
+      "loss": 1.2119,
+      "step": 30275
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.090842335978341e-08,
+      "loss": 1.002,
+      "step": 30276
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.080403082994845e-08,
+      "loss": 0.6094,
+      "step": 30277
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0699745170785796e-08,
+      "loss": 0.9219,
+      "step": 30278
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.059556638341678e-08,
+      "loss": 0.645,
+      "step": 30279
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.049149446896051e-08,
+      "loss": 0.8711,
+      "step": 30280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.038752942853498e-08,
+      "loss": 0.9922,
+      "step": 30281
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0283671263255954e-08,
+      "loss": 1.2485,
+      "step": 30282
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.017991997424032e-08,
+      "loss": 1.2363,
+      "step": 30283
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.007627556260275e-08,
+      "loss": 1.002,
+      "step": 30284
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.997273802945457e-08,
+      "loss": 0.8643,
+      "step": 30285
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9869307375909334e-08,
+      "loss": 0.9629,
+      "step": 30286
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.976598360307838e-08,
+      "loss": 1.0303,
+      "step": 30287
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.966276671207082e-08,
+      "loss": 0.8091,
+      "step": 30288
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9559656703996874e-08,
+      "loss": 1.0767,
+      "step": 30289
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.945665357996121e-08,
+      "loss": 1.0176,
+      "step": 30290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.935375734107295e-08,
+      "loss": 0.8936,
+      "step": 30291
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.925096798843565e-08,
+      "loss": 1.1865,
+      "step": 30292
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.914828552315509e-08,
+      "loss": 1.4316,
+      "step": 30293
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.904570994633151e-08,
+      "loss": 0.9736,
+      "step": 30294
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.894324125907068e-08,
+      "loss": 0.7705,
+      "step": 30295
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.884087946246951e-08,
+      "loss": 0.9463,
+      "step": 30296
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.873862455762934e-08,
+      "loss": 0.7988,
+      "step": 30297
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.863647654564929e-08,
+      "loss": 0.582,
+      "step": 30298
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8534435427625145e-08,
+      "loss": 0.8799,
+      "step": 30299
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.843250120465381e-08,
+      "loss": 1.1812,
+      "step": 30300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.833067387782886e-08,
+      "loss": 1.0225,
+      "step": 30301
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.822895344824607e-08,
+      "loss": 1.0801,
+      "step": 30302
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.812733991699792e-08,
+      "loss": 0.8398,
+      "step": 30303
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8025833285174625e-08,
+      "loss": 0.8608,
+      "step": 30304
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.792443355386644e-08,
+      "loss": 1.293,
+      "step": 30305
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.782314072416361e-08,
+      "loss": 0.7793,
+      "step": 30306
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7721954797153024e-08,
+      "loss": 0.9043,
+      "step": 30307
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.762087577392382e-08,
+      "loss": 1.124,
+      "step": 30308
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.751990365555847e-08,
+      "loss": 0.8042,
+      "step": 30309
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7419038443144995e-08,
+      "loss": 0.6162,
+      "step": 30310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.731828013776474e-08,
+      "loss": 1.0293,
+      "step": 30311
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.721762874050018e-08,
+      "loss": 1.292,
+      "step": 30312
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7117084252432664e-08,
+      "loss": 0.8672,
+      "step": 30313
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.701664667464245e-08,
+      "loss": 1.1733,
+      "step": 30314
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.691631600820867e-08,
+      "loss": 0.7069,
+      "step": 30315
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.681609225420824e-08,
+      "loss": 0.8369,
+      "step": 30316
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6715975413719196e-08,
+      "loss": 0.417,
+      "step": 30317
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6615965487815106e-08,
+      "loss": 0.8486,
+      "step": 30318
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.651606247757068e-08,
+      "loss": 0.8066,
+      "step": 30319
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.641626638405949e-08,
+      "loss": 0.8799,
+      "step": 30320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.631657720835403e-08,
+      "loss": 0.6772,
+      "step": 30321
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.621699495152454e-08,
+      "loss": 1.0352,
+      "step": 30322
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.611751961464017e-08,
+      "loss": 1.0054,
+      "step": 30323
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6018151198770066e-08,
+      "loss": 0.9766,
+      "step": 30324
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.591888970498226e-08,
+      "loss": 0.9629,
+      "step": 30325
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.581973513434146e-08,
+      "loss": 1.2617,
+      "step": 30326
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5720687487913476e-08,
+      "loss": 1.0957,
+      "step": 30327
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.56217467667619e-08,
+      "loss": 0.875,
+      "step": 30328
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5522912971951436e-08,
+      "loss": 1.1582,
+      "step": 30329
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5424186104541246e-08,
+      "loss": 0.7285,
+      "step": 30330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.53255661655927e-08,
+      "loss": 0.8052,
+      "step": 30331
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5227053156164936e-08,
+      "loss": 1.0039,
+      "step": 30332
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.512864707731601e-08,
+      "loss": 0.9785,
+      "step": 30333
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5030347930105077e-08,
+      "loss": 1.0762,
+      "step": 30334
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4932155715584625e-08,
+      "loss": 0.5298,
+      "step": 30335
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4834070434811584e-08,
+      "loss": 0.7456,
+      "step": 30336
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4736092088839556e-08,
+      "loss": 1.0566,
+      "step": 30337
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.463822067871993e-08,
+      "loss": 1.0244,
+      "step": 30338
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.454045620550407e-08,
+      "loss": 0.998,
+      "step": 30339
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.444279867024448e-08,
+      "loss": 1.0996,
+      "step": 30340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.434524807398699e-08,
+      "loss": 0.6382,
+      "step": 30341
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.424780441778076e-08,
+      "loss": 1.168,
+      "step": 30342
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4150467702671615e-08,
+      "loss": 0.8799,
+      "step": 30343
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4053237929707615e-08,
+      "loss": 0.9111,
+      "step": 30344
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3956115099931253e-08,
+      "loss": 1.2812,
+      "step": 30345
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.385909921438502e-08,
+      "loss": 0.9863,
+      "step": 30346
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.376219027411366e-08,
+      "loss": 1.0186,
+      "step": 30347
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3665388280155206e-08,
+      "loss": 0.9204,
+      "step": 30348
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.356869323355106e-08,
+      "loss": 0.8755,
+      "step": 30349
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.347210513534039e-08,
+      "loss": 0.9512,
+      "step": 30350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.337562398655903e-08,
+      "loss": 0.8037,
+      "step": 30351
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3279249788245046e-08,
+      "loss": 0.9321,
+      "step": 30352
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.318298254143205e-08,
+      "loss": 0.8447,
+      "step": 30353
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.308682224715588e-08,
+      "loss": 0.9189,
+      "step": 30354
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.299076890644793e-08,
+      "loss": 0.833,
+      "step": 30355
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2894822520340715e-08,
+      "loss": 0.9941,
+      "step": 30356
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.279898308986452e-08,
+      "loss": 0.894,
+      "step": 30357
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.270325061604963e-08,
+      "loss": 1.3691,
+      "step": 30358
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2607625099923e-08,
+      "loss": 0.7693,
+      "step": 30359
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2512106542512696e-08,
+      "loss": 0.7305,
+      "step": 30360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.241669494484568e-08,
+      "loss": 1.5254,
+      "step": 30361
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2321390307944464e-08,
+      "loss": 1.2676,
+      "step": 30362
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.22261926328349e-08,
+      "loss": 0.7617,
+      "step": 30363
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.213110192053838e-08,
+      "loss": 0.8599,
+      "step": 30364
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.203611817207631e-08,
+      "loss": 0.832,
+      "step": 30365
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1941241388470114e-08,
+      "loss": 1.1367,
+      "step": 30366
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.184647157073673e-08,
+      "loss": 1.2256,
+      "step": 30367
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.175180871989759e-08,
+      "loss": 0.7881,
+      "step": 30368
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.16572528369652e-08,
+      "loss": 0.8149,
+      "step": 30369
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.156280392295986e-08,
+      "loss": 0.8945,
+      "step": 30370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.146846197889187e-08,
+      "loss": 0.9404,
+      "step": 30371
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.13742270057782e-08,
+      "loss": 0.8779,
+      "step": 30372
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.128009900462804e-08,
+      "loss": 0.8584,
+      "step": 30373
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.118607797645502e-08,
+      "loss": 0.8818,
+      "step": 30374
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.109216392226611e-08,
+      "loss": 0.8955,
+      "step": 30375
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.099835684307385e-08,
+      "loss": 1.1064,
+      "step": 30376
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.090465673988297e-08,
+      "loss": 1.5059,
+      "step": 30377
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.081106361370157e-08,
+      "loss": 1.3066,
+      "step": 30378
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.071757746553329e-08,
+      "loss": 0.7495,
+      "step": 30379
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0624198296385084e-08,
+      "loss": 0.8145,
+      "step": 30380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0530926107257284e-08,
+      "loss": 1.0654,
+      "step": 30381
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.043776089915241e-08,
+      "loss": 0.9805,
+      "step": 30382
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.034470267307189e-08,
+      "loss": 0.7432,
+      "step": 30383
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.025175143001603e-08,
+      "loss": 1.167,
+      "step": 30384
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.01589071709807e-08,
+      "loss": 0.998,
+      "step": 30385
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.006616989696621e-08,
+      "loss": 1.1025,
+      "step": 30386
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.997353960896622e-08,
+      "loss": 0.9824,
+      "step": 30387
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9881016307977696e-08,
+      "loss": 0.8306,
+      "step": 30388
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.978859999499318e-08,
+      "loss": 0.7744,
+      "step": 30389
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.969629067100522e-08,
+      "loss": 1.0205,
+      "step": 30390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9604088337006355e-08,
+      "loss": 0.8965,
+      "step": 30391
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9511992993988e-08,
+      "loss": 0.8408,
+      "step": 30392
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.942000464293605e-08,
+      "loss": 0.668,
+      "step": 30393
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.932812328484081e-08,
+      "loss": 0.917,
+      "step": 30394
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9236348920689284e-08,
+      "loss": 0.999,
+      "step": 30395
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.914468155146734e-08,
+      "loss": 0.9131,
+      "step": 30396
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9053121178158625e-08,
+      "loss": 0.8398,
+      "step": 30397
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8961667801747924e-08,
+      "loss": 0.752,
+      "step": 30398
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8870321423217785e-08,
+      "loss": 1.1553,
+      "step": 30399
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.877908204354741e-08,
+      "loss": 0.6768,
+      "step": 30400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.868794966371714e-08,
+      "loss": 0.791,
+      "step": 30401
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.85969242847084e-08,
+      "loss": 0.5854,
+      "step": 30402
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.850600590749709e-08,
+      "loss": 0.9756,
+      "step": 30403
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.841519453305909e-08,
+      "loss": 0.451,
+      "step": 30404
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.832449016237139e-08,
+      "loss": 0.998,
+      "step": 30405
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8233892796407656e-08,
+      "loss": 0.8711,
+      "step": 30406
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.814340243614156e-08,
+      "loss": 1.0391,
+      "step": 30407
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.805301908254455e-08,
+      "loss": 0.9258,
+      "step": 30408
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.796274273658807e-08,
+      "loss": 0.9395,
+      "step": 30409
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.787257339924133e-08,
+      "loss": 0.8779,
+      "step": 30410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.778251107147246e-08,
+      "loss": 0.8867,
+      "step": 30411
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.769255575424846e-08,
+      "loss": 0.9873,
+      "step": 30412
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7602707448538553e-08,
+      "loss": 0.6777,
+      "step": 30413
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7512966155305306e-08,
+      "loss": 1.3164,
+      "step": 30414
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.742333187551239e-08,
+      "loss": 1.0205,
+      "step": 30415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.733380461012348e-08,
+      "loss": 0.9404,
+      "step": 30416
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.724438436010114e-08,
+      "loss": 1.1387,
+      "step": 30417
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7155071126404596e-08,
+      "loss": 1.209,
+      "step": 30418
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.706586490999309e-08,
+      "loss": 0.938,
+      "step": 30419
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6976765711825844e-08,
+      "loss": 1.1797,
+      "step": 30420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.688777353285988e-08,
+      "loss": 1.1064,
+      "step": 30421
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.679888837404999e-08,
+      "loss": 0.9346,
+      "step": 30422
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.671011023635207e-08,
+      "loss": 0.9243,
+      "step": 30423
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.662143912071869e-08,
+      "loss": 0.8662,
+      "step": 30424
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6532875028103544e-08,
+      "loss": 1.0225,
+      "step": 30425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6444417959456966e-08,
+      "loss": 0.9727,
+      "step": 30426
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.635606791572932e-08,
+      "loss": 0.8965,
+      "step": 30427
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.626782489787095e-08,
+      "loss": 0.7617,
+      "step": 30428
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.617968890682777e-08,
+      "loss": 1.0137,
+      "step": 30429
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6091659943546797e-08,
+      "loss": 0.7432,
+      "step": 30430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6003738008975054e-08,
+      "loss": 0.9429,
+      "step": 30431
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5915923104056225e-08,
+      "loss": 0.7393,
+      "step": 30432
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.58282152297329e-08,
+      "loss": 1.25,
+      "step": 30433
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5740614386947644e-08,
+      "loss": 0.8359,
+      "step": 30434
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.565312057664194e-08,
+      "loss": 0.9795,
+      "step": 30435
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5565733799753924e-08,
+      "loss": 0.877,
+      "step": 30436
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.547845405722505e-08,
+      "loss": 0.9551,
+      "step": 30437
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.539128134999126e-08,
+      "loss": 0.8291,
+      "step": 30438
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.530421567898845e-08,
+      "loss": 1.1523,
+      "step": 30439
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.521725704515255e-08,
+      "loss": 1.1572,
+      "step": 30440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5130405449417257e-08,
+      "loss": 1.2974,
+      "step": 30441
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.504366089271627e-08,
+      "loss": 1.3555,
+      "step": 30442
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.495702337597995e-08,
+      "loss": 0.8555,
+      "step": 30443
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4870492900140886e-08,
+      "loss": 0.9678,
+      "step": 30444
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.478406946612722e-08,
+      "loss": 1.1323,
+      "step": 30445
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.469775307486711e-08,
+      "loss": 0.6582,
+      "step": 30446
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4611543727288696e-08,
+      "loss": 1.0967,
+      "step": 30447
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.45254414243168e-08,
+      "loss": 1.0938,
+      "step": 30448
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4439446166877334e-08,
+      "loss": 1.2168,
+      "step": 30449
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.435355795589401e-08,
+      "loss": 0.8066,
+      "step": 30450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.426777679228832e-08,
+      "loss": 1.0801,
+      "step": 30451
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.418210267698285e-08,
+      "loss": 0.5576,
+      "step": 30452
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.409653561089799e-08,
+      "loss": 0.7915,
+      "step": 30453
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.401107559495187e-08,
+      "loss": 0.7227,
+      "step": 30454
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3925722630062665e-08,
+      "loss": 0.6416,
+      "step": 30455
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.384047671714852e-08,
+      "loss": 1.0664,
+      "step": 30456
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3755337857123146e-08,
+      "loss": 1.1943,
+      "step": 30457
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.367030605090249e-08,
+      "loss": 1.1748,
+      "step": 30458
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3585381299399146e-08,
+      "loss": 0.96,
+      "step": 30459
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.350056360352572e-08,
+      "loss": 0.8438,
+      "step": 30460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3415852964193696e-08,
+      "loss": 0.9004,
+      "step": 30461
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3331249382312357e-08,
+      "loss": 1.2793,
+      "step": 30462
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3246752858789864e-08,
+      "loss": 0.8389,
+      "step": 30463
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.316236339453549e-08,
+      "loss": 0.7817,
+      "step": 30464
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.307808099045517e-08,
+      "loss": 1.1064,
+      "step": 30465
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.299390564745264e-08,
+      "loss": 0.9707,
+      "step": 30466
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2909837366434936e-08,
+      "loss": 1.0127,
+      "step": 30467
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.282587614830246e-08,
+      "loss": 1.2031,
+      "step": 30468
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2742021993958925e-08,
+      "loss": 0.9941,
+      "step": 30469
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.265827490430362e-08,
+      "loss": 1.0366,
+      "step": 30470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.257463488023804e-08,
+      "loss": 1.1943,
+      "step": 30471
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.249110192265814e-08,
+      "loss": 1.1689,
+      "step": 30472
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.24076760324632e-08,
+      "loss": 1.4541,
+      "step": 30473
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2324357210548055e-08,
+      "loss": 1.1309,
+      "step": 30474
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2241145457808656e-08,
+      "loss": 0.9512,
+      "step": 30475
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.215804077513762e-08,
+      "loss": 1.2339,
+      "step": 30476
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.20750431634298e-08,
+      "loss": 0.8594,
+      "step": 30477
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.199215262357447e-08,
+      "loss": 0.8369,
+      "step": 30478
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1909369156463145e-08,
+      "loss": 0.7812,
+      "step": 30479
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1826692762984004e-08,
+      "loss": 1.0088,
+      "step": 30480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.174412344402744e-08,
+      "loss": 1.293,
+      "step": 30481
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.166166120047831e-08,
+      "loss": 1.1582,
+      "step": 30482
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.157930603322146e-08,
+      "loss": 1.0879,
+      "step": 30483
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.149705794314506e-08,
+      "loss": 1.1172,
+      "step": 30484
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.141491693112952e-08,
+      "loss": 1.2744,
+      "step": 30485
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1332882998058586e-08,
+      "loss": 0.8926,
+      "step": 30486
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.125095614481266e-08,
+      "loss": 0.9785,
+      "step": 30487
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.116913637227215e-08,
+      "loss": 0.873,
+      "step": 30488
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.108742368131634e-08,
+      "loss": 0.999,
+      "step": 30489
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.100581807282232e-08,
+      "loss": 0.8584,
+      "step": 30490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.092431954766717e-08,
+      "loss": 0.9727,
+      "step": 30491
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.084292810672573e-08,
+      "loss": 0.7222,
+      "step": 30492
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.076164375087287e-08,
+      "loss": 1.085,
+      "step": 30493
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.068046648098122e-08,
+      "loss": 1.5664,
+      "step": 30494
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.059939629792341e-08,
+      "loss": 0.9883,
+      "step": 30495
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0518433202568755e-08,
+      "loss": 1.1768,
+      "step": 30496
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.043757719578988e-08,
+      "loss": 1.3047,
+      "step": 30497
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0356828278451656e-08,
+      "loss": 0.9453,
+      "step": 30498
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0276186451424494e-08,
+      "loss": 0.8125,
+      "step": 30499
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.019565171557215e-08,
+      "loss": 0.7793,
+      "step": 30500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0115224071761704e-08,
+      "loss": 1.04,
+      "step": 30501
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.003490352085692e-08,
+      "loss": 0.6792,
+      "step": 30502
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9954690063718205e-08,
+      "loss": 0.7998,
+      "step": 30503
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.987458370121043e-08,
+      "loss": 0.874,
+      "step": 30504
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9794584434191808e-08,
+      "loss": 0.9922,
+      "step": 30505
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.971469226352275e-08,
+      "loss": 0.8906,
+      "step": 30506
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.963490719006146e-08,
+      "loss": 1.3125,
+      "step": 30507
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9555229214663916e-08,
+      "loss": 0.9893,
+      "step": 30508
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9475658338187217e-08,
+      "loss": 1.1836,
+      "step": 30509
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9396194561485124e-08,
+      "loss": 0.6272,
+      "step": 30510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.93168378854114e-08,
+      "loss": 0.9736,
+      "step": 30511
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9237588310819797e-08,
+      "loss": 1.0791,
+      "step": 30512
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9158445838559645e-08,
+      "loss": 1.3301,
+      "step": 30513
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9079410469482484e-08,
+      "loss": 1.0723,
+      "step": 30514
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.900048220443541e-08,
+      "loss": 0.873,
+      "step": 30515
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8921661044268857e-08,
+      "loss": 1.0596,
+      "step": 30516
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8842946989827702e-08,
+      "loss": 0.824,
+      "step": 30517
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.876434004195794e-08,
+      "loss": 0.6914,
+      "step": 30518
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8685840201503333e-08,
+      "loss": 1.041,
+      "step": 30519
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.860744746930877e-08,
+      "loss": 0.8691,
+      "step": 30520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8529161846214682e-08,
+      "loss": 1.2119,
+      "step": 30521
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.845098333306262e-08,
+      "loss": 0.8608,
+      "step": 30522
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8372911930693025e-08,
+      "loss": 0.8887,
+      "step": 30523
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8294947639943006e-08,
+      "loss": 0.9453,
+      "step": 30524
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8217090461650776e-08,
+      "loss": 1.0723,
+      "step": 30525
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8139340396652338e-08,
+      "loss": 0.894,
+      "step": 30526
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8061697445783685e-08,
+      "loss": 1.1025,
+      "step": 30527
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.79841616098786e-08,
+      "loss": 1.3525,
+      "step": 30528
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7906732889768638e-08,
+      "loss": 1.5586,
+      "step": 30529
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7829411286287577e-08,
+      "loss": 1.1641,
+      "step": 30530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.775219680026364e-08,
+      "loss": 0.9609,
+      "step": 30531
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.76750894325295e-08,
+      "loss": 0.9038,
+      "step": 30532
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.759808918391005e-08,
+      "loss": 1.0156,
+      "step": 30533
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.752119605523351e-08,
+      "loss": 1.0234,
+      "step": 30534
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.744441004732701e-08,
+      "loss": 0.8076,
+      "step": 30535
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.736773116101432e-08,
+      "loss": 0.8916,
+      "step": 30536
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7291159397119238e-08,
+      "loss": 0.7817,
+      "step": 30537
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7214694756464433e-08,
+      "loss": 0.8701,
+      "step": 30538
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.713833723987036e-08,
+      "loss": 1.0977,
+      "step": 30539
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7062086848158587e-08,
+      "loss": 0.9844,
+      "step": 30540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6985943582147344e-08,
+      "loss": 0.532,
+      "step": 30541
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.690990744265487e-08,
+      "loss": 0.8887,
+      "step": 30542
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6833978430498285e-08,
+      "loss": 1.0952,
+      "step": 30543
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6758156546491388e-08,
+      "loss": 0.8752,
+      "step": 30544
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6682441791450187e-08,
+      "loss": 1.3154,
+      "step": 30545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6606834166188478e-08,
+      "loss": 1.3047,
+      "step": 30546
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6531333671517832e-08,
+      "loss": 0.7998,
+      "step": 30547
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.645594030824872e-08,
+      "loss": 1.0635,
+      "step": 30548
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6380654077191593e-08,
+      "loss": 1.0527,
+      "step": 30549
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.63054749791547e-08,
+      "loss": 0.585,
+      "step": 30550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6230403014945172e-08,
+      "loss": 0.9243,
+      "step": 30551
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.615543818537125e-08,
+      "loss": 0.7495,
+      "step": 30552
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.608058049123674e-08,
+      "loss": 0.7419,
+      "step": 30553
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.600582993334544e-08,
+      "loss": 0.4575,
+      "step": 30554
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.593118651250004e-08,
+      "loss": 0.9932,
+      "step": 30555
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5856650229504343e-08,
+      "loss": 1.3047,
+      "step": 30556
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.578222108515771e-08,
+      "loss": 0.9385,
+      "step": 30557
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.57078990802595e-08,
+      "loss": 0.9355,
+      "step": 30558
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5633684215607967e-08,
+      "loss": 0.8516,
+      "step": 30559
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5559576492000248e-08,
+      "loss": 0.8789,
+      "step": 30560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5485575910233484e-08,
+      "loss": 1.1826,
+      "step": 30561
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5411682471101485e-08,
+      "loss": 0.9795,
+      "step": 30562
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5337896175399167e-08,
+      "loss": 0.9492,
+      "step": 30563
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5264217023917014e-08,
+      "loss": 1.2666,
+      "step": 30564
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5190645017447724e-08,
+      "loss": 0.5835,
+      "step": 30565
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5117180156781772e-08,
+      "loss": 0.7399,
+      "step": 30566
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.504382244270853e-08,
+      "loss": 1.1406,
+      "step": 30567
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4970571876015147e-08,
+      "loss": 1.0312,
+      "step": 30568
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4897428457487662e-08,
+      "loss": 1.0315,
+      "step": 30569
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4824392187914326e-08,
+      "loss": 0.749,
+      "step": 30570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.475146306807785e-08,
+      "loss": 1.1211,
+      "step": 30571
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4678641098762057e-08,
+      "loss": 0.7803,
+      "step": 30572
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4605926280749648e-08,
+      "loss": 0.832,
+      "step": 30573
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.453331861482e-08,
+      "loss": 1.042,
+      "step": 30574
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4460818101754714e-08,
+      "loss": 0.8398,
+      "step": 30575
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.438842474233205e-08,
+      "loss": 1.0186,
+      "step": 30576
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.431613853733028e-08,
+      "loss": 0.7676,
+      "step": 30577
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4243959487524337e-08,
+      "loss": 1.4805,
+      "step": 30578
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4171887593691378e-08,
+      "loss": 0.7715,
+      "step": 30579
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4099922856604117e-08,
+      "loss": 1.0234,
+      "step": 30580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4028065277035274e-08,
+      "loss": 0.8828,
+      "step": 30581
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3956314855758667e-08,
+      "loss": 0.9443,
+      "step": 30582
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3884671593543683e-08,
+      "loss": 0.8872,
+      "step": 30583
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.381313549115971e-08,
+      "loss": 0.813,
+      "step": 30584
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.374170654937613e-08,
+      "loss": 1.0078,
+      "step": 30585
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3670384768959e-08,
+      "loss": 0.8281,
+      "step": 30586
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3599170150675478e-08,
+      "loss": 0.9053,
+      "step": 30587
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.35280626952894e-08,
+      "loss": 1.0312,
+      "step": 30588
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3457062403566822e-08,
+      "loss": 0.9863,
+      "step": 30589
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3386169276267134e-08,
+      "loss": 0.7432,
+      "step": 30590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.331538331415417e-08,
+      "loss": 0.7305,
+      "step": 30591
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3244704517988436e-08,
+      "loss": 1.2783,
+      "step": 30592
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.317413288852821e-08,
+      "loss": 0.8525,
+      "step": 30593
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3103668426530667e-08,
+      "loss": 1.207,
+      "step": 30594
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.303331113275409e-08,
+      "loss": 1.2246,
+      "step": 30595
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2963061007954534e-08,
+      "loss": 1.1768,
+      "step": 30596
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2892918052885848e-08,
+      "loss": 1.2178,
+      "step": 30597
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2822882268301872e-08,
+      "loss": 0.6943,
+      "step": 30598
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2752953654953112e-08,
+      "loss": 1.1143,
+      "step": 30599
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2683132213594526e-08,
+      "loss": 1.0654,
+      "step": 30600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2613417944972183e-08,
+      "loss": 0.9819,
+      "step": 30601
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.254381084983659e-08,
+      "loss": 1.2578,
+      "step": 30602
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.247431092893604e-08,
+      "loss": 0.9199,
+      "step": 30603
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.24049181830166e-08,
+      "loss": 1.2256,
+      "step": 30604
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2335632612823233e-08,
+      "loss": 0.8567,
+      "step": 30605
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2266454219100896e-08,
+      "loss": 0.6953,
+      "step": 30606
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2197383002591223e-08,
+      "loss": 0.8818,
+      "step": 30607
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2128418964038057e-08,
+      "loss": 1.0117,
+      "step": 30608
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.205956210418081e-08,
+      "loss": 1.0029,
+      "step": 30609
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.199081242376e-08,
+      "loss": 1.0181,
+      "step": 30610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1922169923513925e-08,
+      "loss": 1.0762,
+      "step": 30611
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1853634604178663e-08,
+      "loss": 1.1758,
+      "step": 30612
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1785206466492516e-08,
+      "loss": 1.3105,
+      "step": 30613
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1716885511189335e-08,
+      "loss": 0.9385,
+      "step": 30614
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.164867173900298e-08,
+      "loss": 1.127,
+      "step": 30615
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.15805651506662e-08,
+      "loss": 1.1377,
+      "step": 30616
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.151256574691063e-08,
+      "loss": 0.9072,
+      "step": 30617
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1444673528466797e-08,
+      "loss": 0.6304,
+      "step": 30618
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.137688849606412e-08,
+      "loss": 0.7822,
+      "step": 30619
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.13092106504309e-08,
+      "loss": 0.9932,
+      "step": 30620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1241639992293228e-08,
+      "loss": 1.083,
+      "step": 30621
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1174176522377188e-08,
+      "loss": 1.3359,
+      "step": 30622
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1106820241407755e-08,
+      "loss": 1.04,
+      "step": 30623
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1039571150108794e-08,
+      "loss": 1.0381,
+      "step": 30624
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0972429249201952e-08,
+      "loss": 0.9326,
+      "step": 30625
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0905394539409984e-08,
+      "loss": 1.2607,
+      "step": 30626
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.083846702145009e-08,
+      "loss": 0.959,
+      "step": 30627
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0771646696042815e-08,
+      "loss": 0.9883,
+      "step": 30628
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0704933563906458e-08,
+      "loss": 0.959,
+      "step": 30629
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0638327625758237e-08,
+      "loss": 1.1699,
+      "step": 30630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0571828882310905e-08,
+      "loss": 0.7944,
+      "step": 30631
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0505437334280564e-08,
+      "loss": 1.0674,
+      "step": 30632
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0439152982379972e-08,
+      "loss": 1.1348,
+      "step": 30633
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0372975827321894e-08,
+      "loss": 1.2441,
+      "step": 30634
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0306905869816874e-08,
+      "loss": 0.9668,
+      "step": 30635
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.024094311057323e-08,
+      "loss": 0.7568,
+      "step": 30636
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0175087550301508e-08,
+      "loss": 0.9697,
+      "step": 30637
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.010933918970781e-08,
+      "loss": 1.0117,
+      "step": 30638
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.004369802949935e-08,
+      "loss": 0.665,
+      "step": 30639
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9978164070380002e-08,
+      "loss": 1.0947,
+      "step": 30640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9912737313054765e-08,
+      "loss": 0.9409,
+      "step": 30641
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9847417758225294e-08,
+      "loss": 0.8877,
+      "step": 30642
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9782205406594367e-08,
+      "loss": 0.8857,
+      "step": 30643
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.971710025886142e-08,
+      "loss": 1.0732,
+      "step": 30644
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9652102315727007e-08,
+      "loss": 0.9756,
+      "step": 30645
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9587211577888344e-08,
+      "loss": 0.9639,
+      "step": 30646
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9522428046043763e-08,
+      "loss": 1.082,
+      "step": 30647
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9457751720887154e-08,
+      "loss": 1.166,
+      "step": 30648
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9393182603115734e-08,
+      "loss": 1.3447,
+      "step": 30649
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9328720693420065e-08,
+      "loss": 1.0635,
+      "step": 30650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.926436599249515e-08,
+      "loss": 0.6953,
+      "step": 30651
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.920011850103043e-08,
+      "loss": 1.0264,
+      "step": 30652
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9135978219717577e-08,
+      "loss": 0.7754,
+      "step": 30653
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.907194514924493e-08,
+      "loss": 1.0908,
+      "step": 30654
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9008019290299718e-08,
+      "loss": 0.9307,
+      "step": 30655
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8944200643570275e-08,
+      "loss": 1.3096,
+      "step": 30656
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8880489209739395e-08,
+      "loss": 0.9521,
+      "step": 30657
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.88168849894943e-08,
+      "loss": 1.0654,
+      "step": 30658
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8753387983515558e-08,
+      "loss": 1.0557,
+      "step": 30659
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8689998192488178e-08,
+      "loss": 0.9287,
+      "step": 30660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8626715617090508e-08,
+      "loss": 0.833,
+      "step": 30661
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.856354025800422e-08,
+      "loss": 0.96,
+      "step": 30662
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8500472115905443e-08,
+      "loss": 0.5159,
+      "step": 30663
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8437511191474745e-08,
+      "loss": 0.9453,
+      "step": 30664
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8374657485386027e-08,
+      "loss": 1.1328,
+      "step": 30665
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8311910998315417e-08,
+      "loss": 0.834,
+      "step": 30666
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8249271730936825e-08,
+      "loss": 1.2754,
+      "step": 30667
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8186739683923037e-08,
+      "loss": 0.8896,
+      "step": 30668
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8124314857945747e-08,
+      "loss": 0.7734,
+      "step": 30669
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8061997253675524e-08,
+      "loss": 1.1055,
+      "step": 30670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.799978687178183e-08,
+      "loss": 0.6401,
+      "step": 30671
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.793768371293303e-08,
+      "loss": 0.8994,
+      "step": 30672
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7875687777795247e-08,
+      "loss": 0.7998,
+      "step": 30673
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.781379906703573e-08,
+      "loss": 0.7466,
+      "step": 30674
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7752017581319504e-08,
+      "loss": 1.2861,
+      "step": 30675
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7690343321308255e-08,
+      "loss": 0.9644,
+      "step": 30676
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7628776287665905e-08,
+      "loss": 1.0908,
+      "step": 30677
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.756731648105414e-08,
+      "loss": 0.8662,
+      "step": 30678
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7505963902131328e-08,
+      "loss": 1.0117,
+      "step": 30679
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7444718551558048e-08,
+      "loss": 1.0029,
+      "step": 30680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7383580429992663e-08,
+      "loss": 1.3516,
+      "step": 30681
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7322549538091316e-08,
+      "loss": 1.2734,
+      "step": 30682
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7261625876507926e-08,
+      "loss": 0.7886,
+      "step": 30683
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7200809445898636e-08,
+      "loss": 0.9502,
+      "step": 30684
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.714010024691737e-08,
+      "loss": 0.999,
+      "step": 30685
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7079498280213603e-08,
+      "loss": 0.7285,
+      "step": 30686
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.701900354644126e-08,
+      "loss": 0.9736,
+      "step": 30687
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.695861604624871e-08,
+      "loss": 0.7759,
+      "step": 30688
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6898335780284323e-08,
+      "loss": 0.9209,
+      "step": 30689
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6838162749196473e-08,
+      "loss": 0.8096,
+      "step": 30690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6778096953630195e-08,
+      "loss": 1.0107,
+      "step": 30691
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6718138394232753e-08,
+      "loss": 0.9346,
+      "step": 30692
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.665828707164585e-08,
+      "loss": 0.7344,
+      "step": 30693
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6598542986514533e-08,
+      "loss": 0.8477,
+      "step": 30694
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6538906139478285e-08,
+      "loss": 1.0278,
+      "step": 30695
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6479376531179925e-08,
+      "loss": 0.8545,
+      "step": 30696
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6419954162258944e-08,
+      "loss": 0.8311,
+      "step": 30697
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.636063903335039e-08,
+      "loss": 1.2871,
+      "step": 30698
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.630143114509597e-08,
+      "loss": 1.2314,
+      "step": 30699
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.62423304981274e-08,
+      "loss": 0.8281,
+      "step": 30700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6183337093083063e-08,
+      "loss": 0.8628,
+      "step": 30701
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6124450930593562e-08,
+      "loss": 0.8115,
+      "step": 30702
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.606567201129394e-08,
+      "loss": 0.8018,
+      "step": 30703
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6007000335814814e-08,
+      "loss": 1.2588,
+      "step": 30704
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5948435904785677e-08,
+      "loss": 0.8442,
+      "step": 30705
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5889978718836018e-08,
+      "loss": 1.127,
+      "step": 30706
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5831628778594233e-08,
+      "loss": 1.083,
+      "step": 30707
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5773386084686482e-08,
+      "loss": 0.8838,
+      "step": 30708
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5715250637738933e-08,
+      "loss": 1.2988,
+      "step": 30709
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.565722243837553e-08,
+      "loss": 0.7327,
+      "step": 30710
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5599301487219108e-08,
+      "loss": 0.7993,
+      "step": 30711
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5541487784893617e-08,
+      "loss": 0.8027,
+      "step": 30712
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5483781332018556e-08,
+      "loss": 1.3301,
+      "step": 30713
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.542618212921454e-08,
+      "loss": 1.144,
+      "step": 30714
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5368690177099965e-08,
+      "loss": 1.0,
+      "step": 30715
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5311305476292114e-08,
+      "loss": 0.8828,
+      "step": 30716
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5254028027407165e-08,
+      "loss": 1.0225,
+      "step": 30717
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5196857831061285e-08,
+      "loss": 0.7542,
+      "step": 30718
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.513979488786843e-08,
+      "loss": 0.9453,
+      "step": 30719
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5082839198440336e-08,
+      "loss": 1.2021,
+      "step": 30720
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.502599076339095e-08,
+      "loss": 1.1582,
+      "step": 30721
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.496924958332868e-08,
+      "loss": 0.8115,
+      "step": 30722
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.491261565886526e-08,
+      "loss": 1.043,
+      "step": 30723
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4856088990606866e-08,
+      "loss": 0.7715,
+      "step": 30724
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4799669579161902e-08,
+      "loss": 1.1719,
+      "step": 30725
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4743357425136551e-08,
+      "loss": 0.8906,
+      "step": 30726
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4687152529134774e-08,
+      "loss": 1.1113,
+      "step": 30727
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4631054891760531e-08,
+      "loss": 1.3389,
+      "step": 30728
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4575064513617787e-08,
+      "loss": 0.8145,
+      "step": 30729
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4519181395306058e-08,
+      "loss": 1.0312,
+      "step": 30730
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4463405537425979e-08,
+      "loss": 0.8579,
+      "step": 30731
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4407736940578177e-08,
+      "loss": 0.9004,
+      "step": 30732
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4352175605358843e-08,
+      "loss": 1.2402,
+      "step": 30733
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4296721532366387e-08,
+      "loss": 1.1387,
+      "step": 30734
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4241374722194778e-08,
+      "loss": 1.1377,
+      "step": 30735
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4186135175439097e-08,
+      "loss": 1.0498,
+      "step": 30736
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4131002892694424e-08,
+      "loss": 0.8818,
+      "step": 30737
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4075977874550284e-08,
+      "loss": 0.9453,
+      "step": 30738
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4021060121599538e-08,
+      "loss": 0.959,
+      "step": 30739
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3966249634430606e-08,
+      "loss": 0.8613,
+      "step": 30740
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3911546413634125e-08,
+      "loss": 1.0713,
+      "step": 30741
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3856950459796293e-08,
+      "loss": 0.9727,
+      "step": 30742
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3802461773504416e-08,
+      "loss": 0.5476,
+      "step": 30743
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3748080355343585e-08,
+      "loss": 0.8076,
+      "step": 30744
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3693806205897776e-08,
+      "loss": 0.9263,
+      "step": 30745
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3639639325748744e-08,
+      "loss": 0.9897,
+      "step": 30746
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3585579715480468e-08,
+      "loss": 1.2734,
+      "step": 30747
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3531627375672485e-08,
+      "loss": 0.7656,
+      "step": 30748
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3477782306905441e-08,
+      "loss": 1.0039,
+      "step": 30749
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.342404450975554e-08,
+      "loss": 0.8809,
+      "step": 30750
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3370413984802321e-08,
+      "loss": 0.9482,
+      "step": 30751
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3316890732619769e-08,
+      "loss": 0.9258,
+      "step": 30752
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.32634747537852e-08,
+      "loss": 0.9844,
+      "step": 30753
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3210166048870377e-08,
+      "loss": 0.9375,
+      "step": 30754
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3156964618448176e-08,
+      "loss": 0.7168,
+      "step": 30755
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3103870463091472e-08,
+      "loss": 0.6909,
+      "step": 30756
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3050883583368701e-08,
+      "loss": 0.6853,
+      "step": 30757
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2998003979850515e-08,
+      "loss": 1.1318,
+      "step": 30758
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.294523165310424e-08,
+      "loss": 0.8735,
+      "step": 30759
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2892566603697199e-08,
+      "loss": 0.8218,
+      "step": 30760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2840008832194495e-08,
+      "loss": 0.7905,
+      "step": 30761
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2787558339161233e-08,
+      "loss": 0.957,
+      "step": 30762
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2735215125160293e-08,
+      "loss": 0.8188,
+      "step": 30763
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2682979190754563e-08,
+      "loss": 0.646,
+      "step": 30764
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2630850536503592e-08,
+      "loss": 0.8984,
+      "step": 30765
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2578829162970264e-08,
+      "loss": 0.8682,
+      "step": 30766
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2526915070709688e-08,
+      "loss": 1.5488,
+      "step": 30767
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2475108260282531e-08,
+      "loss": 0.8545,
+      "step": 30768
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2423408732243902e-08,
+      "loss": 0.7202,
+      "step": 30769
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2371816487150023e-08,
+      "loss": 0.8896,
+      "step": 30770
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2320331525553787e-08,
+      "loss": 0.6187,
+      "step": 30771
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2268953848009191e-08,
+      "loss": 0.6904,
+      "step": 30772
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.221768345506802e-08,
+      "loss": 1.1162,
+      "step": 30773
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.216652034728094e-08,
+      "loss": 1.1396,
+      "step": 30774
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2115464525196408e-08,
+      "loss": 1.2461,
+      "step": 30775
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2064515989365088e-08,
+      "loss": 0.8262,
+      "step": 30776
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2013674740332104e-08,
+      "loss": 0.959,
+      "step": 30777
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1962940778644794e-08,
+      "loss": 1.1426,
+      "step": 30778
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1912314104847166e-08,
+      "loss": 0.7871,
+      "step": 30779
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1861794719484343e-08,
+      "loss": 0.8936,
+      "step": 30780
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.181138262309811e-08,
+      "loss": 1.2383,
+      "step": 30781
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.176107781623026e-08,
+      "loss": 0.6025,
+      "step": 30782
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1710880299421468e-08,
+      "loss": 0.7617,
+      "step": 30783
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1660790073210193e-08,
+      "loss": 0.9922,
+      "step": 30784
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1610807138134894e-08,
+      "loss": 0.9463,
+      "step": 30785
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1560931494732919e-08,
+      "loss": 0.957,
+      "step": 30786
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1511163143538285e-08,
+      "loss": 1.0029,
+      "step": 30787
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1461502085087228e-08,
+      "loss": 1.0107,
+      "step": 30788
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1411948319913768e-08,
+      "loss": 0.7681,
+      "step": 30789
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.136250184854859e-08,
+      "loss": 0.8154,
+      "step": 30790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1313162671522382e-08,
+      "loss": 1.0645,
+      "step": 30791
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.126393078936805e-08,
+      "loss": 0.9316,
+      "step": 30792
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1214806202610729e-08,
+      "loss": 1.2002,
+      "step": 30793
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1165788911781106e-08,
+      "loss": 0.8291,
+      "step": 30794
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1116878917404317e-08,
+      "loss": 0.9395,
+      "step": 30795
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1068076220005497e-08,
+      "loss": 1.043,
+      "step": 30796
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1019380820109781e-08,
+      "loss": 1.0322,
+      "step": 30797
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0970792718240086e-08,
+      "loss": 1.1025,
+      "step": 30798
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0922311914917106e-08,
+      "loss": 1.0889,
+      "step": 30799
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0873938410662643e-08,
+      "loss": 0.7568,
+      "step": 30800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0825672205996285e-08,
+      "loss": 0.9922,
+      "step": 30801
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0777513301437614e-08,
+      "loss": 1.0879,
+      "step": 30802
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0729461697500665e-08,
+      "loss": 1.1113,
+      "step": 30803
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0681517394705021e-08,
+      "loss": 0.8398,
+      "step": 30804
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0633680393563605e-08,
+      "loss": 0.8525,
+      "step": 30805
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0585950694591563e-08,
+      "loss": 1.1045,
+      "step": 30806
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0538328298300703e-08,
+      "loss": 1.0566,
+      "step": 30807
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0490813205202843e-08,
+      "loss": 0.7246,
+      "step": 30808
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0443405415808684e-08,
+      "loss": 0.8306,
+      "step": 30809
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0396104930626704e-08,
+      "loss": 1.0322,
+      "step": 30810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0348911750165391e-08,
+      "loss": 0.9639,
+      "step": 30811
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0301825874931004e-08,
+      "loss": 1.5059,
+      "step": 30812
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0254847305432025e-08,
+      "loss": 1.1504,
+      "step": 30813
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0207976042169165e-08,
+      "loss": 1.0254,
+      "step": 30814
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0161212085648687e-08,
+      "loss": 0.886,
+      "step": 30815
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0114555436372409e-08,
+      "loss": 0.9771,
+      "step": 30816
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0068006094839933e-08,
+      "loss": 1.0889,
+      "step": 30817
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0021564061554189e-08,
+      "loss": 1.2734,
+      "step": 30818
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.975229337010339e-09,
+      "loss": 0.7334,
+      "step": 30819
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.929001921709092e-09,
+      "loss": 0.748,
+      "step": 30820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.882881816146716e-09,
+      "loss": 0.8696,
+      "step": 30821
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.836869020817263e-09,
+      "loss": 0.7959,
+      "step": 30822
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.790963536214781e-09,
+      "loss": 1.001,
+      "step": 30823
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.74516536283443e-09,
+      "loss": 0.8149,
+      "step": 30824
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.699474501166927e-09,
+      "loss": 0.7556,
+      "step": 30825
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.653890951701884e-09,
+      "loss": 0.7056,
+      "step": 30826
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.608414714931125e-09,
+      "loss": 0.7578,
+      "step": 30827
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.56304579134204e-09,
+      "loss": 1.0342,
+      "step": 30828
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.517784181422018e-09,
+      "loss": 0.9199,
+      "step": 30829
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.472629885657336e-09,
+      "loss": 0.8159,
+      "step": 30830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.427582904533161e-09,
+      "loss": 0.9951,
+      "step": 30831
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.382643238533551e-09,
+      "loss": 0.8418,
+      "step": 30832
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.337810888140343e-09,
+      "loss": 0.8442,
+      "step": 30833
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.293085853835371e-09,
+      "loss": 0.6973,
+      "step": 30834
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.248468136099364e-09,
+      "loss": 0.8335,
+      "step": 30835
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.203957735411938e-09,
+      "loss": 0.8838,
+      "step": 30836
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.15955465225049e-09,
+      "loss": 0.6169,
+      "step": 30837
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.115258887093525e-09,
+      "loss": 0.7314,
+      "step": 30838
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.071070440413998e-09,
+      "loss": 1.1309,
+      "step": 30839
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.026989312689305e-09,
+      "loss": 0.6335,
+      "step": 30840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.98301550439129e-09,
+      "loss": 1.0229,
+      "step": 30841
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.93914901599291e-09,
+      "loss": 0.9531,
+      "step": 30842
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.895389847966006e-09,
+      "loss": 1.2012,
+      "step": 30843
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.851738000779098e-09,
+      "loss": 1.0596,
+      "step": 30844
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.808193474902916e-09,
+      "loss": 1.1396,
+      "step": 30845
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.764756270803753e-09,
+      "loss": 1.0518,
+      "step": 30846
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.721426388947907e-09,
+      "loss": 0.9795,
+      "step": 30847
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.678203829802779e-09,
+      "loss": 0.9521,
+      "step": 30848
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.635088593830221e-09,
+      "loss": 1.0493,
+      "step": 30849
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.592080681495418e-09,
+      "loss": 0.6797,
+      "step": 30850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.549180093259112e-09,
+      "loss": 0.9102,
+      "step": 30851
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.506386829583158e-09,
+      "loss": 0.772,
+      "step": 30852
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.463700890926075e-09,
+      "loss": 0.8057,
+      "step": 30853
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.421122277746385e-09,
+      "loss": 0.7852,
+      "step": 30854
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.378650990502613e-09,
+      "loss": 1.3018,
+      "step": 30855
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.336287029649948e-09,
+      "loss": 1.2002,
+      "step": 30856
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.294030395643582e-09,
+      "loss": 1.2354,
+      "step": 30857
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.251881088938706e-09,
+      "loss": 0.9233,
+      "step": 30858
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.209839109986074e-09,
+      "loss": 0.8174,
+      "step": 30859
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.167904459238652e-09,
+      "loss": 0.9414,
+      "step": 30860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.126077137146082e-09,
+      "loss": 1.0488,
+      "step": 30861
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.084357144159116e-09,
+      "loss": 1.3867,
+      "step": 30862
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.042744480724062e-09,
+      "loss": 0.8521,
+      "step": 30863
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.001239147288342e-09,
+      "loss": 1.1953,
+      "step": 30864
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.959841144298264e-09,
+      "loss": 0.6748,
+      "step": 30865
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.918550472199027e-09,
+      "loss": 1.1006,
+      "step": 30866
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.8773671314325e-09,
+      "loss": 0.6702,
+      "step": 30867
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.836291122442773e-09,
+      "loss": 0.8716,
+      "step": 30868
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.795322445669495e-09,
+      "loss": 0.6304,
+      "step": 30869
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.754461101554533e-09,
+      "loss": 0.7729,
+      "step": 30870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.713707090535316e-09,
+      "loss": 1.2676,
+      "step": 30871
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.673060413050382e-09,
+      "loss": 0.9434,
+      "step": 30872
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.632521069536047e-09,
+      "loss": 0.5703,
+      "step": 30873
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.59208906042752e-09,
+      "loss": 0.6782,
+      "step": 30874
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.551764386160009e-09,
+      "loss": 0.9727,
+      "step": 30875
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.511547047166501e-09,
+      "loss": 0.6187,
+      "step": 30876
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.47143704387776e-09,
+      "loss": 1.1396,
+      "step": 30877
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.4314343767256654e-09,
+      "loss": 0.7344,
+      "step": 30878
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.391539046139873e-09,
+      "loss": 1.2031,
+      "step": 30879
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.3517510525478175e-09,
+      "loss": 1.0918,
+      "step": 30880
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.312070396379156e-09,
+      "loss": 1.3789,
+      "step": 30881
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.272497078057994e-09,
+      "loss": 1.0229,
+      "step": 30882
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.233031098010656e-09,
+      "loss": 1.0352,
+      "step": 30883
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.193672456660139e-09,
+      "loss": 1.27,
+      "step": 30884
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.154421154430547e-09,
+      "loss": 1.3896,
+      "step": 30885
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.115277191741543e-09,
+      "loss": 0.689,
+      "step": 30886
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.076240569015013e-09,
+      "loss": 1.123,
+      "step": 30887
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.0373112866706226e-09,
+      "loss": 0.917,
+      "step": 30888
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.998489345124704e-09,
+      "loss": 0.873,
+      "step": 30889
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.959774744796921e-09,
+      "loss": 0.8242,
+      "step": 30890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.921167486100278e-09,
+      "loss": 0.7832,
+      "step": 30891
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.882667569452217e-09,
+      "loss": 1.1914,
+      "step": 30892
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.844274995263522e-09,
+      "loss": 0.6758,
+      "step": 30893
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.805989763949417e-09,
+      "loss": 1.498,
+      "step": 30894
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.767811875919572e-09,
+      "loss": 1.2412,
+      "step": 30895
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.7297413315836616e-09,
+      "loss": 1.0518,
+      "step": 30896
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.691778131350246e-09,
+      "loss": 0.7852,
+      "step": 30897
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.653922275630109e-09,
+      "loss": 0.6648,
+      "step": 30898
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.6161737648262615e-09,
+      "loss": 1.2236,
+      "step": 30899
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.578532599346155e-09,
+      "loss": 0.9746,
+      "step": 30900
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.54099877959502e-09,
+      "loss": 0.7217,
+      "step": 30901
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.503572305973649e-09,
+      "loss": 0.8042,
+      "step": 30902
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.466253178885051e-09,
+      "loss": 1.3203,
+      "step": 30903
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.429041398730018e-09,
+      "loss": 0.9058,
+      "step": 30904
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.39193696590934e-09,
+      "loss": 0.6255,
+      "step": 30905
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.354939880820476e-09,
+      "loss": 0.9824,
+      "step": 30906
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.318050143859777e-09,
+      "loss": 1.248,
+      "step": 30907
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.281267755425813e-09,
+      "loss": 1.1719,
+      "step": 30908
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.2445927159127115e-09,
+      "loss": 1.3945,
+      "step": 30909
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.208025025714603e-09,
+      "loss": 0.8291,
+      "step": 30910
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.171564685223397e-09,
+      "loss": 1.0039,
+      "step": 30911
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.13521169483211e-09,
+      "loss": 0.8945,
+      "step": 30912
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.098966054930433e-09,
+      "loss": 1.0166,
+      "step": 30913
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.062827765906942e-09,
+      "loss": 0.749,
+      "step": 30914
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.026796828151327e-09,
+      "loss": 0.9551,
+      "step": 30915
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.990873242049944e-09,
+      "loss": 0.9219,
+      "step": 30916
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.955057007988041e-09,
+      "loss": 1.0781,
+      "step": 30917
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.919348126351976e-09,
+      "loss": 0.9873,
+      "step": 30918
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.8837465975236655e-09,
+      "loss": 1.3232,
+      "step": 30919
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.848252421886136e-09,
+      "loss": 0.6484,
+      "step": 30920
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.812865599821305e-09,
+      "loss": 0.8965,
+      "step": 30921
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.777586131707758e-09,
+      "loss": 1.0449,
+      "step": 30922
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.742414017926301e-09,
+      "loss": 0.8105,
+      "step": 30923
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.707349258853301e-09,
+      "loss": 1.1631,
+      "step": 30924
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.672391854866233e-09,
+      "loss": 1.0059,
+      "step": 30925
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.637541806340352e-09,
+      "loss": 1.1206,
+      "step": 30926
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.602799113649804e-09,
+      "loss": 0.9517,
+      "step": 30927
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.568163777167623e-09,
+      "loss": 0.6743,
+      "step": 30928
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.533635797265735e-09,
+      "loss": 1.1182,
+      "step": 30929
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.499215174317174e-09,
+      "loss": 0.8896,
+      "step": 30930
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.464901908688314e-09,
+      "loss": 0.9277,
+      "step": 30931
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.430696000749969e-09,
+      "loss": 0.9893,
+      "step": 30932
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3965974508685125e-09,
+      "loss": 1.0674,
+      "step": 30933
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.362606259411429e-09,
+      "loss": 1.002,
+      "step": 30934
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.328722426742872e-09,
+      "loss": 1.082,
+      "step": 30935
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.2949459532269934e-09,
+      "loss": 0.7998,
+      "step": 30936
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.2612768392257265e-09,
+      "loss": 0.8042,
+      "step": 30937
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.227715085103224e-09,
+      "loss": 0.8633,
+      "step": 30938
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.194260691216979e-09,
+      "loss": 0.9146,
+      "step": 30939
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.160913657928923e-09,
+      "loss": 0.7612,
+      "step": 30940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.127673985595438e-09,
+      "loss": 1.1709,
+      "step": 30941
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.094541674574016e-09,
+      "loss": 1.0947,
+      "step": 30942
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.061516725221039e-09,
+      "loss": 1.125,
+      "step": 30943
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.028599137891777e-09,
+      "loss": 0.9795,
+      "step": 30944
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.995788912938171e-09,
+      "loss": 1.1406,
+      "step": 30945
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.9630860507132725e-09,
+      "loss": 0.9248,
+      "step": 30946
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.930490551569023e-09,
+      "loss": 0.7842,
+      "step": 30947
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.898002415855141e-09,
+      "loss": 1.1274,
+      "step": 30948
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.865621643921348e-09,
+      "loss": 1.1719,
+      "step": 30949
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8333482361140325e-09,
+      "loss": 1.0312,
+      "step": 30950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8011821927795856e-09,
+      "loss": 1.0879,
+      "step": 30951
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.769123514265506e-09,
+      "loss": 1.4014,
+      "step": 30952
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.737172200915963e-09,
+      "loss": 1.4062,
+      "step": 30953
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.705328253071795e-09,
+      "loss": 0.7427,
+      "step": 30954
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6735916710771715e-09,
+      "loss": 1.1934,
+      "step": 30955
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6419624552729306e-09,
+      "loss": 0.9746,
+      "step": 30956
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.61044060599769e-09,
+      "loss": 0.9785,
+      "step": 30957
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5790261235911795e-09,
+      "loss": 1.1221,
+      "step": 30958
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5477190083897945e-09,
+      "loss": 0.9697,
+      "step": 30959
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5165192607310445e-09,
+      "loss": 0.7969,
+      "step": 30960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.485426880947996e-09,
+      "loss": 1.0127,
+      "step": 30961
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.454441869377047e-09,
+      "loss": 1.2061,
+      "step": 30962
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.423564226350152e-09,
+      "loss": 0.7935,
+      "step": 30963
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.39279395219927e-09,
+      "loss": 0.8281,
+      "step": 30964
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.362131047253027e-09,
+      "loss": 1.3984,
+      "step": 30965
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3315755118433775e-09,
+      "loss": 1.1338,
+      "step": 30966
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.30112734629673e-09,
+      "loss": 1.001,
+      "step": 30967
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2707865509417075e-09,
+      "loss": 0.9553,
+      "step": 30968
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.240553126102498e-09,
+      "loss": 1.085,
+      "step": 30969
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.210427072105505e-09,
+      "loss": 1.0322,
+      "step": 30970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1804083892726945e-09,
+      "loss": 1.0059,
+      "step": 30971
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.150497077928251e-09,
+      "loss": 0.7246,
+      "step": 30972
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.120693138391918e-09,
+      "loss": 0.7583,
+      "step": 30973
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.090996570984551e-09,
+      "loss": 1.1562,
+      "step": 30974
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0614073760258946e-09,
+      "loss": 1.0674,
+      "step": 30975
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.031925553832361e-09,
+      "loss": 0.854,
+      "step": 30976
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0025511047214745e-09,
+      "loss": 1.1982,
+      "step": 30977
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.973284029007429e-09,
+      "loss": 0.8716,
+      "step": 30978
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9441243270066376e-09,
+      "loss": 0.8267,
+      "step": 30979
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9150719990321826e-09,
+      "loss": 1.1021,
+      "step": 30980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8861270453938175e-09,
+      "loss": 0.707,
+      "step": 30981
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8572894664046234e-09,
+      "loss": 1.2236,
+      "step": 30982
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.828559262374354e-09,
+      "loss": 0.9004,
+      "step": 30983
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.799936433610541e-09,
+      "loss": 0.7305,
+      "step": 30984
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.771420980420715e-09,
+      "loss": 0.9028,
+      "step": 30985
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7430129031124085e-09,
+      "loss": 1.2217,
+      "step": 30986
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7147122019887128e-09,
+      "loss": 0.7739,
+      "step": 30987
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.686518877356049e-09,
+      "loss": 0.7393,
+      "step": 30988
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6584329295152876e-09,
+      "loss": 0.792,
+      "step": 30989
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6304543587695196e-09,
+      "loss": 1.1396,
+      "step": 30990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6025831654173946e-09,
+      "loss": 0.718,
+      "step": 30991
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.574819349760894e-09,
+      "loss": 1.3037,
+      "step": 30992
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.547162912096447e-09,
+      "loss": 0.9873,
+      "step": 30993
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.519613852721593e-09,
+      "loss": 0.8867,
+      "step": 30994
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4921721719316515e-09,
+      "loss": 1.2119,
+      "step": 30995
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4648378700219422e-09,
+      "loss": 0.9014,
+      "step": 30996
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.437610947286674e-09,
+      "loss": 0.6313,
+      "step": 30997
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.410491404017835e-09,
+      "loss": 0.7207,
+      "step": 30998
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3834792405063043e-09,
+      "loss": 1.0234,
+      "step": 30999
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3565744570429606e-09,
+      "loss": 0.7432,
+      "step": 31000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.329777053916461e-09,
+      "loss": 0.9761,
+      "step": 31001
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3030870314143538e-09,
+      "loss": 0.7129,
+      "step": 31002
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.276504389824187e-09,
+      "loss": 1.2031,
+      "step": 31003
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2500291294301766e-09,
+      "loss": 0.8789,
+      "step": 31004
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2236612505176513e-09,
+      "loss": 1.416,
+      "step": 31005
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.197400753369717e-09,
+      "loss": 1.0322,
+      "step": 31006
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.171247638267261e-09,
+      "loss": 0.6421,
+      "step": 31007
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1452019054933892e-09,
+      "loss": 1.1279,
+      "step": 31008
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1192635553267682e-09,
+      "loss": 1.0693,
+      "step": 31009
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0934325880460636e-09,
+      "loss": 0.8262,
+      "step": 31010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0677090039288317e-09,
+      "loss": 0.7998,
+      "step": 31011
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0420928032515174e-09,
+      "loss": 1.0918,
+      "step": 31012
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.016583986288346e-09,
+      "loss": 1.2656,
+      "step": 31013
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9911825533146533e-09,
+      "loss": 1.0732,
+      "step": 31014
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.965888504602443e-09,
+      "loss": 0.6782,
+      "step": 31015
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9407018404237207e-09,
+      "loss": 1.1831,
+      "step": 31016
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9156225610482704e-09,
+      "loss": 0.8877,
+      "step": 31017
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8906506667469857e-09,
+      "loss": 1.1318,
+      "step": 31018
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8657861577874314e-09,
+      "loss": 0.6997,
+      "step": 31019
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.84102903443606e-09,
+      "loss": 0.917,
+      "step": 31020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.816379296959326e-09,
+      "loss": 0.7056,
+      "step": 31021
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.791836945621462e-09,
+      "loss": 0.5371,
+      "step": 31022
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7674019806878118e-09,
+      "loss": 0.7539,
+      "step": 31023
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.743074402419277e-09,
+      "loss": 1.0264,
+      "step": 31024
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7188542110767604e-09,
+      "loss": 1.3828,
+      "step": 31025
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6947414069222744e-09,
+      "loss": 1.0957,
+      "step": 31026
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.670735990212281e-09,
+      "loss": 1.1748,
+      "step": 31027
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6468379612076824e-09,
+      "loss": 0.8604,
+      "step": 31028
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6230473201627193e-09,
+      "loss": 1.1289,
+      "step": 31029
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.599364067333854e-09,
+      "loss": 0.9297,
+      "step": 31030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.575788202976437e-09,
+      "loss": 0.9492,
+      "step": 31031
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.552319727341379e-09,
+      "loss": 0.8867,
+      "step": 31032
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.528958640682921e-09,
+      "loss": 1.0432,
+      "step": 31033
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5057049432519744e-09,
+      "loss": 1.0293,
+      "step": 31034
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4825586352961174e-09,
+      "loss": 1.0342,
+      "step": 31035
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.459519717066261e-09,
+      "loss": 1.1445,
+      "step": 31036
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.436588188808875e-09,
+      "loss": 1.5078,
+      "step": 31037
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4137640507704284e-09,
+      "loss": 1.0742,
+      "step": 31038
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.391047303196281e-09,
+      "loss": 0.9429,
+      "step": 31039
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3684379463295716e-09,
+      "loss": 1.0498,
+      "step": 31040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3459359804145487e-09,
+      "loss": 0.8955,
+      "step": 31041
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3235414056921314e-09,
+      "loss": 1.0176,
+      "step": 31042
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3012542224032375e-09,
+      "loss": 0.6587,
+      "step": 31043
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2790744307865654e-09,
+      "loss": 0.8657,
+      "step": 31044
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.257002031080813e-09,
+      "loss": 0.981,
+      "step": 31045
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2350370235235674e-09,
+      "loss": 1.2988,
+      "step": 31046
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2131794083490865e-09,
+      "loss": 0.7568,
+      "step": 31047
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1914291857949575e-09,
+      "loss": 1.2051,
+      "step": 31048
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1697863560921073e-09,
+      "loss": 0.8086,
+      "step": 31049
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1482509194747923e-09,
+      "loss": 1.0796,
+      "step": 31050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.126822876172829e-09,
+      "loss": 0.8467,
+      "step": 31051
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1055022264182546e-09,
+      "loss": 0.7803,
+      "step": 31052
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0842889704386636e-09,
+      "loss": 0.9355,
+      "step": 31053
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.063183108461653e-09,
+      "loss": 1.2471,
+      "step": 31054
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0421846407148173e-09,
+      "loss": 1.1367,
+      "step": 31055
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0212935674235323e-09,
+      "loss": 1.2646,
+      "step": 31056
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0005098888131735e-09,
+      "loss": 0.8579,
+      "step": 31057
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.979833605104675e-09,
+      "loss": 1.2969,
+      "step": 31058
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9592647165223024e-09,
+      "loss": 1.0098,
+      "step": 31059
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.938803223285879e-09,
+      "loss": 1.1211,
+      "step": 31060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9184491256152295e-09,
+      "loss": 1.1982,
+      "step": 31061
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8982024237290675e-09,
+      "loss": 1.1885,
+      "step": 31062
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.878063117844997e-09,
+      "loss": 0.7896,
+      "step": 31063
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8580312081795116e-09,
+      "loss": 0.9258,
+      "step": 31064
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8381066949479942e-09,
+      "loss": 0.9395,
+      "step": 31065
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.818289578363608e-09,
+      "loss": 1.1777,
+      "step": 31066
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7985798586406257e-09,
+      "loss": 0.8408,
+      "step": 31067
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7789775359888794e-09,
+      "loss": 1.1338,
+      "step": 31068
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7594826106204221e-09,
+      "loss": 0.8972,
+      "step": 31069
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7400950827439755e-09,
+      "loss": 1.1182,
+      "step": 31070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7208149525682615e-09,
+      "loss": 0.8994,
+      "step": 31071
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7016422202997817e-09,
+      "loss": 0.8584,
+      "step": 31072
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6825768861450377e-09,
+      "loss": 1.0068,
+      "step": 31073
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.66361895030831e-09,
+      "loss": 1.1904,
+      "step": 31074
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6447684129938802e-09,
+      "loss": 0.9468,
+      "step": 31075
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6260252744026982e-09,
+      "loss": 0.8711,
+      "step": 31076
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6073895347379354e-09,
+      "loss": 0.8584,
+      "step": 31077
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5888611941994313e-09,
+      "loss": 0.7715,
+      "step": 31078
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5704402529859163e-09,
+      "loss": 1.002,
+      "step": 31079
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5521267112950101e-09,
+      "loss": 0.8848,
+      "step": 31080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5339205693232218e-09,
+      "loss": 0.7861,
+      "step": 31081
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5158218272670611e-09,
+      "loss": 1.1152,
+      "step": 31082
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.497830485320817e-09,
+      "loss": 1.0073,
+      "step": 31083
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.479946543676558e-09,
+      "loss": 0.6885,
+      "step": 31084
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4621700025274633e-09,
+      "loss": 1.1416,
+      "step": 31085
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4445008620644906e-09,
+      "loss": 0.6895,
+      "step": 31086
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4269391224763784e-09,
+      "loss": 1.4502,
+      "step": 31087
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4094847839529745e-09,
+      "loss": 1.2021,
+      "step": 31088
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3921378466807966e-09,
+      "loss": 1.2051,
+      "step": 31089
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3748983108474722e-09,
+      "loss": 1.2002,
+      "step": 31090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3577661766372985e-09,
+      "loss": 0.8882,
+      "step": 31091
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340741444233462e-09,
+      "loss": 0.7246,
+      "step": 31092
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3238241138213703e-09,
+      "loss": 1.0322,
+      "step": 31093
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3070141855797692e-09,
+      "loss": 0.6885,
+      "step": 31094
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.290311659690735e-09,
+      "loss": 1.1768,
+      "step": 31095
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2737165363341242e-09,
+      "loss": 0.8486,
+      "step": 31096
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2572288156875722e-09,
+      "loss": 0.8252,
+      "step": 31097
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2408484979276047e-09,
+      "loss": 1.293,
+      "step": 31098
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2245755832318574e-09,
+      "loss": 1.1006,
+      "step": 31099
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2084100717735247e-09,
+      "loss": 0.7383,
+      "step": 31100
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.192351963726912e-09,
+      "loss": 1.2188,
+      "step": 31101
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.176401259264104e-09,
+      "loss": 0.9268,
+      "step": 31102
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1605579585571846e-09,
+      "loss": 1.1826,
+      "step": 31103
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1448220617749084e-09,
+      "loss": 0.5474,
+      "step": 31104
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1291935690882494e-09,
+      "loss": 1.1729,
+      "step": 31105
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1136724806637411e-09,
+      "loss": 0.8242,
+      "step": 31106
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0982587966690271e-09,
+      "loss": 1.1846,
+      "step": 31107
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0829525172684208e-09,
+      "loss": 1.1367,
+      "step": 31108
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0677536426273449e-09,
+      "loss": 1.3135,
+      "step": 31109
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0526621729090024e-09,
+      "loss": 1.2578,
+      "step": 31110
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0376781082743759e-09,
+      "loss": 1.0557,
+      "step": 31111
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0228014488866677e-09,
+      "loss": 0.9375,
+      "step": 31112
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0080321949035299e-09,
+      "loss": 0.8428,
+      "step": 31113
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.93370346483724e-10,
+      "loss": 0.7695,
+      "step": 31114
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.788159037871226e-10,
+      "loss": 1.2031,
+      "step": 31115
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.64368866966936e-10,
+      "loss": 1.3145,
+      "step": 31116
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.500292361797058e-10,
+      "loss": 1.3232,
+      "step": 31117
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.357970115808635e-10,
+      "loss": 1.0449,
+      "step": 31118
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.216721933202888e-10,
+      "loss": 0.896,
+      "step": 31119
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.076547815523029e-10,
+      "loss": 1.1836,
+      "step": 31120
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.937447764267859e-10,
+      "loss": 0.7939,
+      "step": 31121
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.799421780925077e-10,
+      "loss": 1.2178,
+      "step": 31122
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.662469866982382e-10,
+      "loss": 0.8496,
+      "step": 31123
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.526592023905267e-10,
+      "loss": 1.1426,
+      "step": 31124
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.391788253159227e-10,
+      "loss": 0.4893,
+      "step": 31125
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.258058556187554e-10,
+      "loss": 1.1064,
+      "step": 31126
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.125402934433535e-10,
+      "loss": 0.6021,
+      "step": 31127
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.993821389329359e-10,
+      "loss": 1.1465,
+      "step": 31128
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.863313922262805e-10,
+      "loss": 0.8379,
+      "step": 31129
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.733880534654959e-10,
+      "loss": 1.0713,
+      "step": 31130
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.605521227893598e-10,
+      "loss": 1.0566,
+      "step": 31131
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.478236003355399e-10,
+      "loss": 0.6577,
+      "step": 31132
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.352024862405938e-10,
+      "loss": 0.9521,
+      "step": 31133
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.226887806399685e-10,
+      "loss": 1.2349,
+      "step": 31134
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.102824836691113e-10,
+      "loss": 1.3037,
+      "step": 31135
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.979835954601389e-10,
+      "loss": 1.0859,
+      "step": 31136
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.857921161462777e-10,
+      "loss": 0.959,
+      "step": 31137
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.737080458574241e-10,
+      "loss": 1.1895,
+      "step": 31138
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.617313847234741e-10,
+      "loss": 0.7358,
+      "step": 31139
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.498621328732136e-10,
+      "loss": 0.9873,
+      "step": 31140
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.381002904343181e-10,
+      "loss": 0.6592,
+      "step": 31141
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.264458575333532e-10,
+      "loss": 0.853,
+      "step": 31142
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.14898834295774e-10,
+      "loss": 1.0635,
+      "step": 31143
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.034592208437051e-10,
+      "loss": 0.835,
+      "step": 31144
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.921270173026017e-10,
+      "loss": 0.7495,
+      "step": 31145
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.809022237923678e-10,
+      "loss": 0.8374,
+      "step": 31146
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.69784840435128e-10,
+      "loss": 1.0615,
+      "step": 31147
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.58774867348566e-10,
+      "loss": 0.6553,
+      "step": 31148
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.478723046514755e-10,
+      "loss": 1.2832,
+      "step": 31149
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.370771524615404e-10,
+      "loss": 1.1035,
+      "step": 31150
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.263894108953338e-10,
+      "loss": 1.1641,
+      "step": 31151
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.158090800660986e-10,
+      "loss": 0.8389,
+      "step": 31152
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.053361600881879e-10,
+      "loss": 0.9512,
+      "step": 31153
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.94970651073734e-10,
+      "loss": 0.9697,
+      "step": 31154
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.847125531348695e-10,
+      "loss": 1.2393,
+      "step": 31155
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.745618663803963e-10,
+      "loss": 1.1807,
+      "step": 31156
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.6451859092133676e-10,
+      "loss": 0.8438,
+      "step": 31157
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.545827268642722e-10,
+      "loss": 0.8413,
+      "step": 31158
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.44754274315784e-10,
+      "loss": 1.0234,
+      "step": 31159
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.350332333813434e-10,
+      "loss": 1.3066,
+      "step": 31160
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2541960416642157e-10,
+      "loss": 0.8164,
+      "step": 31161
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1591338677426927e-10,
+      "loss": 0.9707,
+      "step": 31162
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0651458130591683e-10,
+      "loss": 1.0068,
+      "step": 31163
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.9722318786350466e-10,
+      "loss": 0.8809,
+      "step": 31164
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.880392065447325e-10,
+      "loss": 0.8789,
+      "step": 31165
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.7896263745063057e-10,
+      "loss": 0.8242,
+      "step": 31166
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.699934806777883e-10,
+      "loss": 1.1406,
+      "step": 31167
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.611317363227951e-10,
+      "loss": 1.1436,
+      "step": 31168
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.523774044800199e-10,
+      "loss": 1.0762,
+      "step": 31169
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.437304852438317e-10,
+      "loss": 1.542,
+      "step": 31170
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.351909787085994e-10,
+      "loss": 0.9893,
+      "step": 31171
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.267588849631409e-10,
+      "loss": 0.9614,
+      "step": 31172
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.184342041007149e-10,
+      "loss": 0.6343,
+      "step": 31173
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1021693620902904e-10,
+      "loss": 1.125,
+      "step": 31174
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0210708137801137e-10,
+      "loss": 1.0225,
+      "step": 31175
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9410463969314907e-10,
+      "loss": 0.8682,
+      "step": 31176
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.862096112410395e-10,
+      "loss": 1.1816,
+      "step": 31177
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.784219961060597e-10,
+      "loss": 0.8247,
+      "step": 31178
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.707417943725865e-10,
+      "loss": 0.8208,
+      "step": 31179
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6316900612388673e-10,
+      "loss": 0.9268,
+      "step": 31180
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5570363143878617e-10,
+      "loss": 0.9067,
+      "step": 31181
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4834567039944135e-10,
+      "loss": 1.0752,
+      "step": 31182
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.410951230846781e-10,
+      "loss": 0.8389,
+      "step": 31183
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.33951989572212e-10,
+      "loss": 0.6694,
+      "step": 31184
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2691626993753823e-10,
+      "loss": 0.6436,
+      "step": 31185
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1998796425837244e-10,
+      "loss": 0.8086,
+      "step": 31186
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.131670726079893e-10,
+      "loss": 1.3438,
+      "step": 31187
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0645359505966356e-10,
+      "loss": 0.8789,
+      "step": 31188
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9984753168444948e-10,
+      "loss": 0.9883,
+      "step": 31189
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.933488825556218e-10,
+      "loss": 0.7549,
+      "step": 31190
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8695764774201432e-10,
+      "loss": 0.7441,
+      "step": 31191
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8067382731135063e-10,
+      "loss": 1.0234,
+      "step": 31192
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7449742133246462e-10,
+      "loss": 0.9854,
+      "step": 31193
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.684284298708594e-10,
+      "loss": 0.9268,
+      "step": 31194
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6246685299203812e-10,
+      "loss": 1.0547,
+      "step": 31195
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5661269076039377e-10,
+      "loss": 1.1123,
+      "step": 31196
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5086594323809878e-10,
+      "loss": 0.8022,
+      "step": 31197
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4522661048621544e-10,
+      "loss": 1.0635,
+      "step": 31198
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3969469256691625e-10,
+      "loss": 0.9424,
+      "step": 31199
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3427018954015325e-10,
+      "loss": 0.9287,
+      "step": 31200
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.289531014614376e-10,
+      "loss": 0.6455,
+      "step": 31201
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.237434283907213e-10,
+      "loss": 0.9033,
+      "step": 31202
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.186411703812951e-10,
+      "loss": 1.0859,
+      "step": 31203
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1364632749089055e-10,
+      "loss": 0.7173,
+      "step": 31204
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0875889977057796e-10,
+      "loss": 1.1025,
+      "step": 31205
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0397888727364802e-10,
+      "loss": 0.9487,
+      "step": 31206
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.930629005228121e-11,
+      "loss": 1.0234,
+      "step": 31207
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.474110815643755e-11,
+      "loss": 1.3047,
+      "step": 31208
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.028334163496688e-11,
+      "loss": 0.8535,
+      "step": 31209
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.593299053449855e-11,
+      "loss": 1.1064,
+      "step": 31210
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.169005490388238e-11,
+      "loss": 1.0586,
+      "step": 31211
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.755453478752728e-11,
+      "loss": 0.9248,
+      "step": 31212
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.352643022984219e-11,
+      "loss": 0.8477,
+      "step": 31213
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.960574127412578e-11,
+      "loss": 1.0088,
+      "step": 31214
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.579246796367677e-11,
+      "loss": 1.0615,
+      "step": 31215
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.208661033735297e-11,
+      "loss": 1.1982,
+      "step": 31216
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.848816843623262e-11,
+      "loss": 0.9336,
+      "step": 31217
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.499714229917352e-11,
+      "loss": 0.6655,
+      "step": 31218
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.161353196281305e-11,
+      "loss": 0.8623,
+      "step": 31219
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.833733746378855e-11,
+      "loss": 0.6833,
+      "step": 31220
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.5168558837627166e-11,
+      "loss": 1.1152,
+      "step": 31221
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.210719611874581e-11,
+      "loss": 0.7412,
+      "step": 31222
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.915324933934095e-11,
+      "loss": 0.8848,
+      "step": 31223
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.630671853049883e-11,
+      "loss": 0.915,
+      "step": 31224
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.356760372441592e-11,
+      "loss": 1.2061,
+      "step": 31225
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.09359049488478e-11,
+      "loss": 1.1016,
+      "step": 31226
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8411622233770474e-11,
+      "loss": 1.1689,
+      "step": 31227
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5994755604719092e-11,
+      "loss": 0.8291,
+      "step": 31228
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3685305088338994e-11,
+      "loss": 0.709,
+      "step": 31229
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1483270710165317e-11,
+      "loss": 1.0322,
+      "step": 31230
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9388652492402515e-11,
+      "loss": 0.6377,
+      "step": 31231
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.74014504594755e-11,
+      "loss": 1.0068,
+      "step": 31232
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.552166463025806e-11,
+      "loss": 1.1299,
+      "step": 31233
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3749295026954657e-11,
+      "loss": 0.6914,
+      "step": 31234
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2084341667328858e-11,
+      "loss": 0.7202,
+      "step": 31235
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0526804570254456e-11,
+      "loss": 0.9155,
+      "step": 31236
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.076683752384797e-12,
+      "loss": 0.7148,
+      "step": 31237
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.733979228152777e-12,
+      "loss": 0.7317,
+      "step": 31238
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.498691013101521e-12,
+      "loss": 0.9961,
+      "step": 31239
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.370819119443482e-12,
+      "loss": 0.9404,
+      "step": 31240
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.350363561611559e-12,
+      "loss": 0.6831,
+      "step": 31241
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.4373243473773134e-12,
+      "loss": 0.9346,
+      "step": 31242
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.631701488953198e-12,
+      "loss": 1.5,
+      "step": 31243
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.933494994110774e-12,
+      "loss": 0.4985,
+      "step": 31244
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3427048706216029e-12,
+      "loss": 1.0859,
+      "step": 31245
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.593311240367997e-13,
+      "loss": 1.2959,
+      "step": 31246
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.833737599074795e-13,
+      "loss": 1.1689,
+      "step": 31247
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1483278378475748e-13,
+      "loss": 0.7256,
+      "step": 31248
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.3708195668633615e-14,
+      "loss": 0.6709,
+      "step": 31249
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 0.9087,
+      "step": 31250
+    },
+    {
+      "epoch": 1.0,
+      "step": 31250,
+      "total_flos": 3.875429257153662e+19,
+      "train_loss": 1.0998086030273437,
+      "train_runtime": 101056.1494,
+      "train_samples_per_second": 4.948,
+      "train_steps_per_second": 0.309
+    }
+  ],
+  "max_steps": 31250,
+  "num_train_epochs": 1,
+  "total_flos": 3.875429257153662e+19,
+  "trial_name": null,
+  "trial_params": null
+}